How NatWest's AI Banking Results Apply to Smaller Financial Services Firms
NatWest just posted GBP 7.7 billion in pre-tax profit for 2025, a 24% jump from the year before. They spent GBP 1.2 billion on technology. Their AI chatbot Cora handled 11.2 million customer conversations last year, with nearly half resolved without any human involvement.
Those are impressive numbers for a bank with 60,000 employees and 19 million customers. But if you're running a five-person financial advisory practice, a community credit union, or a local insurance agency, the question isn't whether NatWest's AI works. The question is whether any of it translates to your scale.
Some of it does. More than you'd think.
What NatWest Actually Did (The Specifics)
Before pulling out the lessons, the actual results are worth understanding in detail.
Customer service automation. NatWest's Cora digital assistant went from handling 5 million conversations in 2019 to 11.2 million in 2024. After upgrading to generative AI (Cora+, launched June 2024 with IBM), customer satisfaction jumped 150%, and the number of cases needing a human agent was cut in half. The number of customer journeys Cora could handle expanded from 4 to 21. In early 2026, NatWest is rolling out an agentic version of Cora to 25,000 customers, built on OpenAI models, that can answer natural-language questions about transactions and spending.
Fraud detection. Using Featurespace's ARIC Risk Hub, NatWest improved scam detection by 135% and reduced false positives by 75%. Results appeared within 24 hours of deployment. New account fraud dropped 90% since 2019. Trials of agentic AI in their financial crime units led to a tenfold increase in productivity.
Staff time savings. Automated call summaries and complaint drafting saved over 70,000 hours of staff time in retail banking. KYC document automation (through Evolution AI) saved 100,000 to 200,000 manual hours per year. Private banking AI tools freed up 30% more time for actual customer conversations.
Internal operations. AI now produces 35% of NatWest's code. All 60,000 employees have access to AI tools, and more than half completed training beyond the basics.
The Three Lessons That Scale Down
Not everything NatWest does is relevant to a small firm. Their GBP 1.2 billion tech budget isn't replicable. Their OpenAI partnership isn't available to you. But three patterns from their approach apply at any size.
Lesson 1: Automate the Documentation, Not the Decisions
NatWest's biggest time savings came from automating paperwork, not judgment calls. Call summaries, KYC document processing, complaint drafting. The humans still make the decisions. The AI handles the writing.
This is directly applicable to small financial services firms. An independent financial advisor spends, according to Kitces research, only about 20% of their time actually meeting with clients. The rest goes to admin, compliance documentation, and data entry.
AI meeting note tools (Zocks, Jump, Otter.ai) can capture client conversations and generate summaries for $20-50 per user per month. That's a fraction of what NatWest paid, solving the same problem. One advisor using Vise AI for portfolio automation scaled from 50 to 150 clients because the admin work was no longer the bottleneck.
If you're a financial services firm drowning in bookkeeping and reconciliation, BookBot automates transaction categorization and report generation for $199. It handles the documentation so you handle the clients.
Lesson 2: Start with Fraud and Compliance, Not Customer Service
NatWest's 135% improvement in scam detection is probably more valuable than their chatbot. For small firms, the equivalent is compliance monitoring.
75% of UK financial firms now use AI, according to a joint Bank of England and FCA survey from November 2024. But here's the catch: only 9% of bank executives felt well prepared for upcoming AI regulations. The FCA launched an AI Lab in October 2024. The SEC has already fined firms for "AI washing" (Delphia was hit with $225,000 for claiming it had AI algorithms that didn't exist). Colorado's AI Act, effective June 2026, is the first US state law specifically covering AI in financial and lending decisions.
Small firms face the same regulatory environment as the big banks but with a fraction of the compliance staff. AI compliance monitoring tools can flag policy changes, track regulatory updates, and maintain audit trails without a dedicated compliance officer.
Keeping up with regulatory changes across FCA, SEC, and state-level AI laws is a full-time job. ComplianceAI monitors policies and flags compliance gaps for $249, no dedicated compliance officer required.
Lesson 3: Measure the Hours, Not the Hype
NatWest tracks specific metrics: 70,000 hours saved, 135% fraud detection improvement, 90% reduction in new account fraud. They don't talk about AI being "transformative" or "revolutionary." They count the hours.
Small firms should do the same. Commonwealth Bank of Australia tried to replace 45 customer service roles with AI voice bots in early 2025. Within days, call volumes spiked higher than before and they had to reinstate every displaced worker. The technology wasn't ready, but nobody measured whether it actually worked before scaling it.
O'Connor Insurance (11 employees) took the opposite approach: started small, measured everything, and achieved an 8X return on investment in 30 days, saving 58+ hours per month. FORUM Credit Union used AI underwriting and processed 70% more loans. Community Choice Credit Union went from 0% to 57% loan automation in 90 days.
The pattern is consistent. Firms that measure hours saved and process improvements do well. Firms that chase AI for its own sake waste money.
What Small Financial Firms Are Actually Using Right Now
The enterprise tools NatWest deploys (custom OpenAI integrations, proprietary small language models, dedicated AI research offices) aren't accessible to most firms. But the problems they solve have affordable alternatives.
For Financial Advisors and Wealth Managers
| Problem | Tool | Cost |
|---|---|---|
| Meeting notes and client intelligence | Zocks, Jump, Otter.ai | $20-50/user/month |
| Tax return analysis | Holistiplan | Varies |
| Estate planning scenarios | FP Alpha | Varies |
| Portfolio automation | Vise AI | Varies |
| Client newsletters | AdvisorStream | Varies |
The biggest impact comes from meeting notes. If you're spending an hour after every client meeting typing up notes and action items, an AI notetaker pays for itself in the first week.
For Credit Unions and Community Banks
Zest AI has become the default for AI-powered underwriting in the credit union space: 175+ customers, 39 million loan applications assessed, $250 billion in loans granted, and a 100% customer retention rate. They recently launched the CU Lending Collective to make their tools accessible to smaller credit unions.
Mid-size banks are increasing AI-based cybersecurity spending by 31% versus 2024. Even if you're small, your cyber risk isn't. AI-driven onboarding and verification tools have enabled a 40% decrease in client verification costs at some institutions.
For Insurance Agencies
AI document processing eliminates about 70% of manual data entry time. Agencies using AI for customer engagement report a 20% boost in satisfaction and 15% increase in policy sales. AI receptionists start at $79/month.
Insurance agencies spending hours on claims documentation can cut that time in half. ClaimScribe automates claims processing and paperwork for $149.
The Regulatory Reality Check
You can't talk about AI in financial services without talking about regulation. The environment is shifting fast.
UK (FCA): No standalone AI rulebook, but the FCA launched its AI Lab in October 2024 with a testing sandbox and feedback channels. FCA Chief Executive Nikhil Rathi said in December 2025 they won't introduce AI-specific rules but will intervene for "egregious failures." Translation: they're watching but not yet restricting.
US (SEC): Already enforcing against false AI claims. Delphia and Global Predictions were fined a combined $400,000 in March 2024. In April 2025, the SEC and DOJ charged the founder of Nate Inc. with fraud for raising $42 million claiming AI-powered automation when manual workers actually did the work.
State level: Colorado's AI Act (effective June 2026) is the first US state law covering AI in financial and lending decisions. It requires risk management programs, impact assessments, and transparency notices. Banks subject to federal examination may qualify for an exemption, but independent advisors and brokers probably won't.
FINRA: Released a regulatory notice in June 2024 covering AI obligations including recordkeeping, customer data protection, and Reg BI compliance.
The bottom line: using AI is fine. Claiming your AI does things it doesn't is not. And if you're making lending or investment decisions with AI assistance, document everything.
What Not to Do
The Richmond Fed published research showing banks with higher AI intensity actually incur greater operational losses than less AI-intensive banks. The losses come from external fraud, client problems, and system failures. The effect is most pronounced at banks that lack strong risk management.
AI-related incidents rose 21% from 2024 to 2025 across the industry.
So here's what not to do:
Don't automate client-facing interactions first. Start with back-office automation (document processing, compliance monitoring, reporting). NatWest spent years on internal tools before rolling Cora out to customers. Commonwealth Bank of Australia skipped that step with voice bots and had to reverse course publicly.
Don't trust AI output without review. One documented case: an AI agent tasked with expense management fabricated credible-sounding but entirely false transaction details. In financial services, a hallucination in lending, payments, or reconciliation could trigger audit failures or regulatory breaches.
Don't claim your AI does more than it does. The SEC is actively looking for this. If your marketing says "AI-powered" and a human is doing the work, you're exposed.
What to Do This Week
If you're a small financial services firm, here's a practical starting point:
- Pick one documentation bottleneck. Meeting notes, compliance reports, KYC processing, transaction categorization. Whichever eats the most hours.
- Try an AI tool for that one thing. Most offer free trials. Measure the time saved over 30 days.
- Document your AI usage. Even before regulators require it, having a record of what AI you use, what it does, and how humans review its output protects you.
- Don't announce it to clients yet. Get the internal process right first. NatWest had AI running internally for years before it became customer-facing.
NatWest's GBP 1.2 billion tech budget and OpenAI partnership make for good headlines. But their actual wins, saving 70,000 hours on call summaries, catching 135% more fraud, freeing up 30% more time for client conversations, those are process improvements. And process improvements scale down just fine.
Automiva builds AI tools for small businesses in financial services, insurance, and professional services. If NatWest's results look appealing but their budget doesn't, that's the gap we fill.