What problem does it solve?
Banks see a detailed picture of a customer's financial life in their transaction data. A customer who is surprised by a direct debit, or who does not know how much they can safely save, is often left to work it out from a list of transactions.
Forrester's 2025 review of the mobile apps of the four biggest Australian banks found that most still fall short in helping customers improve their overall financial health, while leading banks increasingly use AI powered insights to help customers stay on top of their finances. The opportunity is a coach that does the analysis for the customer, speaks up at the right moment and can act on a simple instruction, without drifting into selling or unlicensed advice.
- Forrester's 2025 review of Australian mobile banking apps found that most banks still fall short in helping customers improve their overall financial health.Conversational AI And Anticipatory Insights, What's New In Australian Mobile Banking In 2025 (2025)
How does it work?
- Understand the customer's money. Models categorise transactions, detect income, bills and subscriptions, and forecast the balance over the coming days.
- Speak up at the right moment. Proactive insights flag a bill that will not be covered, a subscription price rise or a month of unusual spending, with an action attached.
- Answer in conversation. The customer asks "can I afford this trip" or "where did my money go" and gets an answer grounded in their own data and the bank's approved guidance content.
- Act within limits. On instruction it sets up a savings goal, a transfer into a savings pot or a budget, using the same authenticated APIs as the app, and confirms before moving money.
- Know the boundary. Questions that need regulated advice (investments, pensions, debt solutions) or show signs of financial difficulty go to a human or to the right service.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Mobile app
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 | 10 million to 3 billion | 3 | 3 organization |
| Users served | Not pooled | 900,000 to 50 million | 2 | 2 organization |
Value drivers: Customer experience, Revenue growth, Inclusion and access.
Indicative value
A retail bank with 1 million digitally active customers
USD 50,000 to USD 500,000
Revenue retained through lower attrition per year
How this is calculated
Formula: customers * engagedShare * churnPoints * valuePerCustomer. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Digitally active customers customers, customers | 1,000,000 | 1,000,000 | The reference bank. |
| Share of customers who use the coach regularly engagedShare, fraction of customers | 0.1 | 0.25 | Editorial assumption. For scale, RBC reports more than 900,000 clients used NOMI Forecast in its first 19 months. |
| Reduction in annual attrition among engaged users churnPoints, fraction of engaged customers per year | 0.005 | 0.01 | Editorial assumption; no deployment on this page discloses a retention effect. Measure it with a control group. |
| Annual revenue of a retained main bank customer valuePerCustomer, USD per customer per year | 100 | 200 | Editorial assumption, replace with your own customer economics. |
What it leaves out: Retention value only, and the most uncertain input is the attrition effect. It leaves out fees customers avoid (a benefit to them, not the bank), deposit growth from savings features, contact centre calls avoided and the cost of building and running the coach.
Who already uses it?
6 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Hyundai Card
South Korea · Payments and cards · 2026
Hyundai Card, a South Korean credit card issuer, has published an annual statement (연간명세서) in its app since 2021 that summarizes each member's yearly card spending. For the 2025 edition, released in January 2026, Hyundai Card applied an AI agent system it built itself, described as generative AI based on a large language model that runs a predesigned workflow. According to the company, the agent was used across the whole production run: analysing the payment data of 12.6 million members, generating a personalized message for each member and reviewing the results. The statement is a spending review with narrative, persona based insights and peer comparisons, not a legal account statement or tax document.
- Interactions handled: 12.6 million, members whose payment data the AI agent analysed for the 2025 annual statement, with a personalized message generated and reviewed per member
"이번 연간명세서 제작 과정에서는 1260만 회원의 결제 데이터 분석, 회원별 개인화 메시지 생성, 결과 검수까지 전 과정에 AI 에이전트가 사용됐다."
Claimed by: organization
Starling Bank
United Kingdom · Banking · 2026
Starling Assistant is an agentic AI assistant in the Starling app that responds to text and voice, analyses spending patterns, creates savings Spaces and sets up transfers on the customer's behalf. In August 2026 Starling added "smart tools" to it and said new ones would follow every week for the rest of 2026 and at least monthly after that. The launch set includes a tax saver that sweeps a share of the transactions a small business picks into a Space, a Making Tax Digital guide, a spending quiz and a student budget planner. A rainy day saver, which works out with the customer how much they can realistically save and sets up transfers into a savings Space, was announced as a follow up tool. The assistant is built on Google's Gemini models. No outcome figures are disclosed.
No outcome disclosed.
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 centre 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
Royal Bank of Canada
Canada · Banking · 2023
RBC's NOMI is a set of AI features in the RBC Mobile app and RBC Online Banking that give clients personalized insights about their money. NOMI Forecast, built with the bank's research centre Borealis AI, uses deep learning to show a seven day view of upcoming preauthorized payments and cash flow; NOMI Find and Save helps clients put money aside; NOMI Budgets tracks spending. RBC says clients using Find and Save have put aside more than CAD 3.6 billion.
- Users served: at least 900,000, NOMI Forecast, September 2021 to April 2023
"Since its launch in September 2021, more than 900,000 clients have used the feature."
Claimed by: organization - Interactions handled: at least 10 million, NOMI Forecast, 2021 to April 2023
"The addition of NOMI Forecast has led to more than 10 million client interactions since 2021."
Claimed by: organization - Customer savings: at least CAD 3.6 billion, NOMI Find and Save, cumulative as of April 2023
"Clients using NOMI Find & Save have put aside more than $3.6 billion in savings."
Claimed by: organization
Commonwealth Bank of Australia
Australia · Banking · 2022
Commonwealth Bank's Customer Engagement Engine (CEE), built on Pega Customer Decision Hub, suggests in real time the next best conversation to have with each customer, whether in the branch, on the phone, online or on a mobile device. Beyond suggesting conversations, the bank uses it to match customers to government benefits and rebates they may be missing (Benefits finder) and to reach customers hit by natural disasters with same day support such as a loan deferral. The same decisions feed digital channels and prompts for branch and contact centre staff.
No outcome disclosed.
Westpac
Australia · Banking · 2025
Forrester's 2025 review of Australian mobile banking apps, as reported by Mi3, ranked Westpac first for the third year running and credited AI features that nudge customers toward better financial decisions. The same coverage says that the big four Australian banks still fall short of what customers most want, including timely alerts and personalised guidance. This is an analyst assessment of the app, not a result published by the bank.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Transaction history with reliable merchant and category enrichment
- Scheduled payments, direct debits and income patterns
- Approved guidance content on budgeting, saving and financial difficulty
- Customer consent and preferences for proactive messages
Systems to integrate
- Core banking and payments data (read) and savings and transfer APIs (write, with confirmation)
- Transaction enrichment and categorisation service
- Notification and in app messaging
- Referral routes to advice, debt support and the contact centre
Complexity: Medium
Categorisation and forecasting models are well understood; the difficulty is quality (a wrong forecast destroys trust), the boundary with regulated advice, and making insights useful rather than noisy.
- 1
Start with forecasting and alerts
A reliable view of upcoming bills and the projected balance is useful on its own and needs no conversation. RBC's NOMI Forecast (a seven day view of upcoming payments) and Erica's alerts on where balances are trending over the next seven days at Bank of America are examples.
- 2
Add conversation over the customer's own data
Let customers ask about their spending and plans, with answers computed from their data by deterministic functions and explained by the model, never estimated by it.
- 3
Draw the advice line in writing
With compliance, list what the coach may say (facts, general guidance, the bank's own product features) and what triggers a referral (investment, pension or debt advice, signs of financial difficulty).
- 4
Let it act with confirmation
Add savings goals and transfers between the customer's own accounts, each confirmed by the customer, before anything more autonomous.
- 5
Measure outcomes, not clicks
Track fees avoided, savings built and financial difficulty referrals against a control group, not just insight views.
Guardrails
- Numbers come from deterministic calculations on the customer's data, never from the model's own arithmetic
- A written boundary between guidance and regulated advice, with automatic referral when it is crossed
- No sales messages disguised as coaching; product offers are labelled and follow marketing consent
- Signs of financial difficulty trigger support routes, not product offers
- Any money movement is confirmed by the customer and limited to their own accounts
KPIs to instrument
- Regular users and repeat use of insights
- Forecast accuracy on upcoming balances
- Fees avoided and savings built by users versus a control group
- Referrals to advice and financial difficulty support
- Complaints and satisfaction about the coach
Human in the loop
Customers approve every action. Humans take over for regulated advice and financial difficulty. A conduct and quality team reviews samples of insights and conversations for accuracy, tone and any drift towards selling.
Common failure modes
- The coach becomes a sales channel
- Insights turn into product pushes and customers stop trusting them. Separate coaching from offers and review the mix.
- Wrong numbers
- A misclassified income or a missed direct debit gives a wrong forecast. Compute with deterministic functions, show the basis and let customers correct it.
- Advice by accident
- The assistant recommends an investment or a debt product in a way that counts as regulated advice. Enforce the boundary with guardrails and tests.
- Alert fatigue
- Too many nudges and customers mute them all. Cap frequency and measure action rates per insight type.
What are the risks and rules?
EU AI Act
Depends on design
The conversational assistant carries the Article 50 transparency duty: customers must be told they are interacting with an AI system. The system becomes high risk if it is used to evaluate the creditworthiness of natural persons or establish their credit score (Annex III point 5(b)). Article 5(1)(b) prohibits AI that exploits vulnerabilities due to a person's specific social or economic situation to materially distort their behaviour in a way that causes, or is reasonably likely to cause, significant harm.
Guidance
- Advice Guidance Boundary Review (Financial Conduct Authority, Europe). The joint HM Treasury and FCA review of the boundary between financial advice and other forms of support. Its targeted support rules, confirmed as final on 26 February 2026 and expected by the FCA to take effect from 6 April 2026, let firms that hold the new targeted support permission suggest options on pensions and retail investments to groups of customers with common characteristics, which bears on how far a coach may go.
- Article 5: Prohibited AI practices (European Union, Europe). Bans manipulative techniques and the exploitation of vulnerabilities, including those due to a person's economic situation.
Controls to put in place
- Inventory entry for the coach and its personalisation models with an accountable owner
- Documented guidance and advice boundary signed off by compliance
- Fairness monitoring of insights and referrals across customer segments
- Accuracy monitoring of forecasts and categorisation
- Consent management for proactive messages and use of data
Frequently asked questions
- Is a financial wellbeing assistant giving financial advice?
- It should not be. Explaining a customer's own data, general guidance and the bank's own features is guidance; recommending a specific investment or debt product for a customer's circumstances can be regulated advice. Write the boundary down and test it.
- Which banks run AI financial coaches today?
- Bank of America's Erica has delivered more than 1.7 billion proactive, personalized insights (from a predefined set of responses, without generative AI), and RBC's NOMI forecasts cash flow and helps clients put money aside with Find and Save. Starling added smart tools to its agentic assistant in August 2026 and says it will release new ones every week for the rest of 2026.
- How do you prove it helps customers?
- Compare users with a control group on fees avoided, savings built and financial difficulty outcomes, not on clicks. Engagement alone can hide a coach that mostly sells.
How to cite this page
Blits.ai AI Use Case Library, "AI financial wellbeing coach in the banking app", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/financial-wellbeing-coach. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published