What problem does it solve?
Much inbound contact is general: opening hours and fees, how do I, where is my request, what does this letter mean. In a bank it is also time critical: a customer abroad without a local SIM, a card that does not work, a payment that has not arrived. Human agents answer the same questions all day while callers wait in a queue, and the conversations that need care (a bereavement, a scam victim, a complaint) wait behind them.
Touch tone menus and first generation chatbots did not fix this. Menus route by the option a caller picks, and those chatbots matched keywords to a fixed list of answers, so anything outside the list ends in a dead end or a transfer. The shift is an agent that understands free speech and text, answers from the organization's approved knowledge, can look up the customer's own case, and knows exactly when a human must take over.
How does it work?
- Greet and understand. The agent identifies the intent and language from the customer's first words, on the phone or in chat, and discloses that it is AI.
- Answer general questions from approved knowledge. Fees, procedures, product terms and service status come from retrieval over the organization's own content, with a refusal when the content does not cover the question.
- Look up what is personal. After authentication it checks the status of the customer's request, order or case through read only APIs and explains it.
- Route deliberately. Regulated journeys (disputes, fraud reports, complaints) follow a defined path; vulnerability, strong emotion or repeated failure trigger a human.
- Hand over with context. The human receives the transcript, a summary, the authenticated identity and what was already tried, so the customer does not repeat anything.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Mainstream
- Channels
- Phone and voice, Web chat, Mobile app, WhatsApp, Social messaging
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Interactions handled | Not pooled | 400 to 3 billion Not pooled: up to 60,000 | 26plus 1 up to | 16 organization, 10 vendor |
| Containment rate | 55% | 25% to 91% Not pooled: up to 88.5% | 19plus 1 up to | 7 organization, 12 vendor |
| Customer satisfaction | 82.5% vendor reported | 73% to 89% | 3 | 3 vendor |
| Handling time reduction | 30% | 28.6% to 50% | 3 | 1 organization, 2 vendor |
| Response time reduction | 82% | 33% to 99.5% | 3 | 1 organization, 2 vendor |
| First contact resolution | Too few to pool | 60% | 1 | 1 organization |
Value drivers: Lower cost to serve, Customer experience, Inclusion and access, Employee productivity.
Indicative value
A retail bank contact center that receives 2 million contacts a year across phone and chat
USD 900,000 to USD 3.8 million
Human handled contact cost avoided per year
How this is calculated
Formula: contacts * generalShare * containment * costPerContact. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Inbound contacts per year contacts, contacts per year | 2,000,000 | 2,000,000 | The reference contact center. |
| Share of contacts that are general or routine generalShare, fraction of contacts | 0.5 | 0.7 | Editorial assumption, replace with your own contact reason report. |
| Share of those contacts the agent resolves without a human containment, fraction of general contacts | 0.3 | 0.45 | The low end is conservative; the high end sits above most benchmarks on this page (44 to 47% for Airbnb, Ingka, JetBlue and Vodafone Germany), while a few operators report higher containment (66% at Together Credit Union, 70% for Vodafone TOBi, 84.6% for Commonwealth Bank's self service messaging in May 2026), because early months run lower. |
| Cost of a human handled contact costPerContact, USD per contact | 3 | 6 | Editorial assumption for a blended chat and phone contact. Replace with your own fully loaded cost. |
What it leaves out: Gross avoided contact cost only. It leaves out the cost of the AI and the integrations, the value of shorter waiting times, the cost of customers who give up instead of being helped, and any capacity the organization chooses to reinvest in human service rather than save.
Who already uses it?
95 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Air Canada
Canada · Travel and hospitality · 2022
In November 2022, a customer used the AI chatbot on Air Canada's website to ask about the airline's bereavement fare policy. The chatbot told him he could buy a full price ticket and claim the fare difference back within 90 days, which contradicted the airline's actual policy that bereavement fares must be requested before the flight. When Air Canada refused the refund, the customer took the airline to the Civil Resolution Tribunal of British Columbia. The tribunal rejected Air Canada's argument that the chatbot was a separate legal entity responsible for its own actions, and found the airline liable for negligent misrepresentation.
No outcome disclosed.
Arab Bank
Jordan · Banking · 2026
In March 2026 Reflect, the digital banking app of Arab Bank, launched ReVA, an AI powered virtual assistant inside the app. Customers can type, speak or send voice messages in Arabic or English and get written or spoken replies. ReVA answers questions about Reflect services and promotions, shows account details and reviews spending, and also helps prospects who have not yet opened an account. The bank calls the voice to voice feature among the first of its kind in the region. No usage or containment figures are published.
No outcome disclosed.
Deutsche Telekom
Germany · Telecommunications · 2026
At its AI investor day in October 2026, Deutsche Telekom said that its "Frag Magenta" assistant, which the release calls a chatbot, took over about 2.6 million calls in customer service in the first half of 2026. The release presents this as part of its use of AI where large volumes of repetitive work arise, alongside AI that supports service staff during calls and takes over their documentation. It does not say how many of these calls ended without a human, which lines or countries are covered, or which technology runs the assistant.
- Interactions handled: about 2.6 million, first half of 2026
"Der Chatbot „Frag Magenta“ hat bereits im ersten Halbjahr 2026 rund 2,6 Millionen Anrufe im Kundenservice übernommen."
Claimed by: organization
Finnair
Finland · Travel and hospitality · 2026
Finnair relaunched its customer service assistant Sisu in spring 2026 as a generative AI assistant on its website and app, in 15 languages, answering from Finnair's website content and knowledge articles and from booking details the customer shares, and passing a conversation summary to a human agent when needed. Finnair says Sisu serves between 2,000 and 4,000 customers a day and cannot yet change bookings. A Salesforce customer story describes Finnair using Agentforce with Service Cloud, Amadeus booking data and loyalty data to answer customers and escalate with context; neither page says whether Sisu is the Agentforce agent. No measured outcome is recorded.
No outcome disclosed.
Garanti BBVA
Türkiye · Banking · 2026
Garanti BBVA runs Ugi, a smart assistant inside its mobile banking app that answers questions and carries out banking transactions through conversation. The bank reports that Ugi handled more than 60 million chats in 2024 and now averages more than 6.4 million interactions a month. After the bank added generative AI based on large language models, Ugi understands 90% of user requests, and it supports more than 300 end to end banking transactions.
- Interactions handled: at least 6.4 million, per month on average
"Garanti BBVA’s AI journey, launched in 2016, has evolved into Ugi, which today handles more than 6.4 million monthly interactions on average, has 1.6 million active users, and supports more than 300 end-to-end banking transactions."
Claimed by: organization - Accuracy: 90%, share of user requests understood after the generative AI upgrade
"With the integration of generative AI, Ugi can now understand 90% of user requests."
Claimed by: organization - Interactions handled: at least 60 million, chats in 2024
"In 2024, Garanti BBVA’s AI assistant, Ugi, handled over 60 million chats, helping more than 5 million customers resolve their doubts."
Claimed by: organization - Users served: at least 5 million, customers helped in 2024
"In 2024, Garanti BBVA’s AI assistant, Ugi, handled over 60 million chats, helping more than 5 million customers resolve their doubts."
Claimed by: organization
Ministry of Finance (United Arab Emirates)
United Arab Emirates · Government and public sector · 2026
The UAE Ministry of Finance reports that its call centre met every approved operational target in the first half of 2026 after it added generative AI. The AI covers a virtual assistant that answers customer enquiries around the clock, a Chat with Documents tool that answers questions from PDFs and user guides, sentiment analysis of voice tone and language, and speech to text transcripts that produce feedback for staff. The ministry credits the tools for the results but publishes no figure that isolates their effect: its first contact resolution rate of 97.11 percent in the first half of 2026 is slightly below the 97.30 percent it reported for the first half of 2025.
No outcome disclosed.
National Bank of Bahrain
Bahrain · Banking · 2026
National Bank of Bahrain Group, which comprises NBB and Bahrain Islamic Bank, runs its AI work under an enterprise programme called AI@NBB. The group lists an AI powered customer services virtual assistant among the applications already in use, next to cybersecurity response, internal audit and legal AI capabilities, and agentic policy and documentation management. A federated model of a central AI function, AI Champions and cross functional AI Squads governs delivery, and the group says Bahrain based ARRAY Innovation supports distinct areas of implementation. The announcement gives no volumes or outcomes for the assistant.
No outcome disclosed.
Airbnb
United States · Travel and hospitality · 2025
Airbnb runs an AI assistant as the first line of customer support for guests and hosts. It was expanded to all US users in 2025 and then rolled out to more countries and languages, reaching more than 50 languages by mid 2026. Airbnb reports the share of issues resolved without a human agent in each quarterly letter and links part of the fall in support cost per booking to the assistant; it plans an AI voice assistant.
- Containment rate: about 45%, Q2 2026, issues that begin with the AI assistant
"Nearly 45 percent of issues that begin with our AI assistant are now resolved without a human agent, up from Q1, while delivering much faster resolution times."
Claimed by: organization - Cost reduction: about 16%, Q2 2026 year on year, customer support cost per booking
"In Q2, our customer support related cost per booking declined approximately 16 percent year-over-year, driven in part by improvements to our AI assistant."
Claimed by: organization
Bank of America
United States · Banking · 2025
Erica, launched in 2018, is Bank of America's virtual financial assistant in its Mobile Banking app. Beyond answering questions it delivers proactive, personalized insights: BankAmeriDeals cash back deals based on the client's spending, where balances are trending over the next seven days and eligibility for the Preferred Rewards program. It also gives guidance on investment topics for Merrill clients and hands off to people by scheduling appointments.
The bank reports that clients have received and interacted with more than 1.7 billion of these insights, and that most users find the information they need, which it links to lower call center volume. Bank of America says Erica selects answers from a predefined set and does not use generative AI or large language models.
- Users served: about 50 million, since launch in 2018, as of August 2025
"assisting nearly 50 million users since launch, surpassing 3 billion client interactions, and now averaging more than 58 million interactions per month"
Claimed by: organization - Interactions handled: at least 3 billion, client interactions since launch in 2018, as of August 2025
"surpassing 3 billion client interactions"
Claimed by: organization
Bank of the Philippine Islands
Philippines · Banking · 2025
Bank of the Philippine Islands runs BEA Chat, a conversational AI assistant on its website and Facebook page that answers general inquiries around the clock, offers self service for common concerns, lets customers apply for products and track service requests, and escalates complex issues to live agents. It serves both logged in clients and guests, and the bank positions it as a channel for Filipinos working overseas who would otherwise need a local SIM card or international calls. No containment or cost outcome is disclosed.
No outcome disclosed.
BBVA México
Mexico · Banking · 2025
BBVA México eliminated touch tone (IVR) dialing in its telephone banking service and replaced it with Blue, a generative AI voice assistant. Instead of navigating more than 40 menu options, callers answer one open question in natural language and Blue routes them straight to the right option, using the identification already available from the BBVA Mx app. The bank also uses Blue as a sales assistant that resumes an abandoned digital sales conversation when a customer shows interest in an offer.
- Accuracy: at least 90%, from the start of the call
"Blue entiende en este momento más del 90% de las solicitudes desde el inicio de la llamada y utiliza la información de la App para identificar al cliente y personalizar la atención."
Claimed by: organization - Interactions handled: 63 million, calls to the contact centre per year, all now answered first by Blue instead of the IVR
"Cada año, BBVA México recibe 63 millones de llamadas en su centro de atención telefónica."
Claimed by: organization
Dubai Land Department (DLD)
United Arab Emirates · Government and public sector · 2025
At GITEX Global 2025 Dubai Land Department unveiled Malik, an omnichannel chatbot that answers real estate customers on WhatsApp, the official website and the smart application, around the clock by text or voice message. In its first phase Malik answers the most common inquiries. The DLD contact page now offers Malik, labeled as powered by OpenAI, next to a live chat with staff, and DLD service pages direct WhatsApp users to Malik. DLD plans to extend it to the top 10 real estate transactions, such as payments, registrations and data updates, and to connect it to a new unified customer relationship system built on Microsoft Dynamics 365.
No outcome disclosed.
E.ON Next
United Kingdom · Energy and utilities · 2025
E.ON Next, the UK retail arm of E.ON, says its AI voice assistant CLaiRE answers all residential calls and supports more than five million accounts, handling areas such as meter readings and payments. A second assistant, DaiVID, verifies callers for Sainsbury's Energy customers before they speak to an agent. E.ON's chief commercial officer also describes AI voice assistants that can securely take a payment in seconds. The page gives no containment, volume or satisfaction figures.
No outcome disclosed.
KBC
Belgium · Banking · 2025
KBC's digital assistant Kate, launched in KBC Mobile in 2020, has grown from a simple chatbot into a proactive assistant that guides customers through banking and insurance moments, such as spotting duplicate payments or checking before a trip whether their bank card is activated for use abroad. Since October 2025 Kate runs entirely on the GPT-4.1 large language model, which KBC says makes it better at understanding complex questions and giving more personal answers. KBC positions Kate's next step as offering personalized advice on topics such as savings and buying a home.
- Users served: 5.8 million
"Kate actively serves 5.8 million digital customers in Belgium, the Czech Republic, Hungary, Slovakia and Bulgaria"
Claimed by: organization - Interactions handled: 80 million, since launch
"80 million conversations have been started with Kate since launch"
Claimed by: organization - Containment rate: 70%
"Kate currently solves 70% of customer queries independently and we expect this percentage to increase in the future"
Claimed by: organization
Kenya Power
Kenya · Energy and utilities · 2025
Kenya Power (Kenya Power and Lighting Company) launched its first AI chatbot, named Nuru, alongside a redesigned MyPower App and a Kiswahili menu on its USSD short code, for its more than 10 million customers. Nuru answers customer queries on the company's website and its Facebook page, lets customers report incidents such as power outages, and can connect them to a human customer care representative. The launch was part of a wider digital customer engagement upgrade that also added self service billing, token purchase and complaint features to the MyPower App.
No outcome disclosed.
Magyar Telekom
Hungary · Telecommunications · 2025
Magyar Telekom introduced Vanda, an NLP based virtual assistant, to residential customers in 2018, replacing traditional push button IVR menus with voice controlled service. Built on speech recognition, language understanding and a synthetic voice, Vanda started by resolving about twenty common customer requests end to end on its own, or connecting the caller to the right agent when needed. The operator's customer service director says Vanda now resolves on average 25 to 27 percent of cases, freeing staff for the more complex requests that need expertise, and that satisfaction scores are approaching those of expert human agents.
- Containment rate: at least 25%
"Vanda elsősorban az egyszerű, ismétlődő ügyfélkérésekre nyújt teljes körű, end-to-end megoldást, így átlagban az esetek 25-27 %-ában virtuális asszisztensünk oldja meg az ügyeket."
Claimed by: organization
NatWest Group
United Kingdom · Banking · 2025
NatWest routes many kinds of everyday customer queries through Cora, its AI assistant in online banking and the mobile app, now with generative AI (Cora+). Customers whose ATM withdrawal did not pay out are sent to Cora with the phrase "ATM dispute" as the first step of the claim, and the assistant is available before login as well.
NatWest disclosed that Cora handled 10.8 million queries over 2023, more than double the roughly five million it handled in 2019, and that the top five questions asked via Cora are about canceling a transaction, changing an address, asking for a bank statement, requesting a new bank card or opening an ISA.
NatWest says the generative AI version improved customer satisfaction and reduced how often a colleague has to step in, and in 2025 it began a collaboration with OpenAI to extend the assistant to more complex tasks.
- Satisfaction uplift: 150%, Cora+ generative AI functionality
"The GenAI functionality offered by Cora+ has shown a 150% improvement in customer satisfaction, while reducing the number of times a colleague needs to intervene."
Claimed by: organization - Interactions handled: 10.8 million, over 2023
"NatWest revealed the virtual chatbot handled 10.8 million queries over 2023, up from about five million in 2019."
Claimed by: organization
Swisscom
Switzerland · Telecommunications · 2025
Since the end of 2025, callers to the Swisscom hotline are greeted by Sam, an AI voice assistant that takes the request in the caller's own words, including Swiss German, instead of a touch tone menu. Sam answers simple questions, handles administrative tasks such as extending a payment deadline or issuing a login code, does the first steps of authentication and connects the rest to the responsible specialist. Swisscom says Sam handled over 2.3 million requests by August 2026. After the launch, customers reported difficulties with the assistant; Swisscom now lets callers ask for a person at any time, keeps adjusting Sam and says it now sees fewer callbacks.
- Interactions handled: at least 2.3 million, from launch at the end of 2025 to August 2026
"Seit seiner Einführung Ende 2025 hat der KI-gestützte Sprachassistent Sam an der Hotline über 2,3 Millionen Anliegen bearbeitet."
Claimed by: organization
Trainline
United Kingdom · Travel and hospitality · 2025
Trainline Assistant is an agentic AI feature in the Trainline app for customers who bought a domestic UK ticket. It answers questions with knowledge of the specific ticket and journey, shows real time travel details and can issue a refund for an eligible ticket from a plain language request. Where it cannot complete a request, the customer is connected to Trainline's customer service team. At the June 2025 launch Trainline said 88% of queries were handled to completion, and by December 2025 it reported more than a million conversations since the beta.
- Containment rate: 88%
"At present it has an average response time of 12 seconds and 88% of queries are handled to completion – where not possible, users are connected to Trainline’s customer service team."
Claimed by: organization - Interactions handled: at least 1 million, from the beta launch in 2025 to December 2025
"Since its beta launch earlier this year, it's had over a million conversations with customers and almost a third are now repeat users."
Claimed by: organization
Umniah
Jordan · Telecommunications · 2025
Umniah by Beyon, a Jordanian mobile operator, introduced Raya in 2025, a conversational assistant it says is powered by AI rather than rules. Raya handles general inquiries, complaints, billing and balance questions and plan changes, and can diagnose technical issues, on the Umniah website, the Umniah app and WhatsApp, around the clock. The announcement names no vendor and discloses no usage or resolution figures.
No outcome disclosed.
Asiacell
Iraq · Telecommunications · 2024
Asiacell, Iraq's largest mobile operator, launched LAILA on 11 November 2024 as what it calls Iraq's first AI powered customer support assistant. LAILA answers customer inquiries and helps with technical issues around the clock, reachable through the Asiacell app, the operator's website, WhatsApp and Facebook Messenger. The launch announcement names no vendor. A July 2026 webinar promotion run with Druid AI names Druid AI as the platform and states that Laila automates 79% of digital customer interactions end to end; that figure is a vendor claim.
- Containment rate: 79%, digital customer interactions since Laila was deployed
"Since deploying Laila, their Druid AI powered agent, Asiacell has automated 79% of digital customer interactions end-to-end."
Claimed by: vendor
Batelco
Bahrain · Telecommunications · 2024
Batelco, part of the Beyon Group in Bahrain, launched Basma in July 2024, a digital assistant built on generative AI and available by voice and chat in the Batelco mobile app and on its website, in English and Arabic, around the clock. At launch Basma covered FAQs, bill and usage inquiries, package and device installments, fiber order tracking, line reconnection and add on activation. In July 2026 Batelco described Basma as completing both support and sales actions behind the scenes and handling thousands of customer conversations daily. The company discloses no containment, satisfaction or cost figure.
No outcome disclosed.
BT Group
United Kingdom · Telecommunications · 2024
BT Group runs the EE virtual assistant Aimee on Sprinklr's customer experience platform, drawing on BT Group data for personalized answers. The platform lets BT Group use generative AI for EE and BT customers, for example in an Aimee flow that prepares customers for international travel and in billing support, where generative AI gives detailed explanations of billing charges. BT Group says Aimee handles up to 60,000 conversations a week, double the volume of two years earlier, that the travel flow halved the need for chat support, and that it stays model agnostic behind a private cloud instance with safeguards against attempts to make the AI misbehave.
- Interactions handled: up to 60,000, per week
"EE virtual assistant Aimee now handles up to 60,000 customer conversations per week, with automation success rates on several types of customer journey now approaching 50%, freeing time for guides to focus on more complex issues"
Claimed by: organization - Automation rate: up to 50%, several types of customer journey
"EE virtual assistant Aimee now handles up to 60,000 customer conversations per week, with automation success rates on several types of customer journey now approaching 50%, freeing time for guides to focus on more complex issues"
Claimed by: organization
Jordan Ahli Bank
Jordan · Banking · 2024
Jordan Ahli Bank launched "ahliGPT", an AI system that answers customer queries about banking and financial services, and announced it as the first bank in the Middle East to bring generative AI into customer service this way. The bank's own announcement names no vendor or launch date beyond the article's publication, and discloses no usage or resolution figures; CEO Ahmad Al-Hussein is quoted on the bank's ambition for the technology rather than on any measured result.
No outcome disclosed.
Klarna
Sweden · Payments and cards · 2024
Klarna announced in February 2024 that its AI assistant built on OpenAI models had been live globally for a month as the first line of its customer service, handling refunds, returns, payment issues, cancellations and disputes in more than 35 languages across 23 markets. In 2025 the company said it had gone too far in replacing people and began recruiting human agents again so that customers can always reach a person; the assistant still handles the majority of inquiries. The record is useful precisely because it shows both the gain and the correction.
- Interactions handled: 2.3 million, first month after launch
"The AI assistant has had 2.3 million conversations, two-thirds of Klarna’s customer service chats"
Claimed by: organization - Response time reduction: 82%, since launch, as reported in 2025
"Since launch, response times have improved by 82%, and Klarna has seen a 25% drop in repeat issues."
Claimed by: organization
Poste Italiane
Italy · Cross industry · 2024
Poste Italiane's Assistente Digitale was introduced in 2019 alongside human operators and uses artificial intelligence to answer customer service requests by phone, web and app, 24 hours a day, 7 days a week. It resolves simpler requests quickly (for example a duplicate PIN, balance and transaction list, or blocking a prepaid card) and routes more complex problems to human operators. Over the years it has been extended to financial, telephony, logistics, insurance and energy services. In July 2024 Poste Italiane reported that it had intercepted around 80 million requests since activation, a significant share of them handled autonomously.
- Interactions handled: about 80 million, cumulative since activation in 2019, reported July 2024
"Dalla sua attivazione l’Assistente Digitale, operativo 24 ore su 24, 7 giorni su 7, per 365 giorni l’anno, ha intercettato circa 80 milioni richieste, di cui una quota significativa gestita in modo autonomo."
Claimed by: organization
Vodafone
United Kingdom · Telecommunications · 2024
Vodafone rebuilt its TOBi chatbot on Azure OpenAI as SuperTOBi, launched in Italy and Portugal, with Germany and Turkey announced to follow from July 2024 and other markets later that year. A companion SuperAgent helps human agents search the company knowledge base and, in Ireland, sends the human agent a summary of the online customer conversation so customers do not repeat themselves. Vodafone reports that initial tests at one of its call centers showed a 50% improvement in first time resolution of critical journeys such as complex billing inquiries, and that in Portugal first time resolution on appointment booking rose from 15% to 60%, with billing journeys being added next.
- First contact resolution: 60%, appointment booking flow, Vodafone Portugal
"As a result, the first-time resolution rate has increased from 15% to 60% and Vodafone’s online net promoter scores (where respondents are asked to rate their experience) improved by 14 points to 64 points – anything above 50 points is considered a strong result."
Claimed by: organization - NPS change: +14 points, online NPS, Vodafone Portugal
"As a result, the first-time resolution rate has increased from 15% to 60% and Vodafone’s online net promoter scores (where respondents are asked to rate their experience) improved by 14 points to 64 points – anything above 50 points is considered a strong result."
Claimed by: organization
Alizz Islamic Bank
Oman · Banking · 2023
Alizz Islamic Bank in Oman launched a chatbot that answers customer questions about its products and services around the clock and lets customers contact the bank directly. It went live on WhatsApp, on the bank's phone number, and on the bank's Instagram page, with a web version planned for the corporate website. The bank frames it as part of its digital transformation and gives no volume or resolution figures.
No outcome disclosed.
Vodacom
South Africa · Telecommunications · 2023
Vodacom South Africa's TOBi is an AI assistant that answers customer queries such as account balances, SIM swaps, PUK retrieval, invoices and statements, and guides customers through the Self RICA identity verification process, across SMS, WhatsApp, the My Vodacom and VodaPay apps, the web and voice. Vodacom added isiZulu, South Africa's most widely spoken home language, as a second language alongside English, reporting comparable understanding rates for the two languages on both the voice and chat versions of the assistant.
- Interactions handled: about 2 million, per month
"TOBi can solve just over 2 million customer queries a month via SMS, WhatsApp, My Vodacom App, VodaPay App, Web and Voice."
Claimed by: organization
Dukhan Bank
Qatar · Banking · 2021
Dukhan Bank launched Rashid in November 2021, an AI powered virtual assistant that answers customer questions in English and Arabic on the bank's website, mobile app, online banking and WhatsApp. Rashid finds the nearest branch or ATM, shows nearby merchant offers using geolocation, gives information about products and services, checks foreign currency exchange rates and connects the customer to a service team member. It is also available to Dukhan Bank employees on the staff portal. The bank planned to add fund transfers, card activation and blocking, spend analysis and bill payments. No usage figure is disclosed.
No outcome disclosed.
Ingka Group
Sweden · Retail and ecommerce · 2021
Ingka Group, the largest IKEA retailer, rolled out the AI chatbot Billie in its 2021 financial year to answer simpler customer enquiries around the clock. Between 2021 and 2023 Billie resolved about 47% of the enquiries it received. With the chatbot taking simpler enquiries, Ingka reskilled 8,500 call center staff for remote interior design and remote selling, and sales through its remote customer meeting points reached EUR 1.3 billion in its 2022 financial year.
- Containment rate: about 47%, 2021 to 2023
"Since the rollout of the solution in FY21[1], Billie has continued to provide value, and from 2021 to 2023 it resolved approximately 47% of customer enquiries it received, which translates to 3,2 million interactions solved by the chatbot and nearly EUR 13 million in savings thus far."
Claimed by: organization - Interactions handled: about 3.2 million, resolved by the chatbot, 2021 to 2023
"Since the rollout of the solution in FY21[1], Billie has continued to provide value, and from 2021 to 2023 it resolved approximately 47% of customer enquiries it received, which translates to 3,2 million interactions solved by the chatbot and nearly EUR 13 million in savings thus far."
Claimed by: organization - Cost savings: about EUR 13 million, cumulative, 2021 to 2023
"Since the rollout of the solution in FY21[1], Billie has continued to provide value, and from 2021 to 2023 it resolved approximately 47% of customer enquiries it received, which translates to 3,2 million interactions solved by the chatbot and nearly EUR 13 million in savings thus far."
Claimed by: organization
Qatar Islamic Bank
Qatar · Banking · 2021
Qatar Islamic Bank (QIB) launched "Zaki," which it describes as the first conversational AI banking service in Qatar, as an additional channel on its website. Zaki answers customer questions about the bank and its products in Arabic and English, around the clock, without the customer needing to visit a branch or call the QIB call center, and can hand visitors to a list of tailored products or a sales call back. QIB positioned Zaki as the foundation for a later voice and chat enabled conversational banking service, with transaction capability and a mobile app presence planned as future releases.
No outcome disclosed.
Zenith Bank
Nigeria · Banking · 2021
Zenith Bank, one of Nigeria's largest banks, launched ZiVA (Zenith Intelligent Virtual Assistant), an AI powered chatbot on WhatsApp, so customers could transact and get real time customer service around the clock on the encrypted messaging platform they already used. Through ZiVA, customers can open new accounts, receive instant transaction notifications, check balances, transfer funds, top up airtime, confirm checks, pay bills, apply for loans, block their accounts and request mini statements. Group Managing Director Ebenezer Onyeagwu said the launch was driven by the need for additional secured communication channels as the bank deepened its retail penetration.
No outcome disclosed.
Aman Bank
Libya · Banking · 2020
Aman Bank, a private Libyan bank, introduced "Dareen" (spelled "Darren" in the press) in November 2020 as a virtual assistant on its Facebook Messenger page, and says it was the first banking service of its kind in Libya. Customers use it to ask about balances, transactions, pending and blocked payments, products such as opening a current account, how to apply for a loan, and transfers between accounts. The bank states that it is built on Oracle's artificial intelligence. No usage or resolution figures are published.
No outcome disclosed.
Aman Bank
Libya · Banking · 2020
Aman Bank launched "Maarya" (called "Maria" in the press) in October 2020, a virtual assistant that answers the bank's phone banking line around the clock in Arabic or English. Callers state what they need in natural language instead of working through menus, and can ask about balances and transactions, update contact details, request card services such as blocking or a new card, and set up transfers. The caller is verified by national number and a registered phone number, and can press star 9 to reach a human agent. No usage or resolution figures are published.
No outcome disclosed.
Banque Misr
Egypt · Banking · 2019
Banque Misr launched a self service chatbot on its website that it describes as a first among Egyptian banks, answering customer inquiries around the clock and also receiving customer complaints. At the Arab African Youth Forum in March 2019 the bank told the audience that, over its first three months, the service averaged 16,000 calls a month and covered 79% of client calls without call center staff. Later bank news pages present the assistant as the character Abqarino, shown at the Cairo ICT fair in 2022. The release does not say how a call is counted or which languages the chatbot spoke.
- Containment rate: 79%, first three months of service, share of total client calls
"Within 3 months, the service was offered for average 16000 calls per month, representing 79% of the total clients’ calls with no need for the assistance of the call-center staff."
Claimed by: organization - Interactions handled: 16,000, per month, average over the first three months
"Within 3 months, the service was offered for average 16000 calls per month, representing 79% of the total clients’ calls with no need for the assistance of the call-center staff."
Claimed by: organization
Kotak Mahindra Bank
India · Banking · 2018
Kotak Mahindra Bank launched Keya in April 2018, billed as Indian banking's first AI powered voicebot, built with Nuance and integrated into the bank's phone banking helpline alongside the existing IVR. Keya uses speech recognition, natural language understanding and text to speech to identify what a caller wants, verify it and route simple requests to self service while sending complex calls to a human agent, so customers no longer have to work through the traditional IVR menu.
- Handling time reduction: at least 50%
"Keya more than halves the time of delivering an accurate response to a customer's query as he/she no longer has to go through the maze of options offered on the traditional IVR system."
Claimed by: organization
Ooredoo Oman
Oman · Telecommunications · 2018
Ooredoo Oman introduced Saeed in 2018 and calls it the first AI chatbot from a telecom company in Oman. It answers typed questions in English and Arabic around the clock in the Ooredoo app and on the website, covering mini bills, barring status, new SIM and PUK, bill copies, call and usage history, roaming bundles, active plans and postpaid balance. An update announced in July 2019 let Saeed recognize the customer type during the chat (mobile or fixed, prepaid or postpaid) and offer the matching account services. No volume or containment figure has been disclosed.
No outcome disclosed.
Roads and Transport Authority, Dubai (RTA)
United Arab Emirates · Government and public sector · 2018
Dubai's Roads and Transport Authority launched Mahboub in October 2018 as a virtual assistant based on natural language processing and added generative AI to it in 2024. It runs on the RTA website, its smart apps and WhatsApp in Arabic and English and offers 330 services, among them fine payment, vehicle renewal, nol card top up and appointment booking. RTA says it has handled more than 27 million conversations and completed 800,000 transactions, passes only 1.5 percent of chats to a live agent and has cut demand on the call centers by up to 70 percent. RTA's review of 2025 adds that a Madinati service in the Mahboub WhatsApp chatbot uses computer vision and generative AI.
- Interactions handled: at least 27 million, conversations since the October 2018 launch, as stated in August 2025
"Mahboub has since handled over 27 million conversations and completed 800,000 transactions"
Claimed by: organization - Contact deflection: up to 70%, since the October 2018 launch, as stated in August 2025
"This has reduced demand on call centres by up to 70%, saving customers time and effort."
Claimed by: organization
Deutsche Telekom
Germany · Telecommunications · 2026
Deutsche Telekom's chief AI officer said in September 2026 that AI agents are replacing the automated phone menus on its customer service lines. In some of its ten European markets callers no longer reach the menu, and others are to follow within months.
Simple actions such as checking and rebooting a router or explaining a high bill go to AI agents first and complex calls go to humans, with adapters and caching added so that a 30 to 60 second router reboot does not mean hold music.
He said an agent switched on for a new use case starts at no more than 20% of those calls and reaches 40% within a couple of months, and that the share of interactions that never reach a human grows by one to two percentage points a month. No overall containment figure was given.
No outcome disclosed.
Khaleeji Bank
Bahrain · Banking · 2026
Khaleeji Bank in Bahrain launched Sheikha in June 2026, an AI powered voice assistant that clients reach by calling a dedicated number, 24 hours a day, seven days a week. Clients hold natural voice conversations to get banking support and product information, and the assistant also collects client feedback and shares news of offers and campaigns. The bank positions it as a complement to its call center and branch network and says it is the first bank in Bahrain to offer natural voice conversations over the phone. No usage or resolution figure is disclosed.
No outcome disclosed.
KPN
Netherlands · Telecommunications · 2026
KPN, the largest telecom provider in the Netherlands, built a central platform that connects voice agents to its core telecom systems. The agents take customer verification, order status inquiries, technician appointment management and internet troubleshooting, with humans kept in the loop at moments such as a customer moving home. The team targets response times under two seconds per turn, lets callers interrupt, and reviews transcripts of up to 100 real calls every morning.
Early data put customer satisfaction on par with calls handled by people, but the score was published without its scale. KPN has an ambition for agentic AI to handle 10% to 20% of its roughly five million yearly calls by 2027, which is a target, not a result.
No outcome disclosed.
Lascana
Germany · Retail and ecommerce · 2026
Lascana, a fashion brand of the Otto Group, automated standard customer service requests with a chatbot from BOTfriends, selected through the Otto Group venture client unit Otto Dock 6. The bot uses a multi agent design combined with fixed business logic and is meant to move from answering questions to recommending products, following orders and running service processes. In the pilot phase, 82.5% of users rated their experience as positive.
- Customer satisfaction: 82.5%, pilot phase
"Außerdem bewerteten 82,5 Prozent der Nutzer ihre Erfahrungen mit dem KI-Chatbot als positiv."
Claimed by: vendor
Münchener Verein Versicherungsgruppe
Germany · Insurance · 2026
Münchener Verein, a German health, life and property insurance group, replaced its IVR menu with Ella, a Parloa voice agent that identifies the caller's concern, authenticates policyholders by name, insurance number, date of birth and address, answers frequently asked questions and passes the call with full context to the right team. The first use cases went live within ten weeks. Parloa reports break even after about three months and says the proof of concept is now a scaled solution.
No outcome disclosed.
Nordea
Finland · Banking · 2026
Nordea, the leading Nordic bank, built its virtual agent Nova on the boost.ai conversational AI platform, starting with Nordea Norway in 2017 and extending it to Sweden, Denmark and Finland in 2021, so that customers who open chat on the web or in the mobile app in any of the four countries reach Nova first, before any human agent.
Nordea's Group Head of AI Strategy and Acceleration, Mattias Fras, and business owners in each country describe Nova as central to a chat first customer service strategy that, as of a case study last updated in March 2026, runs 12 AI agents across the four markets, in each market's own language as well as English in some countries, covering both customer facing questions and internal support functions such as HR.
The case study reports Nova handling over 220,000 conversations a month for private banking customers across the Nordics, with Nordea Sweden reporting a 91 percent in scope resolution rate for private banking customers and 95 percent for corporate customers.
- Containment rate: 91%, to date, Nordea Sweden
"To date, the AI Agent has a 91% in-scope resolution rate for private banking customers and 95% for corporate customers."
Claimed by: vendor - Containment rate: 95%, to date, Nordea Sweden
"To date, the AI Agent has a 91% in-scope resolution rate for private banking customers and 95% for corporate customers."
Claimed by: vendor - Containment rate: at least 90%, across all of Nordea's customer facing AI agents, as of the case study last updated 20 March 2026
"90%+ in-scope resolution rates across all customer-facing AI Agents"
Claimed by: vendor - Interactions handled: at least 220,000, per month
"This is impressive when you factor in that Nova averages over 220,000 conversations per month across the Nordics, purely for private banking customers"
Claimed by: vendor
Norwegian Air Shuttle
Norway · Travel and hospitality · 2026
Norwegian, the Norwegian low cost airline, runs an AI support agent from Kindly on chat that answers the most frequent passenger enquiries in Norwegian and English and finds bookings. Norwegian's principal advisor for service delivery recalls that the 737 MAX grounding came shortly after the chatbot started, which places the launch around early 2019. The Kindly case study, written with Norwegian's customer service leaders, reports that live enquiries to agents fell by 20 percent, live chats by 30 percent and incoming calls by 5 percent, without giving the comparison period. About a thousand people work in customer service across phone, chat, email and social media.
- Contact deflection: 20%, reduction in live enquiries handled by the customer service team, comparison period not stated, as stated by Kindly in a 2026 case study
"Norwegian's AI Support Agent reduced the number of live enquiries handled by their team by 20%."
Claimed by: vendor
OTE Group
Greece · Telecommunications · 2026
OTE Group, the Deutsche Telekom controlled operator that runs one of the highest volume customer support operations in Greece for its Cosmote TV pay TV service, built a generative AI voice agent with vendor Wonderful, integrated with its Genesys telephony and backend systems. The agent covers ten skill areas, including subscriber authentication, troubleshooting and satellite reset, package and pricing information, profile management, service barring and program lookup. OTE's chief customer operations officer and CIO are both quoted on the deployment, which Wonderful describes as production grade and handling tens of thousands of calls.
- Contact deflection: 50%, within 8 weeks of the agent going live, still rising as of publication
"Deflection rate reached 50% (nearly triple the baseline) and is still climbing as new flows and knowledge are added."
Claimed by: vendor - Handling time reduction: 30%
"Average handling time was reduced by 30% with the AI agent."
Claimed by: vendor
Ryanair
Ireland · Travel and hospitality · 2026
Ryanair runs a single customer service agent on Amazon Bedrock AgentCore and Amazon Nova 2 Lite that answers chats in seven languages across baggage, refunds, bookings and flight disruptions. Every answer passes a multi model guardrail scoring step before the customer sees it, and customers can be routed between chat and the Amazon Connect voice channel. AWS reports an 80% containment rate on about 120,000 chat interactions a day. The case study does not say which share of contacts end with a completed refund or rebooking.
- Containment rate: 80%
"The solution now handles 120,000 customer chat interactions daily across seven languages, achieving an 80% containment rate."
Claimed by: vendor - Interactions handled: 120,000, per day
"The solution now handles 120,000 customer chat interactions daily across seven languages, achieving an 80% containment rate."
Claimed by: vendor
Trust Bank
Singapore · Banking · 2026
Trust Bank, a Singapore digital bank, added a generative AI chatbot to its app that answers customers in natural language using updated product information and customer context. It handles common requests such as deleting a payee, paying a credit card bill, replacing a card and getting statements, and hands complex cases to human agents with the conversation history. Its customer care team helped refine the knowledge base. Trust Bank reports that human handled chats fell by 50 percent and that complaints dropped by about 40 percent within months of the rollout.
- Contact deflection: 50%, customer chats handled by human agents, within months of rollout
"Trust Bank has cut customer chats handled by human agents by 50 percent and reported around a 40 percent drop in complaints within months of rolling out its Gen AI chatbot."
Claimed by: organization
Uelzener Versicherungs-Gruppe
Germany · Insurance · 2026
Uelzener Versicherung, a German specialty animal insurer with more than 500,000 customers, moved from an IVR menu to Clara, a Parloa voice agent that recognizes the request, summarizes the conversation for the service team and forwards the call to the right skill team. The insurer says many callers had miscategorized themselves in the old menu, so the agent also gave a clearer picture of the main service issues. Parloa describes routing accuracy as nearly complete and the abandonment rate as low, without a measured figure.
No outcome disclosed.
UmweltBank AG
Germany · Banking · 2026
UmweltBank, a German bank, uses a voice AI system from the Swiss provider Spitch that captures the caller's request in spoken language, understands dialects and colloquial speech and routes about 400 calls a day to the responsible teams automatically. Callers can book a callback when lines are busy. The system runs in an EU cloud and the bank plans further modules. The bank says it will use AI in a supporting role, not to let a bot take every call.
- Accuracy: at least 90%, recognition rate
"Dank kontextueller Sprachanalyse versteht die Lösung auch Dialekte und Umgangssprache zuverlässig und erzielt Erkennungsquoten von über 90 Prozent."
Claimed by: vendor - Interactions handled: about 400, calls routed per day (not resolved)
"Das System erfasst Kundenanliegen in gesprochener Sprache und leitet rund 400 Anrufe pro Tag automatisch an die zuständigen Teams weiter."
Claimed by: vendor
ABN AMRO
Netherlands · Banking · 2025
ABN AMRO moved to Microsoft Copilot Studio to build two AI assistants: Anna for customers and Abby for employees. Anna now handles a large share of the bank's first line customer contact by text and voice, automating more than half of those interactions with a set of AI agents working together, while Abby gives employees easier access to IT and other internal resources.
- Interactions handled: at least 2 million, per year
"‘Anna,’ the AI agent for customers, now supports more than two million text conversations and 1.5 million voice conversations annually, automating over 50% of interactions with multiple AI agents."
Claimed by: vendor - Interactions handled: 1.5 million, per year
"‘Anna,’ the AI agent for customers, now supports more than two million text conversations and 1.5 million voice conversations annually, automating over 50% of interactions with multiple AI agents."
Claimed by: vendor - Containment rate: at least 50%
"‘Anna,’ the AI agent for customers, now supports more than two million text conversations and 1.5 million voice conversations annually, automating over 50% of interactions with multiple AI agents."
Claimed by: vendor
AOK PLUS
Germany · Insurance · 2025
AOK PLUS, the statutory health insurer for Saxony and Thuringia with about 3.5 million insured people, introduced an AI voice agent from Cognigy in 2025 that answers frequent questions such as branch opening hours, care benefits, sick pay and the electronic patient record, and routes other requests to the responsible teams. Cognigy reported over 30,000 calls supported in the first month, with full automation of selected topics still being tested. In 2026 AOK PLUS brought the AI agents and its 2,400 service staff onto one NICE CXone platform in an EU sovereign cloud.
- Interactions handled: at least 30,000, first month after launch
"Just one month after launch, the new system has already successfully supported over 30,000 calls."
Claimed by: vendor
AtlantaSanad Assurance
Morocco · Insurance · 2025
AtlantaSanad announced Assia in September 2025, an AI assistant on its website that it presents as the first of its kind among Moroccan insurers. Assia is available around the clock and answers policyholders' questions in real time. It explains the steps after a car accident, helps fill in the joint accident report, lists the documents needed for a claim file, details cover and limits in contracts, and gives advice on taking out a policy, handling a claim or surrendering a savings and retirement product. It adds to a chatbot the insurer already ran on messaging apps. No usage figures are disclosed.
No outcome disclosed.
Bausparkasse Schwäbisch Hall AG
Germany · Banking · 2025
Bausparkasse Schwäbisch Hall, Germany's largest building society with about seven million contracts, uses a Parloa voice agent on its service numbers for customers and sales partners. The agent recognizes the request, authenticates callers by dialog, answers building savings questions, updates master data such as addresses end to end and routes the rest to the right department. It covers 16 use cases. Parloa reports 500,000 calls handled in the first six months and 98% intent recognition.
- Interactions handled: 500,000, first six months
"500,000 calls handled by the AI Agent in 6 months"
Claimed by: vendor - Accuracy: 98%, intent recognition
"98% success rate in intent recognition"
Claimed by: vendor
BLS AG
Switzerland · Logistics and transportation · 2025
BLS, the Swiss railway company that carries about 200,000 passengers a day, automated part of its telephone customer support with the Spitch agentic AI platform, built with the IT provider UMB. Callers can choose the AI or a person, the system understands Swiss German dialects and answers many questions by SMS, and calls it cannot answer go to the contact center with the key facts on screen. Fare surcharge complaints are taken by the AI and passed to a callback queue, which reduces handling time and emotional load for staff.
No outcome disclosed.
bunq
Netherlands · Banking · 2025
bunq, a Dutch neobank, launched its Finn assistant in 2023 to help users search their finances and has built it into a multi agent system that investigates issues, processes documents and carries out banking operations for users who ask. In 2024 bunq added real time speech to speech translation so users can talk to support in their own language. Anthropic's customer story says that after bunq moved to Claude models its automated resolution rate rose by 8%, to over 80% of support tickets.
- Automation rate: at least 80%, support tickets
"With Claude’s advanced reasoning, bunq’s automated resolution rate improved by 8%, bringing their total automation to over 80% of all support tickets."
Claimed by: vendor
DATEV eG
Germany · Technology and software · 2025
DATEV, a German IT service provider for tax advisors, auditors and lawyers, uses a Cognigy voice agent on its support line, which receives more than 1.5 million complex support calls a year. The agent authenticates callers in a guided dialog, handles routine inquiries and transfers calls to service representatives with context. Cognigy reports a success rate above 99.9% for phone authentication.
- Accuracy: at least 99.9%, phone authentication
"AI Agent achieves over 99.9% success rate in Phone Authentication"
Claimed by: vendor
Deutsche Rentenversicherung Hessen
Germany · Government and public sector · 2025
Deutsche Rentenversicherung Hessen uses a voicebot on Deutsche Telekom's Conversational AI Suite in customer service, pension processing and rehabilitation. The bot records the reason for the call and identification details, then routes the caller through automatic call distribution to the responsible team with the dialog attached. Simple requests can also be completed by the bot. Telekom reports about 85,000 calls a month pre qualified automatically.
- Interactions handled: about 85,000, calls pre qualified and routed per month (not resolved)
"Aktuell werden monatlich etwa 85.000 Anrufe automatisiert vorqualifiziert."
Claimed by: vendor
Essent
Netherlands · Energy and utilities · 2025
Essent, the largest energy company in the Netherlands with 2.5 million customers under the Essent and Energiedirect brands, handles around 3 million calls and 900,000 chats a year. According to Cognigy's case study, every incoming chat is handled by AI first and Essent has fully automated 55% of them. An AI voice agent collects customer meter readings by phone four times a year. An agent copilot inside Microsoft Teams, trained on more than 1,100 articles, gives customer service representatives real time answers, live chats are summarised and logged in the CRM, and intelligent routing distributes calls and chats across 17 mission based teams.
- Containment rate: 55%, incoming chat conversations
"Across the Conversational AI landscape, Essent has now fully automated 55% of its incoming chat conversations"
Claimed by: vendor - Cost savings: EUR 240,000, per year, voice AI agent for meter reading collection
"The Voice AI for meter reading collection engaged 22,000 customers during a single two-week window, achieving a 22% success rate and contributing to €240,000 in annual savings."
Claimed by: vendor - Automation rate: 85%, internal helpdesk queries from customer service representatives, answered by the agent copilot
"The Agent Copilot tool has transformed internal service by automating 85% of previously helpdesk-handled queries, saving €2 million annually."
Claimed by: vendor - Cost savings: EUR 2 million, per year, agent copilot replacing internal helpdesk work
"The Agent Copilot tool has transformed internal service by automating 85% of previously helpdesk-handled queries, saving €2 million annually."
Claimed by: vendor
Movistar Plus+
Spain · Telecommunications · 2025
Movistar Plus+, the pay television platform of Telefónica in Spain, runs an AI agent on Salesforce Agentforce in its help centre chat that answers customer queries and incidents from knowledge articles and hands harder cases to human staff. After a pilot it went live in March 2025, as the first Agentforce use case in the Telefónica group, and in July 2025 it was extended to sports content questions using Data Cloud. The head of customer support described the rollout at a press event organised by Salesforce and gave no containment or satisfaction figure.
No outcome disclosed.
New Look
United Kingdom · Retail and ecommerce · 2025
New Look, a British fashion retailer with more than 300 UK stores and a strong online presence, launched Zendesk AI agents on email and live chat in late 2024, the day before Black Friday. The care team handles about 9,500 customer inquiries a week. Sam Ashton, Senior Customer Care Manager, says the agents run without backend integrations and resolve an average of 42 percent of customer queries on their own. Zendesk adds that the agents deflected 305,761 contacts and that Copilot helps the human team answer faster.
- Containment rate: 42%, average across email and live chat
"Our AI agents currently operate without integrations, yet they already resolve an average of 42 percent of customer queries independently"
Claimed by: organization
Ooredoo Qatar
Qatar · Telecommunications · 2025
Ooredoo Qatar unified its call center and digital channels (app, web, WhatsApp, Messenger, Instagram and X) on Microsoft Dynamics 365 Omnichannel, so more than two hundred agents who had worked each channel separately now share one platform and one view of the customer. Its AI assistant Obot, built on Microsoft AI Foundry and Azure AI Services, resolves a substantial share of customer queries without human intervention, and Ooredoo later extended Obot with a GPT real time voice avatar and made it available to retail customers across the same social channels. Microsoft's case study on the program reports that average handling time fell from seven to five minutes.
- Handling time reduction: about 28.6%
"The average handling time for customer interactions fell from seven minutes to five, unlocking significant productivity gains across the organization."
Claimed by: vendor
Pegasus Airlines
Türkiye · Travel and hospitality · 2025
Pegasus Airlines retrained FlyBot, the virtual assistant on its website, with Azure OpenAI and integrated it with internal systems, so customers can ask about flights, flight rules, baggage allowances and claims and reissue tickets in the same conversation. The airline reports that satisfaction with the virtual assistant doubled after the change.
- Satisfaction uplift: 100%
"“Since we integrated Azure AI Services into our FlyBot, customer satisfaction rates for our virtual assistant have doubled,” points out Bora."
Claimed by: organization
Royal Air Maroc
Morocco · Travel and hospitality · 2025
Royal Air Maroc announced Atlas in April 2025, a digital assistant inside its RAM Assistant mobile app that it built in house on generative AI. Atlas gives flight status, shows the progress of complaints, explains products such as the Flight Pass, upgrade options and onboard services, and makes recommendations for the traveler. It understands Arabic, French, Darija and Wolof and, according to the announcement, uses ChatGPT for query understanding. At launch it was in beta, with a stated plan to cover the whole journey from trip preparation to after the flight. No usage figures are disclosed.
No outcome disclosed.
Together Credit Union
United States · Banking · 2025
Together Credit Union runs a voice agent on its inbound phone line that answers member questions, resolves routine requests and hands the rest to a person, drawing on one knowledge base that the branch, contact center and chat channels also use. The vendor reports that the agent now contains the majority of inbound calls, 12 points more than at the start, that after hours escalations to an outsourced contact center fell, and that member satisfaction held above 94% during the rollout.
- Containment rate: 66%, total inbound calls, as reported by the vendor
"Heather, Together CU's Posh-powered voice agent, now contains 66% of total inbound calls"
Claimed by: vendor - Contact deflection: 14%, after hours escalations to the outsourced contact centre
"After-hours escalations to their third-party contact center dropped 14%."
Claimed by: vendor
University Hospitals Sussex NHS Foundation Trust
United Kingdom · Healthcare · 2025
University Hospitals Sussex, which serves 1.6 million people across seven hospital sites, moved its outpatient booking centre onto Netcall's Liberty Converse contact centre platform and uses its AI assistant, which the Trust calls AVA, on the phone line. AVA works out in natural language what the patient is asking for, gives the answer or the information patients need to serve themselves, and routes the rest to the right service.
Netcall reports that intelligent routing and self service together deflect about 36% of calls from the booking centre and that waits fell from over 30 minutes to under three after the platform switch, but it does not give a result for AVA on its own. A multilingual AI chatbot was still being implemented.
No outcome disclosed.
Württembergische Versicherung AG
Germany · Insurance · 2025
Württembergische Versicherung, part of the W&W Group, has a Parloa AI agent answer all calls on its main hotline in natural language, classify the request and route the caller to the right expert. The insurer handles about 300,000 calls a year on that line, and a quarter need a handover to a specialist or another W&W company. Parloa reports a 33% reduction in call wait times in the first four weeks.
- Response time reduction: 33%, first four weeks after go live
"33% call wait times reduction"
Claimed by: vendor
Commonwealth Bank of Australia
Australia · Banking · 2024
Commonwealth Bank built a central AI orchestration agent that reads the customer's intent and routes it to a conversational AI, retrieval over public content, a deterministic guarded path for regulated journeys such as fraud disputes, or a human specialist with the full context, on its messaging channel. It migrated nearly 700 chatbot topics and launched a generative AI banking chatbot in November 2024. Voice bots are a planned extension of the orchestration layer.
- Containment rate: about 84.6%, May 2026, self service messaging
"In May 2026, approximately 84.6% of self-service messaging interactions were resolved end-to-end in the messaging channel."
Claimed by: vendor
MONETA Money Bank
Czechia · Banking · 2024
MONETA Money Bank in Czechia saw contact centre call volume rise by 300 percent between 2016 and 2022. With NTT Czech Republic it deployed Tom, a voicebot built on the Feedyou platform and Microsoft Azure AI speech services, to take common calls such as blocking lost cards, changing card limits and mortgage information. The Microsoft customer story reports that Tom handles up to 25 percent of all inquiries without an agent and that the bank cut call centre operating costs by 10 percent while customer satisfaction rose.
- Contact deflection: up to 25%, share of total inquiries resolved without contact centre agents
"It efficiently manages as much as 25 percent of the total inquiries received, resolving tasks such as blocking lost cards, adjusting card limits, providing mortgage information, and more, all without intervention by MONETA Money Bank’s contact center agents."
Claimed by: vendor - Cost reduction: 10%, call center operational costs
"As a result, the bank cut call center operational costs by 10 percent, while increasing customer satisfaction."
Claimed by: vendor
Qatar Airways
Qatar · Travel and hospitality · 2024
Qatar Airways launched Sama 2.0 at ITB Berlin in March 2024, a digital human cabin crew member that answers customer questions in real time through generative AI. Sama covers Qatar Airways FAQs, destinations, on board and on ground services, travel information and support tips. Customers reach Sama by phone or computer in QVerse, the airline's immersive digital platform; the launch release said Sama would also come to the Qatar Airways app. It was fluent in English at launch with Arabic and other languages planned, and it offers a handoff to a human agent when it cannot answer. The airline discloses no usage or outcome figure.
No outcome disclosed.
Salzburg AG
Austria · Energy and utilities · 2024
Salzburg AG, an Austrian utility with more than 450,000 customers, runs LEA, a chat agent that covers energy, telecommunications, water, e mobility, photovoltaics, public transport and tourism questions. In 2024 it added a GPT based layer on Cognigy to make answers more specific. Cognigy reports an 89% customer satisfaction rate and about 2,500 customer service contacts saved each month.
- Customer satisfaction: 89%
"This Generative AI-powered evolution of LEA, based on Cognigy Knowledge AI, has not only optimized operational efficiency but also significantly enhanced customer interactions, earning an 89% customer satisfaction rate."
Claimed by: vendor
Telkomsel
Indonesia · Telecommunications · 2024
Telkomsel, with more than 159 million mobile subscribers, rebuilt its Veronika virtual assistant on Azure OpenAI to handle routine customer questions in Bahasa Indonesia with the local accents and expressions its first chatbot could not follow, so that human agents could focus on complex issues. Telkomsel's chief information officer says the architecture can handle up to 5 million transactions a month. Telkomsel's chief marketing officer says customer self service interactions rose from 19% to 45% after the launch.
- Automation rate: 45%, share of customer interactions that were self service, as Telkomsel describes it
"“I’m thrilled to share that since introducing Veronika, we’ve seen a leap in customer self-service interactions from 19 percent to 45 percent,” enthuses Heng."
Claimed by: organization
Vietcombank
Vietnam · Banking · 2024
Vietcombank, a leading Vietnamese bank, built VCB Digibot with FPT Smart Cloud on the FPT.AI platform to answer routine questions around the clock on cards, loans, interest rates, preferential offers, exchange rates and the branch network, and to hand off anything beyond that scope to a human advisor. The aim was to free the bank's advisory team from repeating the same routine questions so they could spend more time on complex cases, while giving customers an immediate response over chat at any hour.
- Containment rate: up to 88.5%, first 6 months of deployment
"After 6 months of deployment, VCB Digibot has handled up to 88.5% of user requests, with only 11.5% requiring assistance from advisory staff, successfully processing nearly 2 million requests."
Claimed by: vendor - Interactions handled: about 2 million, first 6 months of deployment
"After 6 months of deployment, VCB Digibot has handled up to 88.5% of user requests, with only 11.5% requiring assistance from advisory staff, successfully processing nearly 2 million requests."
Claimed by: vendor
Vodafone
United Kingdom · Telecommunications · 2024
TOBi is Vodafone's digital assistant on the website, the My Vodafone app, messaging and telephony, first launched in Italy and extended to 15 language versions. It handles billing questions, contract updates and simple technical troubleshooting, hands over to a live agent with a summary, and during the pandemic made the same sales offers as human agents, such as data boosts and upgrades. Microsoft reports that TOBi now fully resolves 70% of inquiries arriving through digital channels; an earlier Microsoft story quotes Vodafone on a 12% year on year fall in contacts to call centers after launch.
- Containment rate: 70%, inquiries arriving through digital channels
"Currently, TOBi handles nearly 45 million customer calls a month, fully resolving 70% of customer inquiries coming through the company’s digital channels."
Claimed by: vendor - Interactions handled: about 45 million, per month
"Currently, TOBi handles nearly 45 million customer calls a month, fully resolving 70% of customer inquiries coming through the company’s digital channels."
Claimed by: vendor - Contact deflection: 12%, frequency of customer contacts to call centres, year over year after the TOBi launch
"Since launching TOBi, we’ve reduced the frequency of customer contacts to call centers by 12 percent year-over-year"
Claimed by: organization - Satisfaction uplift: 50%, per chatbot interaction in Italy, about one year after the generative AI version went live
"With millions of customers live we are getting 50% more satisfaction every time they interact with a chatbot – and 20% higher NPS in comparison with the previous AI."
Claimed by: organization
Air India
India · Travel and hospitality · 2023
Air India launched AI.g in May 2023, a virtual assistant on Azure OpenAI that is integrated with the reservation system and answers questions across 1,300 topic areas including bookings, flight status, baggage, check in, frequent flyer awards and lounge access, and escalates automatically to contact center staff when it detects the need. The airline says it has kept contact center call volume flat while its passenger count doubled.
- Automation rate: 97%, cumulative, of nearly 4 million queries
"To date, AI.g has successfully handled nearly 4 million customer queries, 97% of them with full automation."
Claimed by: vendor - Interactions handled: about 10,000, per day
"That's because AI.g is handling about 10,000 a day."
Claimed by: organization
DNB
Norway · Banking · 2023
DNB, Norway's largest bank, has worked with boost.ai since 2017 and runs several virtual agents. Aino, the customer service agent, is the main channel for visitors to DNB's website. In a boost.ai case study, the vendor says Aino automates over half of all incoming chat traffic. The study also describes Juno, a virtual agent launched in March 2020 that gives DNB's own customer service agents quick access to the routines they follow.
- Containment rate: at least 50%, share of incoming chat traffic
"Its customer service virtual agent, Aino, automates over 50% of all incoming chat traffic and is the primary channel for customers visiting DNB’s website."
Claimed by: vendor
Swiss Life Germany
Germany · Insurance · 2023
Swiss Life Germany replaced the touch tone menu of its sales and service hotline with a Parloa voice agent that takes the caller's request in their own words and routes it to the right expert. The pilot ran in one of four sales divisions and was then extended to all sales departments. Parloa reports 96% routing accuracy and that 73% of survey respondents rated the agent 4 or 5 out of 5.
- Accuracy: 96%, call routing
"96% Routing accuracy"
Claimed by: vendor - Customer satisfaction: 73%, survey respondents
"73% of all respondents rated the AI agent 4 or 5 (out of 5)"
Claimed by: vendor
Equity Bank Kenya
Kenya · Banking · 2022
Equity Bank Kenya officially commissioned EVA, the Equity Virtual Assistant, in May 2022 to replace EazzyNet's customer support role with a chatbot that runs on artificial intelligence and machine learning models. On WhatsApp, Facebook Messenger and Telegram, customers can chat with EVA to manage their accounts and carry out transactions such as money transfers, airtime purchases and viewing account balances, in addition to general customer support, and Equity says the assistant will keep learning from its interactions with customers.
No outcome disclosed.
Etihad Airways
United Arab Emirates · Travel and hospitality · 2022
Etihad Airways built a chatbot on Microsoft Power Virtual Agents and Dynamics 365 Customer Service to handle the surge in customer questions during the pandemic, including changing travel restrictions, refunds, cancellations and booking. By February 2022 the chatbot served 52 percent of customers on its own, freeing agents for more complex cases and reducing waiting time, and the airline reports a rise in its Net Promoter Score since the launch.
- Containment rate: 52%, February 2022
"In February 2022, the chatbot successfully served 52 percent of customers, and keeps continuously self-learning through AI technology."
Claimed by: vendor
Mobily
Saudi Arabia · Telecommunications · 2022
Mobily deployed customer facing AI agents on eight channels, including WhatsApp, Twitter and Apple Business Chat, connected to its internal systems. The agents answer billing, balance and data usage questions, change subscriptions, sell add ons, take payments and recharges, and handle feedback and complaints, with a warm handover to a specialist who can take over or hand back. NiCE Cognigy reports that the first response time fell from 20 minutes to about 6 seconds. The deployment was already live in 2022, when the case study described it as conversational AI; the current version presents it as agentic AI.
- Response time reduction: 99.5%, first response time on messaging channels
"An AI agent picks up any inquiry in around 6 seconds, reducing first response times significantly from the previous 20 minutes: a 99,5% improvement."
Claimed by: vendor
Commercial International Bank (CIB)
Egypt · Banking · 2021
CIB introduced the chatbot Zaki in December 2019 as part of a package of new digital products, an AI based assistant that answers customer questions about the bank's products and services in Arabic and English on the bank's website and its Facebook page. CIB's own review of its 2021 digital progress lists Zaki among its channels and says a business banking version lets commercial banking clients and non clients look up products and services, which the bank expects to take pressure off branches and customer service centers. No usage or resolution figures are disclosed.
No outcome disclosed.
Jordan Kuwait Bank
Jordan · Banking · 2021
In March 2021 Jordan Kuwait Bank announced an Arabic speaking assistant on WhatsApp, which it called the first in Jordanian banking. The assistant, MASA, answers general inquiries and frequently asked questions and passes a customer to a contact center agent when it cannot answer. Infobip supplied the WhatsApp Business API and a cloud contact center. The bank later received an industry award for MASA as an AI powered personal financial assistant inside its mobile banking app. No usage or containment figures are published.
No outcome disclosed.
Zain Jordan
Jordan · Telecommunications · 2021
Zain Jordan worked with Labiba, a Jordanian conversational AI vendor, to run a Messenger assistant that uses Arabic language processing, is connected to Zain's internal systems and covers account management, balance and coverage questions, with a handover to a live agent for complex requests (built with RingCentral). A 2021 press release, which presents Facebook's published success story, reports a 50% lower Messenger support cost and a 60% drop in calls to the customer care center. The figures come from the vendor's release; no baseline, period or sample is stated.
- Cost reduction: 50%
"After integrating the Messenger-powered digital assistant the average response time for customer queries dropped substantially, achieving 50% decrease in messenger support costs, and a 60% decrease in calls to the customer care center."
Claimed by: vendor - Contact deflection: 60%
"After integrating the Messenger-powered digital assistant the average response time for customer queries dropped substantially, achieving 50% decrease in messenger support costs, and a 60% decrease in calls to the customer care center."
Claimed by: vendor
Amen Bank
Tunisia · Banking · 2020
Amen Bank's chairman of the management board, Néji Ghandri, told L'Economiste Maghrébin in January 2026 that the bank launched an AI chatbot in 2020 and would deploy an improved version within weeks, built on language models and including Tunisian dialect. The bank plans to grow it into a financial coach covering investments, loans and international operations. He also said AI solutions are already validated in anti money laundering, the money market and credit scoring. No usage figures are published.
No outcome disclosed.
Attijariwafa bank
Morocco · Banking · 2020
Attijariwafa bank launched what it called Morocco's first banking chatbot in May 2020, during the Covid-19 pandemic, as an AI based virtual assistant on the bank's institutional website answering customer questions around the clock. At launch the assistant covered a limited scope in French, mainly pandemic related measures and an Attijarinet password reset, with the bank's own announcement saying future versions would add languages, new use cases and other channels. No usage or resolution figures are disclosed, and no later update confirming a wider rollout was found.
No outcome disclosed.
Etisalat Misr
Egypt · Telecommunications · 2020
Etisalat Misr introduced what it called its first AI virtual assistant in June 2020, built on IBM Watson Assistant on IBM public cloud. Customers can renew bundles, transfer and distribute a family monthly quota, check a balance and pay bills in English and Arabic on their preferred channels, around the clock, so that staff can work on more complex issues. The company's vice president of IT said the bot also carried COVID-19 information and daily case figures and helped keep operating costs down during the pandemic. No usage or resolution figures are disclosed.
No outcome disclosed.
Groupe Al Omrane
Morocco · Real estate · 2020
Groupe Al Omrane, the Moroccan state housing and urban development group, launched an AI based virtual assistant in June 2020 and called it the first multilingual, multiplatform chatbot in Morocco. It answers citizens' questions at any hour on the group's website, on Facebook and on WhatsApp, in French, classical Arabic and Darija, and supports the group's Customer Relations Centre. The launch was part of a digital program accelerated during the Covid-19 pandemic. No usage figures are disclosed.
No outcome disclosed.
Lufthansa Group
Germany · Travel and hospitality · 2020
During the pandemic, when passengers flooded call centers to change or cancel flights, Lufthansa Group replaced its in house chatbot with a conversational AI platform and built self service AI agents that manage rebookings, check alternative flights, give travel information and process refunds. The agents run on the airline websites and through SMS links that open a self service chat, with multilingual support and real time translation, and are used to absorb peaks such as strikes.
- Interactions handled: about 16 million, per year
"By leveraging AI-driven Self-Service Agents, the airline managed to significantly increase its interaction capacity, handling about 16 million conversations throughout the year with AI, with peak days seeing up to 375,000 interactions."
Claimed by: vendor
Mashreq
United Arab Emirates · Banking · 2020
In June 2020 Mashreq put an AI virtual agent from Avaya and Koopid in front of its contact center. A customer who calls Mashreq support is moved into a browser based chat with the bot, which authenticates the customer, interprets the enquiry with native language processing, reads the bank's back end systems and shows the details visually. When a person is needed, the case moves to an agent with the full chat history. Koopid and Mashreq mapped the bot's workflows to the main reasons customers call the contact center. The release states no usage or containment figure.
No outcome disclosed.
Vodafone Germany
Germany · Telecommunications · 2020
Vodafone Germany moved its messaging channels into one central team and put the TOBi chatbot in front of every WhatsApp, Apple Business Chat and SMS conversation, with a handover to a human agent on the same screen. TOBi understands more than 230 intents and can classify the photos and screenshots customers send, such as a bill or a router with a flashing red light. Genesys reports that the share of inquiries resolved by AI (which it calls first contact success) rose from 16% at launch to 44%.
- Containment rate: 44%, share of messaging inquiries resolved by TOBi without a human
"With TOBi fielding 100% of all messaging conversations, first-contact success rates increased from 16% on launch to 44%."
Claimed by: vendor
Bank ABC
Bahrain · Banking · 2019
Bank ABC introduced "Fatema," an AI powered digital employee built on Soul Machines' Digital DNA technology and announced as the first synthetic digital human Soul Machines deployed. Fatema uses an AI powered Digital Brain to engage personally with customers and was set to work alongside the team of Bank ABC's mobile only bank to offer personalized customer service, sharing news and updates on social media and letting customers talk to her by voice on specific digital channels. Bank ABC's Deputy Group CEO framed the launch as part of the bank's push into digital innovation in the region.
No outcome disclosed.
China Unicom Sichuan
China · Telecommunications · 2019
China Unicom's Sichuan provincial unit partnered with JD.com to put JD's AI customer service chatbot into its voice and WeChat channels. Callers to the customer service hotline are greeted by a natural sounding AI voice that directs them and can hand off to a human agent, and the same AI answers inbound questions on China Unicom's public WeChat account. JD also builds China Unicom Sichuan an AI based precision marketing capability as part of the same partnership.
No outcome disclosed.
JetBlue
United States · Travel and hospitality · 2019
JetBlue moved its customer support to an AI platform from late 2019, opening messaging channels (Apple Messages for Business, Google Business Messaging, web and app chat, WhatsApp) with Spanish language support, a virtual agent that resolves routine requests and AI assistance for the crewmembers who handle the rest.
In a January 2026 conference session published by the vendor, a JetBlue customer support leader described weather disruptions, when passengers ask for their options, and said the conversations crewmembers now handle (rebooking, refunds, alternatives weeks away) have many parts, which is why the airline has looked at AI that orchestrates several workflows. The ASAPP speaker in the same session warned against reading containment gains without checking whether customers still have the option to escalate.
- Containment rate: 45%, May 2023, virtual agent
"The integration of virtual agent experiences facilitated streamlined interactions and contributed to a remarkable 36% year-over-year growth in containment, with a 45% containment rate achieved in May 2023."
Claimed by: vendor - Hours saved: 73,000 hours, Q1 2023 only (one quarter, not annualized)
"In Q1 2023 alone, this AI-driven efficiency translated into significant savings of 73,000 workforce hours."
Claimed by: vendor
Emirates NBD
United Arab Emirates · Banking · 2018
Eva is the Emirates NBD virtual assistant for phone banking, launched in February 2017 and live on all major retail helplines by June 2017. It uses natural speech recognition in English and Arabic instead of touch tone menus, covers 150 call reasons and routes calls to the right agent skill. The Asian Banker reported in 2018 that it handled around 500,000 calls a month and cut wrong agent routing by 73%. This is an older deployment recorded as history; it is not tied to the 2023 to 2026 generative AI work.
