What is Capital One's virtual assistant and how does it compare with other banks'?
Eno is Capital One's customer assistant, unveiled at SXSW on March 10, 2017 as the first SMS text-based intelligent assistant from a US bank, with natural-language processing the bank built in-house in under three months and trained on customer-service chat logs. It now answers account questions around the clock in the mobile app, online, by text (227-898) and email, sends notifications including to smart watches, flags surprise charges, free-trial endings and recurring charges, locks and replaces cards, and creates virtual card numbers for online shopping; it runs on AWS serverless streaming (EC2, Lambda, DynamoDB) and is strongest for credit-card customers. Unlike its peers, Capital One has never published an interaction count for Eno, so scale cannot be compared: Bank of America's Erica passed 3 billion interactions and nearly 50 million users by August 2025, Wells Fargo's Fargo passed 1 billion in March 2026, and NatWest's Cora handled 12.9 million conversations in 2025. Capital One's generative and agentic effort went first to dealers (Chat Concierge, 2025) and to employee tools, not to Eno; the bank has not announced a generative Eno.[1][2][3][4][5][6][7][8][9]
Data as of Oct 15, 2026 · Capital One profile updated Oct 1, 2026 · 20 sources cited on this page · 140 banks profiled on this site
What can Eno do, and where does it run?
From Capital One's product pages and engineering posts. Capital One notes Eno 'currently has the most robust capabilities and features for credit card customers' and is being 'rolled out to additional customers over time'.
| Capability | Detail | Channel | Source date |
|---|---|---|---|
| Account questions | Balances, available credit, recent transactions, due dates and minimum payments, authorised users; 24/7; sign-in usually required for account-specific answers | App, web, text | August 2024[4] |
| Card actions | Activate, lock or replace a card from the chat | App, web | August 2024[4] |
| Proactive alerts | Charges that might surprise the customer, free-trial endings, recurring charges, duplicate or unusual transactions; fraud confirmation by text | Push, text, email, smart watch | Retrieved September 2026[3] |
| Virtual card numbers | Creates merchant-specific virtual numbers when shopping online from a desktop | Browser extension | Retrieved September 2026[3] |
| Spending insights | Summaries and insights delivered by email | Retrieved September 2026[3] | |
| Natural-language understanding | Built in-house, concept to launch in under three months, trained on customer-service chat logs; understood more than 2,200 terms and emojis for a balance request at launch | All | September 2018[2] |
| Infrastructure | Serverless streaming architecture on Amazon EC2, AWS Lambda and DynamoDB that analyses account activity in real time | All | AWS case study[5] |
| Design | A gender-neutral character designed by the bank's AI design team | All | September 2017[10][1] |
How does Eno compare with Erica, Fargo and Cora?
A feature table, not a ranking. Each cell is what the bank itself has published; 'not published' means we found no figure. Citi is omitted because its US consumer app has no publicly documented named assistant with published figures; its 2018 Citi Bot ran on Facebook Messenger in Singapore.
| Feature | Eno (Capital One) | Erica (Bank of America) | Fargo (Wells Fargo) | Cora (NatWest) |
|---|---|---|---|---|
| Launched | March 2017 (SMS pilot) | 2018 | 2023 full rollout, with Spanish | Pre-2023; Cora+ generative upgrade June 2024[1][6][7][11] |
| Published scale | Not published | 3 billion-plus interactions, nearly 50 million users (August 2025); 2 million interactions a day (January 2026) | 1 billion-plus interactions in under three years (March 2026); 3 million-plus Spanish-speaking users | 12.9 million conversations in 2025; 7.1 million in H1 2026, 3.8 million fully digital[3][6][12][7][8] |
| Channels | App, web, SMS, email, push, smart watch, browser extension | App (plus EricaAssist for employees) | App, voice and text | App and web chat; voice-to-voice trial from September 2026[3][6][7][13] |
| Generative AI in the customer assistant | Not announced | Generative AI added to EricaAssist (employee-facing) July 2026; CEO calls Erica 'in effect a small language model' | Built on Google conversational AI; no customer data reaches the LLM | Cora+ on IBM watsonx with retrieval-augmented generation (2024); OpenAI models in the spending-insights pilot; proprietary small language models for voice (2026)[3][14][7][11][13] |
| Agentic actions | Lock, replace, activate cards; virtual card numbers. Capital One's agents act elsewhere: Chat Concierge for car buyers | Summarises client calls and recommends next steps for employees (EricaAssist) | Not published | Fraud Triage Agent routes suspicious-transaction cases (nearly 19,000 conversations by September 2026); agentic spending insights pilot with 25,000 customers[4][9][14][13] |
| Who built the language layer | In-house NLP (2018); open-source LLMs customised in-house for newer products | In-house | Google Cloud | IBM watsonx for Cora+; OpenAI collaboration; in-house small language models[2][15][6][7][11][13] |
| Reported business effect | Not published | Does 'the work of 11,000 people'; fraud losses halved (bank statements) | Not published in the milestone release | About 20-point lift in queries resolved without a human on GenAI journeys; 150% satisfaction improvement on some questions (IBM case)[3][12][7][11][8] |
How has Capital One's conversational AI evolved?
| Date | Event | Detail |
|---|---|---|
| March 10, 2017 | Eno unveiled at SXSW | SMS assistant answering natural-language texts and emojis; called the first SMS text-based intelligent assistant from a US bank[1] |
| September 26, 2018 | Why the NLP was built in-house | Concept to launch in under three months; trained on customer-service chat logs[2] |
| February 21, 2019 | Explainability for conversational AI | Margaret Mayer, who set Eno's technology strategy, on embedding explainability through the lifecycle[16] |
| 2020 | Eno cited as a cloud-native example | Built on EC2, Lambda and DynamoDB with serverless streaming after the data-center exit[5] |
| August 13, 2024 | Chat scope documented | Account management, card activity and bill-payment questions; text to 227-898[4] |
| March 4-5, 2025 | Chat Concierge | First proprietary multi-agent assistant, for car dealers rather than Eno's customers[9] |
| August 13, 2026 | MACAW described | Multi-agent workflow summarising fraud calls for several hundred specialist agents[17] |
| September 29, 2026 | Service agents edit one-fifth of AI drafts | Natarajan: customer-service agents now edit only a fifth of AI-drafted copy; smaller open-weight models cut latency and token cost[18] |
What does the Capital One profile record on this topic?
The rows below are the entries of the Capital One AI strategy profile that bear on this page, as of Oct 1, 2026; the profile holds the full record of 31 dated moves, 13 use cases and 46 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| Chat Concierge | Multi-agent assistant on dealer websites; compares vehicles, schedules appointments, routes to sales. | In production | [9] |
| Eno | Customer assistant in the mobile app, by text and email, in notifications (including on smart watches) and online from a computer: answers account questions and watches for charges that might surprise customers, free-trial endings and recurring charges. Natural-language processing built in-house. | In production | [3][2] |
| Digital and customer-service AI | Part of the enterprise AI program presented at re:Invent. | In production | [19] |
| MACAW fraud-call summaries | Understanding, reasoning, validation and explanation agents on fine-tuned open-weight models summarise long fraud calls for several hundred specialist agents; the same architecture runs Chat Concierge. | In production | [17] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 29, 2026 | Natarajan: cutting latency and token costs with smaller, open-weight models | Chief enterprise AI officer Prem Natarajan says the bank has put more emphasis on agentic AI infrastructure to reduce latency and lower token costs, building customised open-weight and specialised models with less context targeted at the desired output; customer-service agents now edit only one-fifth of AI-drafted copy. | [18] |
| Mar 4, 2025 | Chat Concierge launches for auto dealers | A multi-agent assistant that compares vehicles, books appointments and hands off to sales, on the Auto Navigator platform. | [9][20] |
| Sep 26, 2018 | Why Eno's language technology was built in-house | The bank says it built Eno's natural-language processing itself, from concept to launch in under three months, training it on customer-service chat logs; Eno understood more than 2,200 terms and emojis for a balance request. | [2] |
| Mar 10, 2017 | Eno, a text-message chatbot, launches as a pilot | Unveiled at SXSW as an SMS assistant that answers natural-language texts and emojis about balances and card payments; Capital One called it the first SMS text-based intelligent assistant from a US bank. | [1][10] |
What is Eno?
Eno is Capital One's virtual assistant, launched as an SMS chatbot in March 2017 and now available in the Capital One Mobile app, on the website, by text to 227-898, by email and through push and smart-watch notifications. It answers account questions (balances, transactions, due dates, authorised users), lets customers activate, lock or replace cards, warns about surprise charges, free-trial endings and recurring payments, confirms suspicious transactions by text and creates virtual card numbers in a browser extension. Its natural-language processing was built in-house.
How many people use Eno?
Capital One has not published an interaction or user count for Eno, unlike Bank of America (Erica: 3 billion-plus interactions, nearly 50 million users by August 2025), Wells Fargo (Fargo: 1 billion-plus interactions by March 2026) or NatWest (Cora: 12.9 million conversations in 2025). The bank says Eno is available to 'most Capital One customers', with the fullest features for credit-card customers, and serves more than 100 million customers in total.
Does Eno use generative AI?
Capital One has not said so. Its public generative and agentic work has gone to Chat Concierge for car dealers (Llama-based multi-agent assistant, March 2025), to the MACAW fraud-call summarisation workflow and to employee-facing drafting, where customer-service agents now edit only a fifth of AI-drafted copy. Eno's language layer is the in-house NLP described in 2018. No generative upgrade to Eno has been announced as of October 2026.
How does Eno compare with Bank of America's Erica?
Erica is larger by every published measure: 3 billion-plus interactions, nearly 50 million users, 2 million interactions a day and an employee version (EricaAssist) that gained generative AI in July 2026. Eno's advantages are channel breadth (SMS, email, smart watch and a browser extension that generates virtual card numbers) and proactive alerts. Capital One publishes no usage figures for Eno, so the two cannot be compared on scale or cost.
How does Eno compare with Wells Fargo's Fargo and NatWest's Cora?
Fargo (Google Cloud conversational AI, Spanish-language support, 1 billion-plus interactions by March 2026) and Cora (IBM watsonx with retrieval-augmented generation since 2024, a Fraud Triage Agent and an agentic spending-insights pilot in 2026) both run generative AI inside the customer assistant and publish volumes. Eno predates both, keeps its language layer in-house and publishes neither volumes nor a generative roadmap; Capital One's agentic products sit outside Eno.
What is Capital One Chat Concierge and is it part of Eno?
No. Chat Concierge is a separate, business-facing product launched in March 2025 for auto dealers: a multi-agent assistant on dealer websites that compares vehicles, schedules test drives and routes buyers to sales, built on Meta's Llama customised with Capital One data. It is the bank's first proprietary agentic assistant; Eno remains the consumer account assistant.
Who built and runs Eno?
Capital One built Eno's natural-language processing in-house; its design team created the character. Margaret Mayer, then VP of Software Engineering for Conversational AI Platforms, set Eno's technology strategy in 2019. Conversational AI now sits under Prem Natarajan, EVP, Chief Scientist and Head of Enterprise AI, with Milind Naphade's AI Foundations team building the newer agent systems.
- Capital One launches Eno, gender-neutral virtual assistant
- How and Why Capital One Built Eno's NLP In-House
- Eno, your Capital One assistant
- What can you use Capital One chat for?
- Capital One Completes Migration from Data Centers to AWS, Becomes First US Bank to Announce Going All In on the Cloud (case study; page retrieved 2026-10-15)
- A Decade of AI Innovation: BofA's Virtual Assistant Erica Surpasses 3 Billion Client Interactions
- Wells Fargo Reaches Major Digital Milestones (Fargo passes 1 billion interactions)
- NatWest Group Interim Results 2026
- AI for Car Dealerships: Smarter Car Shopping
- Bot AI: How Capital One's Design Team Created SMS Bot Eno
- AI-led answers, empathy-led service (NatWest Cora+ on IBM watsonx)
- How Bank of America's Erica does the work of 11,000 people
- NatWest Group first to unveil next generation of AI tools to help customers manage money and receive fraud support
- Bank of America Enhances EricaAssist with Generative AI
- GAI Training Data Transparency Disclosure
- An integral, pragmatic approach to explainable AI (Margaret Mayer)
- Why Capital One built its multi-agent AI platform around open-weight models (presented by Capital One)
- How bottom-line-focused financial firms are finding ROI in AI
- How AI is transforming financial services & banking
- Capital One's 'Chat Concierge' Puts Agentic AI on Dealer Websites
When Capital One moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning