Skip to main content
Back to blog
  • Fintech

Updated 14 min read

Will AI widen the gap between big banks and fintechs?

Big versus small is the wrong split. 11 industry reports and Evident’s 50-bank index point to rebuilt versus bolted-on

Sophie Walker

Enterprise UX and research

ShareLinkedInXEmail

JPMorgan expects to spend about $19.8 billion on technology this year. Bank of America says it spends $14 billion. Yet when McKinsey tracked AI launches in finance, fintechs were 40% of the firms and launched nearly 70% of the initiatives.

In our library of 65 reports, 11 from McKinsey, BCG, Visa and Accenture take on who pulls ahead. We set them next to the banks’ own investor materials and Evident’s new ranking of 50 banks.

The answer is less comfortable than “big versus small”. The gap that grows runs between firms that rebuilt their work around AI and firms that bolted AI onto the old work. One group of banks ends up squeezed in between.

Landing on the wrong side has a price. In McKinsey’s illustrative model, AI pioneers gain up to 4 extra points of return on tangible equity, while slow movers lose 2 to 4.

We found six patterns in the data, each with a move for 2027. The first is about the kind of AI fintechs ship, and why it earns more.

Sign up with email

The short version

  • Fintechs move faster. In McKinsey’s tracking they were 40% of the firms and launched nearly 70% of the AI initiatives.
  • The biggest banks are not falling behind. Every bank in Evident’s 2026 top 10 has been there before, even though the average bank’s score rose 26% in a year.
  • The squeeze is in the middle. McKinsey counts about 200 to 250 staff per million customers at leading UK neobanks, 700 to 1,100 at large incumbents and 1,200 to 1,350 at smaller ones.
  • Cost-only AI hits a ceiling. In Visa’s survey, 48% of European banks invest in AI mainly to cut cost, and banks led by customer outcomes were 40% more likely to see transformational gains.
  • Fintechs are exposed too. McKinsey says what once took five years to build can now be built in six months.

1. Fintechs ship more AI, and more of the kind that earns money

McKinsey’s Panorama team tracked 639 AI initiatives launched in finance between late 2022 and mid-2025. Fintechs made up 40% of the firms in the data set. They launched nearly 70% of the initiatives.

The mix tells you more than the count. The fastest-growing use cases are agentic and revenue-driving, things like multi-asset trading platforms and predictive decision tools. Fintechs lead there. Banks, McKinsey writes, mostly use AI to make what they already do more reliable: treasury automation, chat assistants, personalized advice. That work is useful. It is also the easiest to copy.

Even the total is small. AI launches were only about 3% of all product launches McKinsey tracked over the period, and McKinsey reads the gap as hesitation on the bank side.

The revenue numbers follow the launches. McKinsey and QED Investors put fintech revenue at about $650 billion in 2025, up about 21%. BCG and FT Partners count 22% growth, more than four times the pace of incumbents. Fintech is still only about 4% of the financial services revenue pool, so there is plenty of room left to take.

McKinsey’s 2026 banking review cuts it more sharply. Take the 1,000 largest banks and the 1,000 largest fintechs. Between 2021 and 2025, the fintechs’ share of their joint revenue rose from 10% to 17%.

For 2027, sort your AI work by purpose. If almost every use case makes internal work cheaper and none changes what the customer gets or pays for, you are building the part of the portfolio a fintech can match fastest.

2. The biggest banks are pulling away too

The giants are not standing still. In February, JPMorgan told investors it expects to spend about $19.8 billion on technology in 2026, up 10%. It said it had doubled the number of AI use cases in production, and Jamie Dimon said 150,000 people use its internal LLM platform every week. (McKinsey puts JPMorgan’s 2025 technology investment at nearly $18 billion.)

Bank of America says it spends $14 billion a year on technology, with more than $4.1 billion of that going to new initiatives, AI among them.

Evident published its 2026 AI Index this month. It ranks 50 of the world’s largest banks on 75 indicators. The average score rose 26% in a year, about three times the earlier pace. The top barely moved. Every bank in this year’s top 10 has placed there before, and six have made it every year. JPMorgan, Capital One and RBC hold the top three.

The leaders also prove more. Evident finds the top 10 account for 25% of use cases with impact claims, and their share grows at each tier up to hard ROI. Twelve banks now report a realized or projected return on AI, up from 8. And 80% of leading banks run sophisticated AI controls, against 40% of the rest.

At the top, then, the race is not bank versus fintech. A handful of scaled banks run the same race as the best fintechs, with far more money behind them.

In 2027, stop benchmarking against “banks” as a group. Compare yourself with the top 10 and with the best fintech in your segment, counting use cases in production that have a measured result.

3. The squeeze is in the middle

The weakest position in the data belongs to a bank that is neither scaled nor rebuilt.

McKinsey’s UK team compared retail staffing across banks, using company annual reports. Smaller incumbents run the heaviest. They need up to about 7 times the staff of a leading neobank for the same number of customers.

Figure 1. Retail staff per million retail customers, selected UK banks, 2024-25 (range)
GroupLowHigh
Leading neobanks200250
Large incumbents7001,100
Smaller incumbents1,2001,350

Source: McKinsey, “The decade to redefine UK banking”, September 2026.

The neobanks keep getting leaner as they grow. Revolut nearly tripled its customer count while its cost to serve fell from about £25 per customer in 2022 to about £21 in 2025.

BCG sees the same split in costs. Many traditional banks are stuck with a cost-to-income ratio above 60%, against about 35% at well-run digital banks, and BCG calls it an expanding gap.

McKinsey and QED name who feels this most. For many midsize incumbents, they write, the choice is now to invest for scale or slowly become irrelevant. Large players that cannot afford to buy fintechs face a bind: stay the course and stay exposed, or buy technology on worse economics just to hold share. Their third option is to rewire the bank itself, and they warn that this window will close. An analyst in McKinsey’s 2025 banking review made the same point from the investor side: focus matters most for midsize banks, because they cannot invest as much as their larger peers.

If you run a mid-sized bank, you will not outspend JPMorgan in 2027. Pick the two or three journeys where you can be the best in your market, rebuild those end to end, and partner for the rest. McKinsey notes that “horizontal” fintechs, software firms that modernize incumbents from the inside, already earn about 13% of fintech revenue, and they help especially the banks without the scale to build alone.

4. The real line runs between rebuilt and bolted-on

BCG’s fintech report draws the line somewhere else. It sees a growing divide between AI-native firms and firms applying AI to their existing operations. Incumbent banks and older fintechs land on the same side of it. BCG says both are structurally behind, held back by operating-model inertia rather than any lack of awareness.

The AI-native end looks very different. BCG profiles CloudWalk, a payments and banking fintech serving small merchants in Brazil and the US. Its revenue per employee has reached nearly $2.5 million, and its CEO says agents handle 99% of customer support. More broadly, BCG says small teams that use AI well now deliver at five times the speed, at a scale that rivals much larger organizations.

Visa’s survey of European banks finds the divide inside banks too. In product development and lending, AI is often bolted on rather than embedded, so progress is slower and stalls sooner. In transaction management, neobanks pull ahead, because their AI sits inside the core workflow. At incumbents it usually runs alongside legacy processes.

Accenture explains how banks got here. For two decades they put most of their technology money into the customer layer and layered new features on decades-old cores. Nearly 70% of IT spending now goes to keeping those systems running and meeting regulatory demands. Software costs have grown about 8% a year since 2017, faster than banking revenue.

The payoff from rebuilding is large. McKinsey’s UK team estimates that a gen AI copilot beside a banker lifts productivity about 1.2 times. Supervised multiagent workflows could lift it as much as 20 times, by McKinsey’s forecast. Where banks rewired a single frontline workflow end to end, they saw 20 to 40% lower cost to serve and 10 to 12 hours a week returned to each banker.

BCG’s Future of Finance report names the most common failure: automating existing processes instead of redesigning them. Its rule of thumb is that 10% of an AI transformation’s success depends on the algorithm, 20% on the technology and 70% on people and process.

In 2027, choose one workflow that matters to customers (onboarding, credit, disputes) and redesign it around agents from the first screen to the last. Decide what the human does before you decide which model to use.

5. Most banks are optimizing cost, and cost has a ceiling

Visa surveyed 325 AI decision-makers at banks and neobanks across 17 European markets. The belief is there: 86% expect AI to change the economics of retail banking by 2030, and 61% think early adopters will outperform or dominate.

The spending goes elsewhere. 30% of banks name cost reduction as their main AI driver, nearly twice as many as put customer experience first. Yet banks focused on the customer were 40% more likely to report transformational gains. Visa sorts banks into four types.

Figure 2. European banks by AI investment type, 2026 (% of banks)
TypeShare
Efficiency Seekers48
Compliance Keepers7
Competitor Chasers15
Trust Builders30

Source: Visa, “From AI Promise to AI Performance”, July 2026.

Strategy is not the gap. 75% of the banks have a defined AI strategy and 74% have a dedicated AI function, but only 43% find and test use cases across the whole organization. The rest grow AI team by team.

BCG’s retail banking report shows what execution looks like at the leaders. Its “future-built” companies have deployed 62% of their AI initiatives, against 12% at laggards. More than 60% of them rigorously track AI value, against 17% of slow movers. They reach impact in 9 to 12 months instead of 12 to 18.

Evident’s index shows how rare proof still is. Only 12% of the use cases it tracks report impact against operational KPIs. Barely 1% disclose a concrete financial return.

In 2027, give every AI use case a P&L owner and a business metric before it starts. A use case nobody can measure tends to stay a pilot.

6. Fintechs do not get a free pass

None of this makes fintechs safe. McKinsey and QED call AI double-edged for scaled fintechs. It powers their lead today and lowers the barriers that protect them from the next wave. What once took five years to build can now be built in six months.

They single out scaled players whose moat is only the technology they built over the past five to ten years, and firms that offer a thin front end on someone else’s product. Both may struggle. Raman Bhatia of Starling Bank put it in one line, quoted in the report: “Technology barriers are coming down, but business model barriers are going up.”

The fintechs that look strongest are the ones that turned scale into lean economics. Revolut serves 69 million customers with about 12,000 employees, and Nubank serves 131 million with about 10,000, according to McKinsey’s 2026 review.

If you run a fintech, plan for 2027 as if your interface lead will be copied within a year. Put the effort into trusted distribution, compliance and the parts of the product a competitor cannot rebuild over a quarter.

Where the reports disagree, and what I’m watching

  • Does size matter? McKinsey’s 2025 banking review calls precision, not heft, the great equalizer, and says even smaller banks can win. Six months later, McKinsey’s fintech report tells midsize incumbents to invest for scale or risk irrelevance. I read them as compatible: a smaller bank can win in a narrow field, not everywhere.
  • How fast are fintechs growing? BCG says more than four times as fast as incumbents. McKinsey says 3.5 times faster than the broader industry. The samples differ.
  • How big is fintech? About 4% of all financial services revenue, or 17% if you compare only the 1,000 largest banks with the 1,000 largest fintechs.
  • Do AI savings last? McKinsey’s 2025 review sees gross cuts of up to 70% in some cost categories but a net drop of 15 to 20% after higher technology costs, with competition passing most of the gain to customers. BCG sees AI-first banks running cost bases 30 to 40% lower, then giving up 15 to 20% of potential revenue as they pass savings on. Either way, a cost lead alone will not hold.
  • The pool itself may shrink. McKinsey estimates that customers’ own AI agents could cut global bank profit pools by $170 billion, or 9%, if banks do not adapt.

In 2027 I’ll be watching three things. Whether anyone breaks into Evident’s top 10. Whether mid-sized banks answer the squeeze with mergers or with partners. And whether the biggest spenders start showing AI returns in their reported results, beyond use-case counts.

Last word

Our team has talked with 10 design teams, from the best motorsport series to leading EU banks, about building or owning their own AI. Most of them worried less about which model to pick than about who owns the data and the workflow around it.

Our own small example points the same way. We built an internal tool that reads Mixpanel and Amplitude data and writes quick reports. It runs on Kimi, an open model, chosen over Gemini and ChatGPT for cost and speed. Evident makes the wider point: as frontier models become easier to swap, the advantage moves to everything built around them.

So yes, AI will widen the gap. The side a bank ends up on will depend on which workflows it rebuilds, and how soon.

Sources

Read next

Let's talk

Write Sergey on LinkedIn
Write Alexandra on LinkedIn
Write Emin on LinkedIn
Have questions? Ask AI
ChatGPT
Claude
Perplexity
Opens a new chat with context about us pre-loaded — ask anything

We’ll reply within 24 hours with case studies, a timeline, and an estimate.

Prefer email? Write to hi@humbleteam.com
All set – our team’s on it. Expect a reply soon.
Send another one
Oops! Something went wrong while submitting the form.
Have questions? Ask AI
ChatGPT
Claude
Perplexity
Opens a new chat with context about us pre-loaded — ask anything
Europe
Národní 135/14, Prague
Middle East
UAE, Dubai, Internet City Offices
Services by industryCompare usHealthtechSaaSDubaiLondonBlogFor AI SaaS
We use cookies to enhance your browsing experience,
serve personalised ads or content, and analyse our traffic.
By clicking "Accept All", you consent to our use of cookies.
Accept all
Decline
Privacy policy