AI chatbots in banking are conversational assistants that let customers check balances, move money, dispute charges and get explanations in natural language, by text or voice. Every one of the ten largest US commercial banks runs one, the CFPB found in 2023. The best documented are Bank of America's Erica, with more than 3.2 billion interactions since 2018, Wells Fargo's Fargo and Capital One's Eno. No chatbot-specific banking rule exists: consumer law, third-party risk guidance and state disclosure laws apply.
Which banks have deployed AI chatbots and virtual assistants?
The banks' own newsrooms document at least 8 named customer-facing assistants among the largest US banks. Bank of America's Erica (2018) is the most used, with more than 3.2 billion client interactions; Wells Fargo's Fargo (2023) passed 1 billion in under three years; Capital One's Eno dates from 2017 and its agentic Chat Concierge from 2025; U.S. Bank, Truist, Fifth Third and Citizens run their own. The table gives the launch year and the usage figures each bank has published, with the source.
| Assistant | Bank | Launched | What it does | Published usage | Source |
|---|---|---|---|---|---|
| Erica | Bank of America | 2018 | AI virtual financial assistant in the mobile app: answers client questions from a library of more than 700 responses, sends proactive personalised insights and schedules appointments; also guides Merrill clients on investment topics. | More than 3.2 billion client interactions since launch; 20.6 million users interacted nearly 700 million times in 2025. | Bank of America newsroom, Mar 10, 2026 ↗ |
| Fargo | Wells Fargo | 2023 | AI-powered virtual assistant in the Wells Fargo Mobile app, built with Google Cloud AI: Zelle payments, bill pay, routing numbers, spending and balance insights, in English and Spanish. | More than 1 billion interactions in under three years; more than 3 million Spanish-speaking customers, over 160 million interactions. | Wells Fargo newsroom, Mar 26, 2026 ↗ |
| Eno | Capital One | 2017 | Capital One's assistant, unveiled at SXSW 2017 as a text-message bot with in-house natural-language processing; now in the mobile app, notifications and the desktop browser, where it creates virtual card numbers and flags unusual charges. | No interaction total published. Capital One said in 2018 that Eno understood more than 2,200 terms and emojis customers use to ask for a balance. | Capital One Tech, Sep 26, 2018 ↗ |
| Chat Concierge | Capital One | 2025 | Multi-agent assistant on car dealers' websites through Auto Navigator: compares vehicles, schedules test drives, hands buyers to sales staff. | No usage figures published by the bank. | Capital One Tech, Mar 5, 2025 ↗ |
| U.S. Bank Smart Assistant | U.S. Bancorp | 2020 | Voice-first assistant in the U.S. Bank Mobile App: customers speak or type requests such as sending money with Zelle. | No usage figures published in the launch announcement. | U.S. Bank, Jul 23, 2020 ↗ |
| Truist Assist | Truist Financial | 2022 | AI-enhanced virtual assistant in mobile and online banking using natural-language understanding, with a handoff to a contact-centre teammate. | Up to 440,000 conversations a month. | Truist investor relations, Dec 18, 2025 ↗ |
| Jeanie | Fifth Third Bancorp | Not stated in the bank's release | AI-powered banking assistant in the Fifth Third mobile app handling more than 150 banking needs through natural-language requests, including in Spanish. | Self-service containment up from 3% to 42% and intent-recognition accuracy from about 20% to about 90% since launch. | Fifth Third media center, Aug 12, 2026 ↗ |
| CiZi | Citizens Financial Group | Not stated on the bank's page | AI-powered assistant in the Citizens mobile app: answers questions, walks customers through tasks and completes certain transactions on their behalf. | No usage figures published by the bank on its product page. | Citizens, Mar 31, 2026 ↗ |
Only assistants the bank itself has named and described are listed, and only usage figures the bank has published. Internal assistants for employees are on the bank pages; agents that act for customers are covered in AI agents in banking.
What do regulators expect of AI chatbots in banking?
No US banking rule is written for chatbots; the expectations come from consumer law applied to the channel. The CFPB's June 2023 issue spotlight found that all ten of the largest US commercial banks used chatbots and that about 37% of the US population interacted with a bank chatbot in 2022, and it warned that a chatbot that gives inaccurate information, fails to recognise a dispute or traps customers in 'doom loops' without a route to a human can violate federal consumer financial law. States are adding disclosure duties: Colorado's HB 26-1263 requires conversational AI services to say they are AI from January 1, 2027, and New York's DFS already requires that of virtual-currency firms.
| Rule | Authority | What it requires | Applies |
|---|---|---|---|
| CFPB issue spotlight: Chatbots in consumer finance | CFPB | Institutions remain responsible for accurate answers, recognising disputes and giving access to a human; deficient chatbots risk non-compliance with federal consumer financial law. | Published Jun 6, 2023 |
| CFPB comment to Treasury on AI | CFPB | 'No fancy new technology carveout': chatbots named among the areas of compliance risk under existing law. | Published Aug 12, 2024 |
| Colorado HB 26-1263 (conversational AI) | Colorado AI Act | Operators of publicly accessible conversational AI must disclose that the user is talking to AI, keep safety protocols and not present output as coming from a licensed professional. | Effective Jan 1, 2027 |
| Colorado AG proposed rules | Colorado AI Act | Draft rules implementing the ADMT Act and HB 26-1263. | Proposed Aug 11, 2026 |
| NY DFS customer-service guidance for virtual currency entities | NY DFS | Tell the customer at the start that they are dealing with an AI tool and let them escalate to a human. | In force from May 30, 2024 |
| Interagency third-party risk guidance (SR 23-4 / OCC 2023-17 / FIL-29-2023) | Federal Reserve | A chatbot bought from a vendor, or built on a vendor's language model, is a third-party relationship: due diligence, contract terms, ongoing monitoring. | In force from Jun 7, 2023 |
| NIST AI 600-1 (Generative AI Profile) | NIST | Confabulation, data privacy and information security among the twelve generative-AI risks, with suggested actions. | Voluntary |
| GDPR Article 22 | EU AI Act | For EU customers: limits on decisions based solely on automated processing, relevant once an assistant decides rather than informs. | In force from May 25, 2018 |
What are the risks of AI chatbots in banking?
The CFPB's review grouped them in three: a limited ability to solve complex problems (failing to recognise a dispute, giving inaccurate or insufficient information, offering no meaningful help), obstacles to timely human intervention, and technical and security weaknesses (downtime, impersonation and phishing, exposure of personal information). Generative models add confabulation, which NIST names among its twelve generative-AI risks. For the bank the consequences are legal as well as reputational: the CFPB says deficient chatbots risk non-compliance with federal consumer financial law, diminished service and trust, and harm to customers.
| Risk | What goes wrong | Who has flagged it |
|---|---|---|
| Unrecognised disputes | A customer raises an error or dispute in chat and the bot does not treat it as one, so legal timelines are missed. | CFPB issue spotlight (2023) |
| Inaccurate or insufficient information | Wrong answers about fees, terms or account status that the customer relies on. | CFPB issue spotlight (2023); NIST AI 600-1 (confabulation) |
| Doom loops | Repetitive, unhelpful responses with no offramp to a human representative. | CFPB issue spotlight (2023) |
| Impersonation and phishing | Fraudsters imitate a bank's chatbot to harvest credentials. | CFPB issue spotlight (2023) |
| Personal data exposure | Chat logs and prompts hold personally identifiable information that must be protected whatever the technology. | CFPB issue spotlight (2023) |
| Undisclosed AI | A customer does not know they are talking to a machine. | Colorado HB 26-1263; NY DFS guidance for virtual currency entities |
| Vendor dependence | The assistant runs on a third party's platform or language model the bank must still oversee. | SR 23-4 / OCC Bulletin 2023-17 / FDIC FIL-29-2023 |
SOURCES: CFPB, Chatbots in consumer finance (June 6, 2023) ↗ · NIST AI 600-1, Generative AI Profile ↗
Should a bank build or buy its AI chatbot?
The public record shows both routes at the largest banks. Capital One built Eno's natural-language processing in-house, saying it needed a platform that 'deeply understands financial services terms'; Bank of America describes Erica as trained by its own data scientists with a library of more than 700 responses; Wells Fargo built Fargo on Google Cloud's conversational AI. The regulatory difference is where the oversight burden lands. A built assistant is the bank's to develop, test and govern end to end; a bought one is a third-party relationship under the 2023 interagency guidance, and the CFPB holds the bank responsible for what the chatbot tells customers either way.
| Route | Named example | What the bank takes on | Rule that bites |
|---|---|---|---|
| Build in-house | Eno (Capital One): NLP built in-house from concept to launch in under three months, trained on customer-service chat logs | Own development, testing, monitoring and the talent to sustain them | Enterprise AI governance; consumer law on the answers given |
| Build on a cloud provider's AI | Fargo (Wells Fargo): Google Cloud's Dialogflow conversational AI | Design and controls stay with the bank; the platform is a critical vendor | SR 23-4 / OCC Bulletin 2023-17 third-party risk management |
| Proprietary, curated responses | Erica (Bank of America): more than 700 responses, over 75,000 updates since launch | Tight control of what the assistant can say, at the cost of flexibility | CFPB expectations on accuracy and access to a human |
| Agentic, acting for the customer | Chat Concierge (Capital One): multi-agent assistant that schedules and hands off | Controls on actions, disclosure that the customer is dealing with AI, human handoff | Colorado HB 26-1263 from Jan 1, 2027; CFPB spotlight |
SOURCES: Capital One Tech: How and why Capital One built Eno's NLP in-house (2018) ↗ · Wells Fargo newsroom: Fargo to be powered by Google Cloud AI (2022) ↗ · Bank of America newsroom: Erica surpasses 3 billion client interactions (2025) ↗ · Capital One Tech: Chat Concierge (2025) ↗
Which chatbot-related documents has each authority published?
| Date | Event | Document |
|---|---|---|
| Aug 11, 2026 | Colorado AG proposes rules implementing the chatbot and ADMT laws | Colorado AG proposed ADMT rules |
| May 29, 2026 | Colorado HB 26-1263 signed: conversational AI disclosure from January 1, 2027 | HB 26-1263 |
| Aug 12, 2024 | CFPB to Treasury: 'no fancy new technology carveout' | CFPB comment to Treasury on AI in financial services (2024) |
| Jul 26, 2024 | NIST generative-AI profile names confabulation among twelve risks | NIST AI 600-1 (Generative AI Profile) |
| May 30, 2024 | NY DFS: virtual currency firms must disclose AI in customer service and allow escalation | DFS Virtual Currency Customer Service Guidance (May 2024) |
| Jun 6, 2023 | CFPB issue spotlight: all ten largest banks use chatbots; 'doom loops' warning | CFPB Chatbots in Consumer Finance (issue spotlight, 2023) |
Which of the 100 largest US banks use AI in customer service?
49 of the 100 bank pages on this site record a customer-facing AI use case, from named assistants to contact-centre tools and personalised insights. Each links to the bank's page, where the claim is sourced.
| Bank | What the record shows | Status |
|---|---|---|
| JPMorgan | Call-centre and client assistants: Generative AI in call centres for agents serving Chase customers; AI concierge for clients is the stated end state. | Rolling out |
| BofA | Erica: Consumer virtual assistant; 2 million interactions a day. | In production |
| Citi | Citi Sky: AI wealth advisor for Citigold clients, phased US rollout from summer 2026; demonstrated handling CD renewals, college-fund checks and mortgage-refinancing prompts as an 'always-on' member of the Citi Wealth team. | Rolling out |
| Wells Fargo | Fargo: Consumer virtual assistant: bill pay, transfers, transaction lookups; 245 million interactions in 2024. | In production |
| Capital One | Eno: Customer assistant in the mobile app, by text, in notifications and in the desktop browser: answers account questions, watches for unusual charges, free-trial endings and recurring charges, and creates virtual card numbers for online shopping. Natural-language processing built in-house. Digital and customer-service AI: Part of the enterprise AI program presented at re:Invent. | In production |
| PNC | Call-centre models: Small and large models for call-centre tasks, built in-house. | Announced |
| Truist | Truist Insights: Personalisation engine delivering proactive insights and advice in digital channels. AI Call Summaries: Automatic generative-AI note-taking for care-centre conversations, piloted in Q3 2025 and fully deployed by September 2026; 70% of inbound interactions supported. | In production |
| Schwab | Portfolio Insights: Gen-AI explanations of portfolio performance with Schwab research and market news. NLP models in client service: Natural-language processing supporting service interactions. | In production |
| TD | TD AI Prism personalisation: Predictive foundation model for customer needs and marketing. | In production |
| Fifth Third | Jeanie and the AI app interface: AI assistant plus natural-language navigation with 150 intents. AI deposit personalisation: AI tools on customer data to tailor deposit offers. | In production |
| Huntington | AI-personalised deposit offers: AI on customer data to tailor deposit products. | In production |
| BMO | Next Best Offer with Codat: Machine learning on payments data recommends products to bankers. | In production |
| Citizens | CiZi: Mobile-app AI assistant; call volume down 44%. | In production |
| Amex | Graphite card agentic search and checkout: Agentic AI supporting product search and checkout for business customers. | In production |
| KeyBank | Conversational AI in the contact centre: AI-handled calls at roughly $0.25 each versus $9 for human-handled. Enriched transactions and personalised insights (Personetics): Enrichment live since April 2026; insights piloting for Q3; conversational assistant not yet dated. | In production, Rolling out |
| Ally | Call summarisation for customer care: Gen-AI summaries of customer calls; about 5 million calls supported. | In production |
| HSBC | Contact-centre AI tools: Rolled out to improve customer experience, per the group CEO. | In production |
| Regions | Personalised insights in mobile banking: Transaction analysis for daily spending and savings notifications and subscription tracking. | In production |
| Pinnacle | Digital-wallet primacy through advice and alerts: Spending advice and fraud alerts as the basis for card use in digital wallets. | In production |
| UBS | STAAT proactive insights for US advisors: Machine-learning alerts on external assets, life events and network connections. AI avatars of research analysts: Digital twins delivering AI-generated research scripts to clients. | In production |
| Santander | Hyper-personalised customer journeys (ONE Transformation): Data and AI embedded in the businesses for cross-selling and primacy. Openbank digital banking: The digital bank that gathered $6 billion of US deposits in its first year. | Rolling out, In production |
| City National | AI-driven product innovation: Led by the SVP for product, AI and innovation; specifics undisclosed. | Announced |
| Zions | Automation of service emails: 330,000 annual address-change and similar emails targeted for AI handling. | Announced |
| Flagstar | Customer 360 and proactive private-banking insights: The S2 architecture's stated customer-facing goal. | Announced |
| First Horizon | In-house digital banking platform: A proprietary customer-facing platform under construction. | Announced |
| East West | Mobile and online banking: Digital account opening and CD products via the mobile app. | In production |
| Old National | Embedded BNPL in the banking app (Affirm, in talks): Affirm Edge would surface purchasing power and installment offers inside the app. | Announced |
| CIBC | Core-tex personalised offers and insights: AI-generated product offers and timely client insights. | Rolling out |
| Valley | AI prospecting and lead generation: Helping bankers develop relationships 'in a not-creepy way'; AI used to compete for deposits. | In production |
| Frost | Conversational, personalised mobile banking: Stated direction: customers describe tasks in their own words; not yet launched. | Announced |
| FNB | eStore Common app with AI-driven insights: One application for 50-plus consumer and business products; AI reduces keystrokes and time. AI customer insights and recommendations: Analyses eStore transactions, suggests products and cash-flow improvements. | In production, Announced |
| SoFi | SoFi Coach: Chat-based AI financial guide on member and external account data; planner-built. Cash Coach: AI insights engine that recommends moving idle cash and paying down high-rate debt. Conversational AI support (Galileo Konecta): 50% lower chat abandonment, 65% faster responses. | Rolling out, In production |
| Associated | Digital and omnichannel strategy: Digital sales and servicing under the new chief digital officer. | Rolling out |
| Deutsche Bank | Conversational AI for inbound inquiries: Target: more than 80% of inbound inquiries assisted by personalised conversational AI. | Rolling out |
| Hancock Whitney | Digital and mobile modernisation with MX: Money-experience platform; new mobile app; online account opening. | Rolling out |
| BankUnited | 'Digital plus human' service model: Digital self-service paired with bankers for complex, high-value work. | In production |
| Texas Capital | Cloud-native client onboarding: Integrated platform cited for onboarding speed. | In production |
| Eastern Bank | Client education on AI chatbots and generative AI: Middle-market guide to selecting AI services and local learning resources. | In production |
| Axos Bank | Arc 'Archie' CFO agent and conversational AI: Agentic finance tools for small businesses via Arc; omnichannel conversational AI on Google's contact-centre suite. | Rolling out |
| Sallie Mae | AI-simulated training for contact-centre agents: Conversation and chat simulation replaces agent-to-agent role-play. | In production |
| United Community | AI-powered contact-centre agent tools: Efficiency tools for agents while maintaining service quality. | In production |
| WaFd Bank | AI-powered 24/7 call-centre agent: Support automation to free bankers for higher-value interactions. | Rolling out |
| First Interstate | Customer guidance on AI search answers: Explains inaccurate AI-generated rate summaries. | In production |
| Customers Bank | AI call-centre scoring: 85% greater efficiency and full visibility into customer interactions. | In production |
| Simmons Bank | Centralised call centre and automated teller line: Legacy service automation credited to the retiring CDO. | In production |
| Bank of Hawaii | Digital banking scale: 350,000-plus enrolments; 6.4 million monthly logins; Branches of Tomorrow. | In production |
| Community Bank N.A. | Small-business AI and automation guidance: Client education on AI tools. | In production |
| FirstBank | Insights personal-finance tool: Automated spending categorisation, budgets and net-worth tracking in the app. | In production |
| Banner Bank | Banner Virtual Assistant: Conversational phone and chat assistant for account servicing with human handoff. | In production |
Which banks have AI chatbots?
All ten of the largest US commercial banks had deployed chatbots by 2023, according to the CFPB. Named assistants documented by the banks themselves include Bank of America's Erica (2018), Capital One's Eno (2017) and Chat Concierge (2025), Wells Fargo's Fargo (2023), the U.S. Bank Smart Assistant (2020), Truist Assist (2022), Fifth Third's Jeanie and Citizens' CiZi.
What is conversational AI in banking?
Software that lets a customer or employee bank in natural language, by text or voice: checking balances, moving money, disputing a charge or getting an explanation. It ranges from rule-based chatbots with curated responses to assistants built on large language models and agents that complete tasks on the customer's behalf.
What is the most used bank chatbot?
By published totals, Bank of America's Erica: more than 3.2 billion client interactions since its 2018 launch and nearly 700 million in 2025, per the bank's March 2026 release. Wells Fargo reported more than 1 billion Fargo interactions in under three years.
Is there a regulation for bank chatbots?
Not a dedicated federal one. The CFPB applies existing consumer financial law to the channel, third-party risk guidance covers vendor-built assistants, and state laws add disclosure: Colorado's HB 26-1263 requires conversational AI services to identify themselves as AI from January 1, 2027.
Do bank chatbots use generative AI?
Increasingly, but the largest deployed assistants grew out of intent-recognition systems with curated responses. US model risk guidance places generative and agentic AI outside its scope since April 2026, so a generative assistant is governed through the bank's broader AI governance, third-party risk management and consumer-compliance controls.
Every new bank assistant and every chatbot rule, the morning after.
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning