Prem Natarajan joined Capital One in May 2023 as Chief Scientist and Head of Enterprise AI; a November 2026 Capital One page calls him chief scientist and chief enterprise AI officer. He came from leading Amazon's Alexa AI organization. He argues for building AI in-house on open-weight models tuned with proprietary data, and says distilled models cut Chat Concierge latency fivefold.
- Joined Capital One in May 2023 from Amazon, where he led the Alexa AI organization.[1][2]
- In October 2026 he said distilled, fine-tuned open-weight models cut Chat Concierge latency fivefold and raised dealership customer engagement 50%.[12]
- He said an AI assistant for audit, risk and compliance cut quality errors 85%; MIT Sloan Management Review (October 2026) reported it saved about 5,000 hours across 30 auditors.[11]
- Fortune reported in September 2026 that, with specialised open-weight models, customer service agents edit just one-fifth of the case-summary copy drafted by AI.[10]
- He spearheaded Capital One's USC responsible-AI center (April 2024) and was named a primary contributor to the ASKS generative-AI safety center Capital One launched with Illinois (June 2025).[5][8]
Bank
Runs
Technology strategy, architecture and development for enterprise data, analytics and machine learning, including AI capabilities, tools and research (2023 announcement)[1]
Chat Concierge
Latency cut fivefold, dealership engagement up 50% (October 2026)[12]
Other roles
Faculty appointment in computer science at the University of Southern California (2023)[1]
Bank's place in the AI index
#9 of 120 US banks
What has Natarajan said about AI?
Interview with Thomas H. Davenport and Randy Bean, MIT Sloan Management Review
“You can’t achieve transformation if you’re focused on quarterly ROI.”[11]
Oct 28, 2026 · MIT Sloan Management Review ↗
Evident AI Symposium, New York, reported by Banking Dive
“The engineering and architectural decisions you make lower down in the stack are either massive enablers or massive constrainers of what you can do up top.”[12]
Oct 22, 2026 · Banking Dive ↗
Fortune, on Chat Concierge and agentic AI
“We want to start off at the low end of the risk spectrum, but also find use cases with impact and enough complexity that we can learn from it.”[9]
Dec 15, 2025 · Fortune ↗
NVIDIA AI Podcast, reported on NVIDIA's blog
“Your data advantage is your AI advantage.”[7]
Apr 23, 2025 · NVIDIA Blog ↗
Interview with Peter High, Forbes, on generative AI for developers and associates
“While the technology is transformative, we emphasize the importance of a human-in-the-loop approach to mitigate risks.”[6]
Dec 16, 2024 · Forbes (Peter High, contributor) ↗
Remarks to the U.S. Senate AI Insight Forum, published by Capital One
“As with any emerging technology, it is essential to balance excitement about potential benefits of AI with a clear-eyed view of associated consequences and risks.”[3]
Nov 1, 2023 · Capital One ↗
What has Natarajan done on AI, and when?
- Oct 28, 2026
Audit assistant cut quality errors 85%[11]
He told MIT Sloan Management Review that an assistant for audit, risk and compliance built on open-weight models cut quality errors 85%; the authors report it saved an estimated 5,000 hours across 30 auditors in a few months.
- Oct 22, 2026
Distilled models speed up Chat Concierge[12]
At the Evident AI Symposium he said open-weight models fine-tuned on proprietary data were distilled into smaller models, cutting latency fivefold and lifting dealership engagement 50%.
- Sep 29, 2026
Open-weight, specialised models to cut token costs[10]
Fortune reported his emphasis on agentic AI infrastructure to reduce latency and token costs; customer service agents now edit just one-fifth of the case-summary copy drafted by AI.
- Sep 24, 2026
ASKS symposium: safety for autonomous agents[13]
In the opening fireside chat at Illinois he described the bank's "destination-back" approach and said autonomous agency in regulated finance requires rigorous safety mechanisms.
- Dec 15, 2025
Chat Concierge: own multi-agent stack[9]
Fortune reported him saying the dealer tool is 55% more successful at converting leads and latency fell fivefold because the bank builds its own stack.
- Jun 4, 2025
Capital One launches ASKS center at Illinois[8]
Capital One and the University of Illinois launched the Center for Generative AI Safety, Knowledge Systems and Cybersecurity; the bank named him first among its primary contributors and quoted him on the launch.
- Apr 11, 2024
Spearheaded USC responsible-AI center[5]
He led development of CREDIF, Capital One's responsible-AI center with the University of Southern California.
- Feb 27, 2024
Capital One joins Partnership on AI[4]
In the bank's announcement he cited multi-stakeholder frameworks to advance responsible AI.
- Nov 1, 2023
Spoke at Senate AI Insight Forum[3]
He proposed four pillars: workforce training, research partnerships, risk management frameworks, and inclusion of underrepresented groups.
Where does Natarajan stand on the questions AI raises for banks?
| Topic | Position | As of | Source |
|---|---|---|---|
| Spending and returns | Measuring AI on short-term returns is premature for a fast-changing technology; he says the ultimate return is a happy customer. | Oct 28, 2026 | MIT Sloan Management Review[11] |
| Build or buy and vendors | Build: open-weight foundation models customised with proprietary data into specialised models and multi-agent workflows on the bank's own stack. | Oct 22, 2026 | NVIDIA Blog · Fortune · Banking Dive[7][9][12] |
| AI agents | Start agents at the low end of the risk spectrum; autonomous agency in regulated finance needs rigorous safety mechanisms. | Sep 24, 2026 | Fortune · Capital One[9][13] |
| Risk and regulation | Told the Senate that Capital One manages AI risk through model risk management, compliance and governance; in 2024 he stressed a human-in-the-loop approach for generative AI. | Dec 16, 2024 | Capital One · Forbes (Peter High, contributor)[3][6] |
| Jobs and workforce | AI should shift cognitive burden from people to systems; new tools come with heavy training, because technology revolutions need tradecraft to evolve alongside them. | Oct 28, 2026 | Capital One · MIT Sloan Management Review[1][11] |
What does Capital One's AI record look like?
Capital One's profile on this site records 13 AI use cases, 13 of them in production.
- How does Capital One use AI in auto finance?
- What is Capital One's approach to responsible AI?
- What AI research, patents and partnerships does Capital One have?
- What is Capital One's virtual assistant and how does it compare with other banks'?
- What is Capital One's technology stack?
- What proprietary AI solutions has Capital One built?
- What is Capital One's AI strategy and who sets it?
- How is Capital One building its AI workforce?
- What innovative products and services has Capital One launched in 2025 and 2026?
- What digital banking innovations and customer experience technology does Capital One use?
- What is Capital One Software and what does it sell?
- Who leads AI and technology at Capital One?
- May 2023–present
Capital One · Chief Scientist and Head of Enterprise AI (described in 2026 as chief enterprise AI officer)[1][13]
- About five years to 2023
Amazon · Vice president leading the Alexa AI organization[1][2]
- Earlier
University of Southern California · Senior Vice Dean of Engineering, Viterbi School; Executive Director, Information Sciences Institute[1]
- Earlier
BBN (now Raytheon BBN Technologies) · Executive Vice President and Principal Scientist for Speech, Language, and Multimedia[1]
Who is Prem Natarajan?
Prem Natarajan is Capital One's chief scientist, appointed in May 2023 as Chief Scientist and Head of Enterprise AI. He previously led Amazon's Alexa AI organization and was a senior vice dean of engineering at USC.
What has Prem Natarajan said about AI and jobs?
He says AI should shift cognitive burden from people to systems. In 2026 he said every new tool comes with heavy training, because technology revolutions need tradecraft to evolve alongside them.
What is Prem Natarajan's role in Capital One's AI strategy?
He leads strategy, architecture and development for enterprise data, analytics and AI. He champions building on open-weight models tuned with proprietary data, the approach behind Chat Concierge.
How does Natarajan judge returns on AI?
He told MIT Sloan Management Review in 2026 that transformation cannot be achieved with a focus on quarterly ROI, and that the ultimate return is a happy customer.
Head of AI, Wells Fargo
Chief AI Officer, Synchrony
Chief Analytics Officer, JPMorgan
Group Head of AI, Citi
Chief AI Officer, U.S. Bank
Head of Wealth Management AI, Morgan Stanley
Chief Data & AI Officer, BNY
Chief AI Officer, Raymond James
Founder, Chairman, and Chief Executive Officer
- Introducing Prem Natarajan
- Capital One's new chief scientist says 'responsible, thoughtful' generative AI is key
- AI Leader Speaks at US Senate AI Insight Forum
- Partnering with AI alongside industry leaders
- Capital One and USC partner to drive responsible AI (CREDIF)
- Prem Natarajan On Leading AI Innovation At Capital One
- Capital One Banks on AI for Financial Services
- Advancing generative AI partnerships with UIUC
- 2025 was the year of agentic AI. How did we do?
- How bottom-line-focused financial firms are finding ROI in AI
- AI Is Thriving at Capital One
- Smaller AI models help Bank of America, Capital One manage costs
- Advancing safe agentic AI: ASKS symposium highlights
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 PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning
