Don Tapscott has a habit of being right too early. In 1994, when most executives were still debating whether the internet was a fad, he published The Digital Economy and mapped the transformation that would take another decade to fully arrive. Now, in You to the Power of Two, he's doing it again - this time with what he calls identic AI: personal agents that learn who you are, reflect your values, and operate as cognitive extensions of yourself.
It's a compelling vision. But read it from inside a bank, or from the strategy floor of a fintech, and the framing shifts uncomfortably. Tapscott's identic AI is a superpower for individuals. For banking, the more urgent question is what happens when your customers deploy that superpower in the direction of your products, your pricing, your relationships - and your margins.
Because that's the scenario most banking strategists are not yet taking seriously enough. Not AI inside the bank. AI on behalf of the customer.
01
The Customer Who Doesn't Show Up Anymore
Start with a simple thought experiment. It's 2028. A customer - let's call her Maya - needs to remortgage. She doesn't open her bank's app. She doesn't call the branch. She doesn't even Google it. She delegates the task to her financial agent. The agent pulls her income data, her credit profile, her existing mortgage terms, and her stated priorities - lowest cost over a five-year horizon, with flexibility on overpayments. It queries forty-two lenders simultaneously. It reads the small print. It stress-tests three scenarios against her income trajectory. Within ninety seconds, it returns a ranked shortlist with a recommendation.
Maya's bank of twenty-three years is fourth on the list.
This is not science fiction. The component technologies exist today. What doesn't yet exist at scale is the orchestration layer that makes it seamless - but that gap is closing fast. Gartner's work on machine customers - non-human economic actors that independently evaluate, negotiate, and transact - identified this shift as one of the most significant commercial disruptions of the decade. Their projection: by 2028, machine customers could account for a meaningful share of B2C revenue across financial services, retail, and utilities. Not through dramatic displacement, but through the quiet accumulation of delegated decisions.
Tapscott frames identic AI as personal sovereignty - the individual owning and directing their cognitive extension. That's the right frame for the individual. For the institution, it's a different word: disintermediation.
02
What Coase Tells Us About What's Coming
Tapscott invokes Ronald Coase's transaction cost theory to argue that AI will dissolve the rationale for traditional firm structures. It's a sharp observation, and it applies with particular force to retail banking.
Coase's insight, for which he won the Nobel Prize, was elegantly simple: firms exist because the cost of transacting in open markets - finding the right people, coordinating activity, building trust - is higher than the cost of internalising those functions. Bring it inside the boundary of a firm, and you reduce friction. The firm is, at its core, a transaction cost reduction machine.
Banking has historically been one of the most extreme expressions of this logic. The reason customers maintain a relationship with a single bank rather than sourcing each financial product from the optimal provider is almost entirely a function of transaction costs: the friction of switching, the opacity of alternatives, the cognitive load of comparison, the inertia built into direct debits and salary mandates. These aren't loyalty. They're switching barriers dressed as loyalty.
Now introduce an agent that eliminates those barriers - that can search, compare, switch, and transact with near-zero friction on the customer's behalf. Coase's equation inverts. The cost of transacting in the open market collapses. The rationale for the bundled relationship weakens. And the bank that relied on inertia rather than genuine value suddenly finds itself exposed.
This is the real Coasian reckoning for banking: not that firms will restructure internally (though they will), but that the customer's relationship with their bank will be mediated by an agent that has no brand sentiment, no loyalty to the familiar, and infinite patience for optimisation.
03
The Trust Paradox at the Heart of Agentic Commerce
There's a tension at the centre of all of this that Tapscott gestures at but doesn't fully resolve, and it's the tension that makes banking genuinely interesting as an Agentic Commerce battleground.
He asks who should own the identic AI. His answer - the individual, not the platform - is philosophically correct and practically difficult. Most people will not build their own agents from scratch. They will adopt agents embedded in platforms: Apple Intelligence, Google Gemini's personal assistant layer, whatever Amazon deploys through Alexa's financial services integrations. These platforms will be the de facto owners of customer intent at the moment of financial decision-making.
For banks, this creates what I've been calling the trust paradox. The bank that most aggressively helps customers deploy AI-powered financial management - genuinely acting in the customer's interest, even when that means recommending a competitor product - is the bank most likely to be selected as the trusted data source and execution partner in an agent-mediated world. But that same behaviour, taken to its logical conclusion, is also the behaviour most likely to compress the bank's own product margins.
Trust, in other words, may require a short-term commercial sacrifice in exchange for a long-term positional advantage. The banks willing to make that trade will become the preferred infrastructure layer in an agent-mediated financial ecosystem. The banks that resist it will find themselves progressively excluded from the agent's consideration set altogether.
The FCA's Consumer Duty framework, which came into full force in 2024, turns out to be an inadvertent blueprint for agent-readiness. Its core requirement - that firms must deliver good outcomes for customers, not merely compliant products - is precisely the behaviour an agent would reward. Banks that have genuinely embedded Consumer Duty into their product design are, whether they know it or not, building the trust architecture that agent-mediated commerce will select for.
04
The Intelligence Layer: From Visible to Invisible Banking
The strategic response to all of this is not to build a better app. The app is visible banking: the bank as interface, as destination, as the place the customer comes to. That model is weakening. The customer is not going to come to the bank anymore. The agent will come on their behalf.
The opportunity - and it is genuinely a significant one - lies in what I've been calling the Intelligence Layer: the bank's capacity to be embedded in the customer's financial life at the level of data, insight, and trusted execution, rather than at the level of interface.
Think of it as a shift from being the place your customer banks, to being the infrastructure their agent relies on. The bank as a source of verified financial identity. As a real-time data feed that agents can query. As a trusted execution partner for transactions that require regulatory compliance, fraud protection, and capital backing that pure fintech pipes cannot provide. As a proactive financial intelligence function that surfaces relevant insights to the customer's agent before the customer even knows they need them.
This is invisible banking: the bank working hardest when it's least visible, embedded in the automated decisions that shape the customer's financial life rather than waiting to be consulted.
It requires a fundamental rethinking of what banking products are for - not destinations that customers visit, but capabilities that agents invoke. The banks best positioned for this transition are not necessarily the largest. They're the ones with the clearest data architectures, the most trustworthy APIs, the strongest compliance records, and - crucially - the most genuine track record of acting in the customer's interest. Because agents, over time, will learn which institutions can be trusted with delegation design authority and which cannot. That trust signal will be worth more than any marketing budget.
05
Know Your Agent: The Regulatory Horizon
There's a regulatory dimension to all of this that the industry has barely begun to grapple with. Tapscott flags the issue of agent accountability when he discusses the need for "checks and balances" on delegated decision-making. In regulated financial services, that abstraction becomes a concrete compliance question very quickly.
When a customer's agent switches their mortgage, who is responsible for the suitability assessment? When an agent negotiates a credit limit increase on the customer's behalf, who verified that the customer's financial position supports it? When an agent makes a series of investment transactions that, in aggregate, represent a strategy the customer never explicitly authorised, who is liable?
The FCA has not yet issued definitive guidance on agent-mediated financial transactions, but the direction of travel is clear. The "Know Your Customer" framework that underpins AML, suitability, and affordability assessment will need to evolve into something like "Know Your Agent" - a framework for verifying the provenance, authorisation scope, and accountability chain of the agents acting on behalf of customers.
This is not a distant regulatory consideration. Banks that begin building the infrastructure for agent verification and authorisation scoping now will have a meaningful head start when the FCA inevitably moves. More importantly, they'll be the institutions that agents are authorised to interact with, because they'll be the ones that can demonstrate the compliance architecture that enterprise-grade agent deployment requires. The Agent Payments Protocol (AP2) and its use of Verifiable Digital Credentials offer a glimpse of what this infrastructure will look like.
06
What Banking Leaders Need to Do Now
Tapscott's advice for executives is to get knowledgeable, use the technology themselves, and map where augmentation will be uneven. It's sound guidance. For banking specifically, I'd add four imperatives.
Audit your transaction cost dependency. Be honest about how much of your customer retention is genuine preference versus switching friction. Because that friction is being systematically eliminated. Any product or relationship that relies on inertia rather than value is at structural risk in an agent-mediated market.
Build for agent legibility. Your products, your pricing, your data - can an agent parse them? Are your APIs agent-accessible? Is your terms and conditions documentation machine-readable in a meaningful sense? The banks that become legible to agents will be considered. The banks that remain opaque will be skipped.
Invest in the trust signal. Consumer Duty compliance is a floor, not a ceiling. The banks that go beyond compliance to genuinely demonstrate customer-first behaviour - transparent pricing, proactive switching recommendations, honest product matching - are building the trust architecture that agent-mediated commerce will select for. This is a long-term investment with a structural payoff.
Begin your KYA thinking now. Start conversations with your compliance and risk functions about what agent verification and authorisation scoping looks like. Model the liability scenarios. Begin building the infrastructure before the regulatory framework crystallises, not after.
07
The Reckoning
Tapscott is right that identic AI represents a fundamental shift in human capability. He's right that it will reshape the architecture of firms, the nature of management, and the very concept of work. And he's right that the question of sovereignty - who owns the extended intelligence - is the central issue of our time.
But for banking, the reckoning is more immediate and more specific than Tapscott's broad civilisational canvas suggests. It arrives not in some distant AI-saturated future, but in the next mortgage renewal cycle, the next current account switch, the next insurance renewal, when a customer's agent does in ninety seconds what used to take three hours of comparison shopping and a phone call.
The banks that survive that moment won't be the ones who built the most beautiful app or ran the most compelling campaign. They'll be the ones who built genuine value, genuine trust, and genuine legibility into their products and infrastructure - so that when the agent arrives on the customer's behalf, it finds something worth choosing.
The app era is not over yet. But its successor is already being trained.
Tony Wood
Tony Wood is an AI Transformation Consultant at the UK's leading retail bank and founder of AgenticCommerce.design, where he writes about the intersection of agentic AI, customer experience, and the future of financial services.