The AI Question Banks Keep Dodging: Will the Brand Survive?
Most coverage of AI in financial services is stuck on the wrong question. It asks whether AI will make banks faster, cheaper, smarter — as if the answer were ever in doubt. It is not. Underwriting will get cheaper. Fraud detection will get sharper. Service costs will fall. None of this is interesting, because none of it is contested.
The question that should keep bank and NBFC boards awake is different and far less comfortable: in a market where AI handles the decision and embedded finance handles the delivery, does the institution's brand still matter to the customer at all?
The bank is becoming a back end
For most of banking history, the institution was a destination. Customers went to a branch, then a website, then an app — but they went *to the bank*. That relationship was the asset. It justified the deposit franchise, the cross-sell, the pricing power.
That destination is dissolving. In India, the plumbing for its disappearance is already laid. UPI turned payments into a layer that sits underneath everything and is attributed to no one — most users could not name the bank moving their money. The Account Aggregator framework lets a borrower's financial history travel, with consent, to whichever lender wants it. Co-lending and embedded credit are pushing loan origination out of bank channels and into the checkout pages of retailers, travel sites and B2B marketplaces.
AI is the accelerant. When an intelligent agent can read a customer's cash flows, compare every credit and savings product in the market in real time, and execute the best option automatically, the customer stops choosing a bank. The agent chooses. And the agent does not care about your logo, your relationship manager or your forty years of trust. It cares about price, speed and data.
Two futures, and you have to pick
This splits the industry into two strategies, and most institutions are sleepwalking into the weaker one.
The first is to remain a customer-facing brand — to own the interface, the agent, the relationship. This is expensive, requires genuine product and technology depth, and pits you directly against fintechs and platforms that are better at experience than you are. Few incumbents will win here, but the winners take the economics.
The second is to become infrastructure — the balance sheet and the regulatory licence behind someone else's experience. This is a viable, profitable business. It is also a commodity business. Infrastructure providers compete on cost of capital and API uptime, not loyalty. Margins compress, and the customer relationship — the thing that was once the whole point — belongs to the platform on top.
There is no comfortable third option where you keep the brand *and* hand off the work. Yet that is precisely the position most banks and NBFCs are drifting toward: ceding the interface to embedded partners while still believing the customer thinks of them as their bank.
Trust is the only defensible asset left
Here is the part the efficiency narrative misses entirely. As the interface disappears and AI commoditises the decision, the one thing that cannot be abstracted away is trust — specifically, the question of who is accountable when an autonomous system gets it wrong.
When an AI agent misprices risk, approves a fraudulent identity or gives advice that blows up a customer's finances, the regulator does not summon the algorithm. It summons the licensed institution. That liability is not a burden to minimise. It is the moat. The institutions that thrive will be the ones that turn accountability into a product — the trusted entity customers and platforms *deliberately* route through because someone has to stand behind the outcome.
The next decade will not be decided by which bank deploys AI fastest. They all will. It will be decided by which institutions understand, early, whether they are building a brand or building a back end — and price, staff and invest accordingly.
The ones that refuse to choose will discover the market chose for them.
