What problem does it solve?
Business customers buy complex bundles: fibre and dedicated lines at several sites, mobile fleets, devices, security and cloud services, each with its own availability, pricing rules and service levels. Small businesses want quick answers and quotes without waiting for a sales call; large accounts issue tenders with long questionnaires. Sellers spend hours researching accounts, checking availability and assembling quotes, and service teams handle routine requests that the customer could have done alone.
The knowledge needed lives in product sheets, price books, contract templates and past proposals spread across systems. Configure, price and quote tools help specialists, but they do not answer a small business owner's question at night or draft the first version of a tender response.
How does it work?
- Answer and qualify. On the business website or portal, the assistant answers questions about products, prices and contracts, checks basic eligibility such as fibre availability at an address, and qualifies the need.
- Configure and quote. Using the product catalogue and configure, price and quote tools, it prepares a configuration and an indicative quote within standard price rules.
- Hand over to a seller. Anything non standard (discounts, multi site designs, tenders) goes to a seller with the configuration, the conversation and the account context.
- Assist sellers. For sellers the assistant researches the account, suggests the next best offer, and drafts tender answers and proposals from approved content.
- Serve after the sale. Business customers raise and track orders and fault tickets, and the assistant resolves routine requests, keeping them informed during service interruptions.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Web chat, Agent desktop, Email, Internal tools
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 |
|---|---|---|---|---|
| Containment rate | Too few to pool | 70% | 1 | 1 organization |
| Conversion uplift | Too few to pool | 15% | 1 | 1 vendor |
| Interactions handled | Not pooled | about 100 | 1 | 1 vendor |
| Handling time reduction | Too few to pool | Not pooled: up to 20% | 0plus 1 up to | 1 vendor |
| Time saved per task | Too few to pool | Not pooled: up to 225 minutes | 0plus 1 up to | 1 vendor |
Value drivers: Revenue growth, Employee productivity, Lower cost to serve, Speed and cycle time.
Indicative value
A telecom operator's business unit with 50,000 small and medium business customers
USD 450,000 to USD 3.6 million
Sales and service handling cost avoided per year
How this is calculated
Formula: businessCustomers * inquiriesPerCustomer * resolvedShare * costPerInquiry. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Business customers businessCustomers, customers | 50,000 | 50,000 | The reference business unit. |
| Routine sales and service inquiries per customer per year inquiriesPerCustomer, inquiries per customer per year | 2 | 4 | Editorial assumption, replace with your own business contact volumes. |
| Share of routine inquiries the assistant resolves resolvedShare, fraction of inquiries | 0.3 | 0.6 | Conservative against the benchmark on this page (SoftBank reports a self resolution rate of 70% for its business website agent). |
| Cost of an inquiry handled by a seller or business service agent costPerInquiry, USD per inquiry | 15 | 30 | Editorial assumption; business inquiries take longer and are handled by more expensive staff. Replace with your own cost. |
What it leaves out: Counts only routine inquiries handled without a person. It leaves out additional sales from faster quotes and better prepared sellers, faster tender responses, the cost of the AI and the catalogue and quoting integrations, and discounts that still need approval.
Who already uses it?
5 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Telefónica España
Spain · Telecommunications · 2026
In June 2026 Telefónica España launched a marketplace for companies and government agencies to select and buy workplace devices from a catalogue of 160 models. Its GenIA virtual assistant lets buyers describe needs in natural language, compare alternatives, get personalised recommendations and resolve technical questions in real time, alongside a productivity calculator. No outcome figures were published.
No outcome disclosed.
Vodafone Business
United Kingdom · Telecommunications · 2025
Vodafone Business and ServiceNow agreed a five year collaboration to run service management for business customers on ServiceNow's telecom service assurance products with agentic AI. Vodafone says it will give a single view of each customer's networks and applications, and that it can detect and fix service anomalies within minutes rather than hours, using AI and machine learning to predict, minimise and manage service interruptions, and to route customers to the right department first time across online, email and phone. Vodafone reports that an initial deployment in Ireland raised digital channel use by 45%, and says satisfaction levels rose fourfold there.
- Satisfaction uplift: 4x, initial deployment in Ireland
"An initial deployment in Ireland led to a 45% rise in customers using digital channels, boosting satisfaction levels by 4x."
Claimed by: organization
SoftBank Corp.
Japan · Telecommunications · 2026
SoftBank's Customer Growth Division, which sells to Japan's small and medium sized businesses, put a conversational AI agent on its corporate website that answers questions about products, services, pricing and contracts. What it cannot resolve goes through an inquiry form to a sales representative. SoftBank says the self resolution rate, measured by the platform's own AI assessment, rose from about 50% at launch to 70%; Sierra puts the volume at about 100 inquiries a day.
- Containment rate: 70%, website inquiries from business customers
"SoftBank's corporate business covers a wide range of services, and the self-resolution rate* — which started at around 50% at launch — has now risen to 70%, thanks to iterative improvements to help the AI agent better understand customer intent and the relevant service."
Claimed by: organization - Interactions handled: about 100, inquiries per day
"Currently, the AI agent handles approximately 100 inquiries per day, with 70% resulting in customers finding the information they need and resolving their inquiries on their own."
Claimed by: vendor
Lumen Technologies
United States · Telecommunications · 2024
Lumen's sellers use Microsoft Copilot to summarize past sales interactions, gather recent news, identify business challenges and industry trends, and suggest next steps for an account. Microsoft reports that this work took up to four hours per seller and that Lumen cut it to 15 minutes in 2024. Lumen's projected annual value of USD 50 million is a projection and is not recorded as a result.
- Time saved per task: up to 225 minutes, summarizing past interactions and researching an account, per seller
"This process traditionally took up to four hours per seller. In 2024, Lumen reduced that time to just 15 minutes, projecting annual time savings worth USD50 million."
Claimed by: vendor
Verizon
United States · Telecommunications · 2022
Verizon's business unit built a single Next Best X decisioning engine on Pega across five lines of business and more than 200 products and offers. It brings AI guided selling into the tools that front line sellers and business customers already use, across sales cycles that range from six month account based programmes to real time transactional sales for small businesses. Pega reports a 15% improvement in win rate, a doubling of the attach rate for value added services and a 10 to 20% reduction in handling time.
- Conversion uplift: 15%, sales win rate
"15% improvement in win rate"
Claimed by: vendor - Handling time reduction: up to 20%
"10-20% improvement in handling time reduction"
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Current business product catalogue, price books and discount rules
- Address level availability for fixed products
- Approved contract templates, service levels and past tender answers
- Account data in the CRM, including installed base and contract dates
Systems to integrate
- CRM and account management
- Configure, price and quote system
- Availability and serviceability checks
- Service management and ticketing for business customers
- Document and proposal repositories
Complexity: Medium
Answering product questions is easy once content is clean. Quoting needs the product catalogue, availability checks and price rules behind APIs, and binding offers must stay in the approved configure, price and quote process.
- 1
Start with questions and qualification
Put an assistant on the business website that answers product, price and contract questions from approved content and routes qualified leads, as SoftBank did first because it had the most customer touchpoints.
- 2
Connect availability and indicative quotes
Add address availability and standard price quotes through tools, and keep every binding quote and discount in the existing approval process.
- 3
Give sellers an account and tender assistant
Let sellers ask for account research, next best offers and draft tender answers from an approved library, and measure time saved per proposal.
- 4
Bring service into the same assistant
Let business customers raise and track orders and tickets and get status during service interruptions, so the relationship does not end at the sale.
- 5
Improve from unresolved questions
Review the questions the assistant could not answer each week and add the missing content. SoftBank credits this kind of iterative improvement for raising its self resolution rate from about 50% at launch to 70%.
Guardrails
- Prices, discounts and availability only from tools, never from the model
- Binding quotes and non standard discounts approved by a seller
- Tender answers drawn only from an approved answer library, with a seller reviewing every submission
- Customer contract data visible only to authorised users of that account
- Handover to a named seller or service agent on request
KPIs to instrument
- Self resolution rate of business inquiries, checked against repeat contacts
- Time from inquiry to quote
- Win rate of assisted quotes versus unassisted ones
- Seller hours spent on research and tender responses
- Quote errors found at approval
Human in the loop
Sellers approve every binding quote, discount and tender response, and service specialists handle complex faults and escalations. Product and pricing teams own the content the assistant uses and review unresolved questions weekly.
Common failure modes
- Quotes the operator cannot honour
- The assistant quotes a price or service that is not available at the address. Check availability and price in tools and label quotes as indicative.
- Tender answers that overpromise
- A generated answer commits to a service level the operator does not offer. Use an approved library and seller review.
- Leaking account data
- A user sees another company's contract details. Enforce account level access in the tools.
- Leads lost in handover
- Qualified leads sit in an inbox. Route them to named sellers with service levels for follow up.
What are the risks and rules?
EU AI Act
Limited risk (transparency)
Article 50(1): people must be informed that they are interacting with an AI system, unless that is obvious from the context. Quoting, sales support and service for business customers are not listed in Annex III. The use would become high risk under Annex III point 5(b) only if the assistant itself evaluated the creditworthiness of a natural person, such as a sole trader, to decide whether to offer a contract.
Rules that apply
Guidance
- 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.
- European Electronic Communications Code (Directive (EU) 2018/1972) (European Union, Europe). Several end user protections, such as contract information, also apply to microenterprises and small businesses unless they waive them, so a B2B assistant cannot treat every business customer as a large enterprise.
Controls to put in place
- AI disclosure on the business website and portal
- Price and discount approval workflow outside the model
- Approved answer library with owners for tender content
- Account level access control and audit logging
- Weekly review of unresolved questions and quote errors
Frequently asked questions
- Can business customers really self serve with an AI agent?
- For routine questions, yes. SoftBank reports that the self resolution rate of its website agent for small and medium business customers rose from about 50% at launch to 70%. Sierra, the platform vendor, measures that rate with its own AI monitoring and puts the volume at about 100 inquiries a day. Complex designs and discounts still go to sellers.
- What does AI do for B2B sellers?
- Mostly research and preparation. Microsoft reports that Lumen cut account research that took a seller up to four hours to 15 minutes with Copilot, and Pega reports a 15% improvement in win rate from Verizon's AI guided selling engine.
- Should the assistant produce binding quotes?
- Indicative quotes, yes; binding ones, no. Keep discounts and binding offers in the approved quoting process, with a seller signing off.
How to cite this page
Blits.ai AI Use Case Library, "AI assistant for B2B telecom quoting, sales and service", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/business-connectivity-quoting-and-service-assistant. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published