Backbase 2027 Banking Predictions Report

Nearly every bank has run an AI pilot, but far fewer have scaled one, and in 2027, that gap will start to show.

As boards ask for proof that AI pays its way, agents join the workforce with governance of their own, and customers start talking to their bank the way they talk to AI, expecting it to understand what they need and act on it, the distance between leaders and laggards will become increasingly hard to miss.

Our third annual predictions report brings together global banking leaders and experts to map what’s coming across 5 industry-wide trends and 5 banking segments, so you can see where the gaps will open and decide where your bank should move first.

Every bank has run an AI pilot. Very few have figured out how to scale one. That gap is the real story of 2026, and it won’t stay open for long.

In 2027, the pilot era will end. The banks that move AI into production will start compounding their advantage with every use case they ship, and the banks that don’t will spend the rest of the decade explaining to their boards why they fell behind. Soon, the distance between the leaders and laggards will be visible to everyone: shareholders, regulators, and above all, customers.

The predictions in this report describe what it takes to be on the right side of that line. Boards will demand proof that AI pays its way, use case by use case. As banks move from single assistants to teams of agents, the hardest question will be who has authority to act and what the bank can prove after the fact. Banks can rent AI models, but they have to own the accountability for what those agents do. That changes the operating model itself, as agents become colleagues and people stay accountable for the decisions that matter. And it changes the front door of the bank. Digital banking put the bank in the customer’s pocket. The next step lets customers simply say what they need done and have the bank’s agents take care of it, with the bank in control of every step.

None of this is a distant scenario. It’s already happening at the banks that have decided not to wait. So read this report as a to-do list for the next 12 months. Pick the journeys where AI will pay off first, build the controls before you scale, and redesign the work around the agents while you still have the choice. In 2027, the gap between the banks that act and the banks that watch will become the defining fact of our industry. AI waits for no bank. Neither will your customers.

Last year’s report predicted 2026 would be the year AI moved from pilots to productivity, and the industry got only partway there. A few banks have scaled AI with measured value, while the majority remained stuck. Pilots have multiplied, but mass rollout has yet to follow.

In 2027, the AI pilot era will finally end. Banks must build the production model that agentic AI needs: capabilities built once and reused across the bank, automation aimed at the steps where it pays off, and controls designed in from the start, so AI becomes part of how the bank runs every day.

A handful of banks are already showing what production looks like. DBS runs more than 2,000 models across over 430 use cases, worth about SGD 1 billion in economic value, and Commonwealth Bank measured about A$200 million in AI benefits in fiscal 2026, a number it expects to double in 2027.

Banks have never spent more on AI, but the returns are still hard to find. Of the AI use cases announced by the world’s 50 largest banks, only 30% came with any impact metric at all. And the cost of running AI is rising faster than many expected.

In 2027, banking executives and boards will demand a clear return on AI spend, use case by use case. ROI must be defined before launch and tracked for as long as the AI runs, and that evidence will shape how banks choose and buy AI, from which model to use to which partner to hire.

One of the fastest ways to improve the return is to stop using the most expensive model for every job. JPMorganChase researchers found that a small in-house model matched larger models’ accuracy while processing transactions in about 95 milliseconds, which they said could save more than $13 million a year.

The first wave of AI gave bank employees a personal assistant, but the next is giving the bank a whole new workforce. Many banks are still governing agents with tools built for humans, and policy documents simply can’t keep pace with software that acts in milliseconds.

In 2027, as banks move from single copilots to teams of agents, regulators will sharpen their expectations. Banks will need to move governance from policy documents to real-time controls, with an identity for every agent, so they can show who authorized what and switch off any agent on demand.

The first question any auditor will ask is simple: which agent did this? Today, many banks can’t answer it. Their agents typically run on a customer’s login or a shared service account, so the bank can’t tell a read-only agent from one that moves money, or switch off a single agent without stopping them all.

Much of a bank’s work still happens in the gaps between its systems. In this “messy middle,” people copy data from one screen to another, chase follow-ups by email, and pass cases from team to team. McKinsey estimates that operations make up 60-70% of a bank’s cost base.

In 2027, agents will become teammates that scale expertise across the bank, and banks must redesign roles, structures, and processes around them. Agents will prepare the case, gather the evidence, and route the work to the right place, so people can make the call.

The clearest results so far come from the back office. First Abu Dhabi Bank reports a productivity uplift of more than 20% and a 70-80% cut in manual effort across key workflows, and in Thailand, the tech arm of KBank saved an estimated 30,000 workdays by giving employees a dedicated space to build their own AI tools.

Outside of banking, people have gotten used to saying what they want and letting the technology work out how. Most banking, meanwhile, still works like a map the customer has to read: menus, forms, and product pages leave customers to find the right journey and do the work themselves.

The shift is already under way. In the U.S., ChatGPT now lets users connect their bank accounts through Plaid and ask questions about their own spending, savings, and investments. For now, it can only read the data, but the step from reading accounts to acting on them is a short one.

In 2027, banking will move from web and app-based journeys to natural, personalized agentic experiences. The bank knows the customer, coaches them, and acts for them. The customer brings the goal, and the bank finds the best route there.

Retail banking has spent the last decade getting faster, cheaper, and always-on, and lost much of what made customers feel connected along the way. Only 32% of consumers say retail banking service quality has improved in the last five years, and their money is following.

In 2027, conversational AI will become how retail banks make their customers feel known again: learning about customers’ lives (know me), guiding them towards their goals (coach me), and acting for them with consent and proper guardrails (act for me). The biggest shift is from deflection to resolution.

The proof points are already in. DBS’s generative AI assistants now reach 10 million customers, and in H1 2026, they resolved 9 in 10 queries fully digitally, with calls and emails falling 7% and satisfaction rising 17%.

Wealthy clients want answers on the spot, on whatever channel they happen to be using, and they want to hear from their advisor when markets move or their life changes. Few firms deliver both: only 17% of high-net-worth individuals describe their advisory experience as seamless and personalized.

In 2027, leading wealth firms will move from AI pilots and point tools to journeys that run end to end, with clients, advisors, and AI agents working on the same journey. Success is measured on client outcomes: how quickly a client is onboarded, how often they hear from the firm at the right moment, and how much of their wealth they choose to bring.

Take a market move that affects a client’s portfolio. Today, the client often hears about it from the news first and repeats the story while their advisor pieces together the picture from several systems. In an end-to-end journey, the client and advisor get the signal at the same time, and the conversation continues on any channel, with its full history.

On paper, private banking has rarely looked stronger, but record assets are funding a cost base that grows faster than revenue. Much of a relationship manager’s day still goes to preparation, documentation, and compliance checks, so the RM’s time is the most obvious place for AI to prove its worth.

In 2027, the gap will widen between private banks that can prove what AI delivers and those that can’t. Only a handful of leaders, likely fewer than three of the top 10, will publish controlled evidence of AI’s impact on RM productivity or revenue, and they will earn the confidence and the budget while their peers are still running pilots.

Clean evidence is scarce, which is exactly what makes it valuable. Across the world’s 50 largest banks, only 27% of new AI use cases carried a disclosed outcome in Q2 2026, and just 1 in 10 of those reported a revenue uplift.

At most small businesses, the finance department is the owner, working after hours on invoicing, payroll, tax, and forecasting, often across dozens of tools they have to stitch together themselves. Every hour spent chasing an invoice is an hour taken away from the business itself.

The first place owners feel the strain is cash flow. In the U.S., the most common reason small firms sought financing was to meet operating expenses (56%), and 22% of applicants received none of the financing they sought. These shortfalls are often predictable, yet they still arrive as surprises.

In 2027, leading banks become the operating system small businesses run on. Through APIs, the bank links to the tools a business already uses, and on top of that connected view, AI agents act as the business’s CFO: forecasting cash flow, raising and chasing invoices, preparing tax filings, and running payroll.

Commercial clients now judge their bank by the standard of the apps they use every day, and many banks are falling short. Inside the bank, relationship managers typically spend only 25-30% of their time in client dialogue, and much of that dialogue is purely tactical.

In 2027, commercial banks will run the client relationship on real-time intelligence, on two fronts. For bankers, conversation becomes how RMs get into the bank’s data, so the RM becomes the client’s strategic advisor. For clients, agents take on onboarding, treasury, and reconciliation, working within policies the client sets.

On the banker side, the early returns are already measurable. Agentic AI can reportedly return 10-12 hours a week to each banker, roughly a quarter of a standard working week, and banks applying it have reported 20-40% lower cost to serve.