My Bank Will Have to Compete for My Agent
Once a personal agent can compare institutions continuously, the incumbent bank has to earn its place in the client's shortlist.
The renewal offer that no longer arrives alone
Imagine my fixed-rate mortgage is due to expire in six months.
My bank sends me a renewal offer. Today, that offer has a useful advantage before I have even read the rate: it is already here. The bank knows the mortgage, holds much of the paperwork and can place a large button in the app I already use. Comparing the wider market means finding a broker or comparison site, entering the same details again and working out whether a lower headline rate is still lower after the fees.
Quite often, people stay where they are. The friction does some of the selling.
Now imagine that my personal agent receives the offer instead. It knows the monthly payment I want to stay below, that I may move within three years and that I care more about flexibility than squeezing out the final few pounds. It asks my bank’s agent for the complete terms in a form it can compare, then requests equivalent offers from other providers using only the verified information I have permitted it to share.
It comes back with three credible options. One is cheaper. One has better early-repayment terms. My existing bank is not the best on either measure, but it can complete the renewal with less paperwork and the relationship manager has offered to review the rate if I keep more of my savings there.
At that point, the agent has prepared the choice and shown its working. I still decide whether any saving is worth the disruption.
The important change is that my bank’s offer no longer arrives alone.
Incumbency has a quiet value
Banks have always competed on price, service, brand and trust. They also benefit from information and inertia.
An existing bank knows when a product expires, where a salary arrives and which channels a client already uses. A competing provider usually begins with less context and more questions. Even when switching is financially sensible, the customer has to gather documents, compare terms, judge unfamiliar institutions and move standing arrangements without creating a mess.
A personal agent could reduce some of that work. It can carry the client’s side of the context, maintain verified documents, remember preferences and prepare the minimum information required for a particular enquiry. The shared receipt I described in the previous article can establish what was disclosed and what each provider returned.
This does not make switching free or automatically wise. Identity checks, advice requirements, tax, transfer times, fraud controls and product rules do not disappear because an agent can read them. But the administrative advantage of staying put becomes smaller when the client no longer has to rebuild the case from scratch.
The FCA’s July 2026 Mills Review puts low switching among the long-standing problems that AI could address. Its consumer research found that one in five UK adults are already open to AI making decisions for them within agreed boundaries. The report goes further than describing a new service channel. It examines what happens when agents can compare products, prepare switches and eventually act across providers.
Working through the GoUpscale Evolution model has made one consequence harder for me to ignore: client context would no longer be trapped inside the bank’s moment of need. The client could bring it to the market.
The bank’s agent and my agent want different things
There will be valuable agents inside banks. Visa’s AI Financial Assistant, due to enter a US bank pilot in August 2026, is an early example of that direction. It is designed to sit inside a bank’s app, use cardholder and institutional information, answer questions and support actions such as setting alerts or locking a card.
That kind of assistant can know the institution deeply. It can explain the products the bank actually offers, see servicing information and move work through controlled internal systems. It can make the existing relationship much better.
It is still the bank’s agent.
If I ask it where to place £50,000 of cash for the next year, it can find the best answer available inside that institution. A personal agent should be able to ask a wider question: which suitable options could I use, across the providers I allow you to inspect, once tax, access, protection limits, service and the value of my existing relationships are considered?
Both questions are useful, although only one tests the wider market.
The personal agent could then return to my bank with a fairly specific challenge. Another provider pays more, but requires a new account and offers weaker support. Is the bank willing to improve its rate, waive a fee or explain why keeping the money in the wider relationship is worth the difference?
That is a better starting point for a human conversation than a generic retention call. The relationship manager sees the real trade-off. The client does not have to pretend that loyalty has no price.
What the bank will have to compete on
Price will matter, but a client agent should be able to understand more than a rate table.
A bank that wants to remain in the shortlist will need to make several things legible:
- The complete economics. Rates, fees, penalties, tax treatment and conditions need to survive comparison rather than looking attractive only on the first screen.
- Eligibility and evidence. The agent needs to know which facts support an offer, how current it is and what could change before approval.
- Service. Response times, access to a person, exception handling and what happens when something goes wrong are part of the product.
- Relationship value. Credit flexibility, joined-up planning, specialist access and the cost or benefit of holding several products together need to be explainable rather than implied.
- A safe route to act. The bank must be able to identify the client and agent, inspect the mandate, enforce the limits and return evidence of what happened.
Some providers will be easier for agents to use than others. Structured, current product information will help. So will reliable interfaces, clear permissions and quick answers. A beautifully formatted feed will not rescue a poor product, but an excellent product that an agent cannot inspect may never reach the client.
At the high-net-worth end, this becomes more interesting than simple rate shopping. A client may already use several banks for custody, lending, deposits, foreign exchange and access to investments. Their personal agent could compare one part of the relationship without disclosing the whole picture, then ask each institution to explain the value of keeping the pieces together.
The private bank may still win because its relationship manager understands a family structure, can negotiate a lending exception or brings useful judgement at the right moment. Those are real advantages. They will have to be visible in the decision rather than sheltered by the effort of looking elsewhere.
The shortlist is a new place to hide an incentive
There is an appealing version of this story in which the agent searches the entire market, understands the client perfectly and returns the objectively best answer.
That is not a safe assumption.
The FCA review describes three broad ways an agent might reach financial providers: direct deals with selected firms, private gateways such as aggregators, or open and standardised access. To a client, all three could look like one neat conversational answer. Underneath, they produce very different markets.
An agent may search only providers that paid for an integration. A platform may rank a partner more highly, exclude a smaller institution whose systems are harder to connect or optimise for the commission it receives. A subscription tier might buy wider coverage. The client may never see the missing options because a conversational recommendation does not naturally show the empty spaces around it.
The shortlist therefore needs its own plain evidence:
- which providers and products were searched;
- which were unavailable or excluded, and why;
- which criteria the agent used and how the client set them;
- whether a commercial relationship affected inclusion or ranking; and
- how sensitive the result is to a different assumption about price, service, risk or convenience.
Otherwise, power moves from the bank’s app to the agent’s answer box and the client remains in roughly the same position.
Open standards can help more providers participate on comparable terms. They cannot decide what “best” means for a particular person. That remains a mandate and governance question, especially when the agent begins initiating transactions rather than only preparing them.
Better competition can still create a worse habit
An agent that can compare continuously may also switch continuously.
The Mills Review calls this possible effect “hyper-switching.” It could push banks to improve rates and reduce the reward for leaving existing customers on poor products. It could also make deposits less stable, fragment a client’s finances and encourage constant movement for gains too small to justify the operational risk.
A good mandate should not simply say “always maximise the rate.” It might say that my current bank remains preferred unless another option improves the outcome by a meaningful amount. It might value deposit protection, tax position, access to a known person, the hassle of another account and the fact that moving one product could affect credit terms somewhere else.
People do not experience their finances as a leaderboard. A client may reasonably accept a slightly lower return in exchange for fewer institutions, quicker help or a relationship that has proved useful when the situation was unusual.
The agent’s job is to make that preference explicit and test it against the alternatives. It should not replace one form of inertia with a restless compulsion to optimise everything.
The relationship can still win
If personal agents become an important financial interface, banks will lose some control over when and how their products are discovered. The customer may arrive through an agent-prepared comparison rather than the bank’s campaign, app or relationship calendar.
The banking relationship is still valuable, but it now faces a harder test.
The bank can still know its own capabilities better than an outside agent. It can respond quickly, recognise permitted context, make a strong offer and bring in a person who understands why the client may value something the comparison cannot reduce to one number. Over time, the personal agent may learn that this bank is dependable when a request falls outside the standard path.
The bank may keep its preferred position, but it will have to keep earning it.
My first three articles in this sequence were about making an agent-to-agent relationship useful, governable and repairable. If that architecture works, the next consequence is commercial. The client gains a practical way to ask whether the current relationship is still the right one.
When the next renewal offer arrives, my bank should still have every opportunity to win. It will be one offer among several, understood in the context of my life and chosen for reasons I can see.
Sources
- FCA: The Mills Review — AI and the future of retail financial services, published July 2026.
- FCA: Mills Review summary, published 6 July 2026.
- Visa: AI Financial Assistant, announced 14 July 2026 for an August 2026 US pilot.