- Error reduction: 73%, Wrong agent skill routing and customer recalls
"Positioning for the right agent to the customers is a key challenge for any bank but with EVA, the bank could reduce wrong agent skill routing and thus recalling of the customer by 73%."
Claimed by: organization - Interactions handled: about 500,000, Per month, 2018
"The assistant currently handles around 500,000 customer calls per month with a Recognition accuracy of around 90% for English and 86% for Arabic."
Claimed by: organization
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A contact reason report with volumes per intent and channel
- Approved, current knowledge articles with an owner and review date each
- A written list of intents and signals that always go to a human
- Historical transcripts to build test sets
Systems to integrate
- Telephony or contact center platform (routing, queues, transfer with context)
- Messaging channels (web chat, app, WhatsApp, social)
- Authentication (app confirmation, one time passcode, voice biometrics)
- CRM or case management, read only at first
- Knowledge management system
Complexity: Medium
Answering from knowledge is straightforward. The build needs clean, owned knowledge content, authentication on the phone, integration with the contact center platform for a warm handover, and a decision on which intents must never be contained.
- 1
Choose intents by volume and by risk
From the contact reason report pick the top general intents and write down, just as explicitly, the intents the agent must route and never contain: complaints, disputes, fraud, hardship and bereavement.
- 2
Clean the knowledge before you connect it
Retire duplicate and outdated articles, give every article an owner and a review date, and make the agent refuse when retrieval finds nothing. Many wrong answers are content problems.
- 3
Design the handover first
Agree with the contact center what a human receives (summary, identity, attempted steps) and how the customer keeps their place in the queue. Always offer a clear route to a person.
- 4
Test on real conversations
Build test sets from historical transcripts per intent, including angry customers, vulnerable customers and attempts to push the agent off policy, and run them on every change.
- 5
Launch in chat, then voice
Chat is easier to monitor and correct. Add voice once containment and satisfaction per intent are stable, and measure repeat contacts alongside containment.
Guardrails
- Answers only from approved knowledge, with a refusal and a handover when it is not covered
- Mandatory routing of complaints, disputes, fraud, hardship and vulnerability signals to people
- A visible route to a human at any time, on every channel
- AI disclosure at the start of every conversation
- Masking of personal and card data before text reaches a model or the logs
KPIs to instrument
- Containment per intent, counting a repeat contact within seven days as not contained
- Handover rate and handover reasons per intent
- Customer satisfaction on contained conversations versus human handled ones
- Time to first meaningful response and total time to resolution
- Complaints that mention the assistant
Human in the loop
Human agents take every conversation the agent routes and can take over live. A quality team reviews a weekly sample of contained conversations for fluent but wrong answers and signs off every new intent before it goes live.
Common failure modes
- Containment that is really abandonment
- Customers give up rather than get helped, which looks like success on the dashboard. Count repeat contacts and measure satisfaction per intent.
- Replacing people instead of routing to them
- Klarna reported that its assistant handled two thirds of chats, then said in 2025 that it had gone too far and began hiring human agents again so customers can always reach a person. Keep human capacity for the moments that matter.
- Confident wrong answers
- An agent that invents a policy creates liability, as the Air Canada tribunal case showed. Ground answers in approved content and refuse when unsure.
- Missed vulnerability
- A customer in distress is kept in automation. Detect vulnerability signals and hand over early.
What are the risks and rules?
EU AI Act
Limited risk (transparency)
An AI system that interacts directly with people must be designed so that they know they are dealing with AI, unless that is obvious from the context (Article 50(1)).
It is not high risk under Annex III as long as it does not evaluate eligibility for essential public assistance benefits and services (point 5(a)), creditworthiness (point 5(b)), risk and pricing for life and health insurance (point 5(c)) or emergency calls (point 5(d)). This holds only if emotion or vulnerability signals are inferred from what the customer says (text or transcript content), not from voice or other biometric features; an agent that infers emotion from a caller's voice is an emotion recognition system (Article 3(39)), which is high risk under Annex III point 1(c) and triggers the deployer disclosure duty in Article 50(3).
Guidance
- Regulation (EU) 2024/1689 (AI Act), Article 50: transparency obligations for providers and deployers of certain AI systems (European Union, Europe). People must be informed that they are interacting with an AI system unless this is obvious from the context.
- Chatbots in consumer finance (Consumer Financial Protection Bureau, North America). Warns that deficient chatbots that prevent access to live, human support can lead to law violations and customer harm.
- FG22/5: Final non-Handbook Guidance for firms on the Consumer Duty (Financial Conduct Authority, Europe). Says firms will likely need a real time human interface, such as a phone service, for security, fraud and other complex or sensitive journeys, and gives an automated phone system without a route to other support as an example of poor practice.
Controls to put in place
- Inventory entry with an accountable owner and a documented list of contained and routed intents
- Content governance with owners and review dates for every knowledge article
- Transcript logging and retention in line with record keeping rules
- Regression tests on every change to prompts, content or model
- Monitoring of outcomes for vulnerable customers and of complaint trends
When it went wrong elsewhere
- Incident 639: Air Canada Chatbot Reportedly Provides Inaccurate Bereavement Fare Information, Leading to Customer Overpayment. A Canadian small claims tribunal held the airline responsible for what its website chatbot told a customer about bereavement fares, a reminder that the organization owns every answer its agent gives.
Frequently asked questions
- What share of contacts can an AI agent resolve on its own?
It depends on the intent mix and the channel. Microsoft reports that about 84.6% of Commonwealth Bank's self service messaging interactions were resolved end to end in May 2026, and that Vodafone's TOBi fully resolves 70% of inquiries arriving through digital channels; Posh reports that Together Credit Union's voice agent contains 66% of inbound calls. Airbnb itself reports a lower figure: nearly 45% of issues that begin with its AI assistant were resolved without a human agent in Q2 2026. Treat these as upper references and count repeat contacts before you celebrate.
- Should we remove the option to speak to a person?
No. Klarna said in 2025 that customers must always have the option to speak with a human and began recruiting customer service agents again. The US Consumer Financial Protection Bureau warns that chatbots that block access to human support can break the law, and the UK Financial Conduct Authority expects a real time human route for complex or sensitive journeys.
- How is this different from account and card servicing?
The first line agent answers and routes general inbound contact across the whole contact center. Account and card servicing is the deeper, authenticated layer that performs transactions such as blocking a card. You can launch the first line first and add servicing actions behind it once authentication and handover work.
- Which banks use AI agents to answer customer calls and chats, and what results did they report?
Banks with reported results include ABN AMRO, Nordea, Vietcombank and NatWest. Microsoft says ABN AMRO's Anna supports more than two million text conversations and 1.5 million voice conversations a year and automates over 50% of interactions. Boost.ai reports a 91% in scope resolution rate for Nordea Sweden's private banking customers. FPT Smart Cloud says Vietcombank's VCB Digibot handled up to 88.5% of user requests after six months of deployment. NatWest says Cora handled 10.8 million queries over 2023. These figures come from vendors or the banks themselves, and they do not measure the same thing.
- Does a customer service chatbot need to disclose it is AI under the EU AI Act?
Yes, in most cases. Article 50(1) of the EU AI Act requires that people are told they are dealing with an AI system, unless that is obvious from the context. A first line agent is classed as limited risk, not high risk under Annex III, as long as it does not evaluate creditworthiness, eligibility for essential public assistance benefits and services, risk and pricing for life and health insurance, or emergency calls. That holds only if emotion or vulnerability is inferred from what customers say, not from their voice. An agent that infers emotion from a caller's voice is an emotion recognition system, which is high risk. Disclose at the start of every conversation.
- What is the ROI of an AI voice agent in a bank's contact center?
None of the banks in the evidence publishes an ROI figure for a voice agent, so none can be quoted. What is published are operating results. Posh says Together Credit Union's voice agent contains 66% of total inbound calls and that after hours escalations to its outsourced contact center dropped 14%. Kotak Mahindra Bank says its Keya voicebot more than halves the time to an accurate answer. BBVA México says Blue understands over 90% of requests from the start of a call. The share of general contacts and the cost per contact in the page's value model are editorial assumptions, so replace them with your own contact data.
How to cite this page
Blits.ai AI Use Case Library, "AI agent for first line contact center service", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/first-line-contact-centre-agent. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 11 November 2026: Added deployments: AOK PLUS, BLS AG, bunq, DATEV eG, Deutsche Telekom, Deutsche Rentenversicherung Hessen, E.ON Next, Essent, Finnair, KBC, KPN, Lascana, Movistar Plus+, Münchener Verein Versicherungsgruppe, Norwegian Air Shuttle, Ooredoo Oman, Ryanair, Salzburg AG, Bausparkasse Schwäbisch Hall AG, Swiss Life Germany, Swisscom, Trainline, Ministry of Finance (United Arab Emirates), Uelzener Versicherungs-Gruppe, Umniah, UmweltBank AG, University Hospitals Sussex NHS Foundation Trust, Württembergische Versicherung AG, Zain Jordan.
- 11 November 2026: Added deployments: Alizz Islamic Bank, Aman Bank, Amen Bank, Arab Bank, DNB, Emirates NBD, Garanti BBVA, Jordan Kuwait Bank, MONETA Money Bank, National Bank of Bahrain, Trust Bank.
- 14 October 2026: Added deployments: Banque Misr, Etisalat Misr, Commercial International Bank (CIB), Royal Air Maroc, AtlantaSanad Assurance, Groupe Al Omrane, Batelco, Khaleeji Bank, Dukhan Bank, Qatar Airways.
- 14 October 2026: Added deployments: Roads and Transport Authority, Dubai (RTA); Dubai Land Department (DLD); Mashreq.
- 14 October 2026: Added deployments: New Look.
- 1 October 2026: Added deployments: Vietcombank, China Unicom Sichuan, Kotak Mahindra Bank.
- 30 September 2026: Added deployments: ABN AMRO, Nordea, Poste Italiane, OTE Group, Magyar Telekom.
- 30 September 2026: Added deployments: Qatar Islamic Bank, Bank ABC, Ooredoo Qatar, Asiacell, Jordan Ahli Bank, Attijariwafa bank, Equity Bank Kenya, Zenith Bank, Vodacom, Kenya Power, BBVA México.
- 27 September 2026: First published