You don’t choose an electricity provider the way your grandparents did. You don’t choose a phone company the way your parents did. Within a decade, you won’t choose a bank the way you do now.

That sentence will make most people uncomfortable. Banks feel permanent. They feel structural. Your chequing account isn’t a subscription you cancel. It’s where your paycheque lands. It’s the routing number on your rent payment. It’s the debit card in your wallet. Switching banks feels like moving house. Which is exactly why most people don’t do it, even when they should.

But the relationship between a person and their financial products is about to be rearranged. Not by a new bank. Not by a better app. By a layer that sits above all of them and makes the question “which bank should I use?” as irrelevant as “which cell tower is carrying my call right now?”

This is a perspective on how that happens. What people believe today, what’s actually true, what behaviours already prove the shift is underway, and what conditions have to hold for the whole thing to work.

What people believe right now

The dominant mental model for personal finance is institutional loyalty. You have a bank. Maybe you inherited it from your parents. Maybe you opened an account at the university because there was a branch near campus. The relationship persists through inertia, not satisfaction.

This belief system has three layers.

The first is that your bank knows you. People assume their primary financial institution has a meaningful understanding of their financial life. It doesn’t. Your bank sees what passes through its own pipes. It sees your chequing balance, your mortgage payment, and your credit card transactions on its card. It does not see your Wealthsimple portfolio. It does not see your Amex spend. It does not see the HISA you opened at EQ Bank because the rate was three times better. Your bank has a partial view of your money and presents it as though it were the whole picture.

The second belief is that financial products belong together under one roof. That your mortgage, your chequing, your savings, your credit card, and your investments should all live at the same institution. Banks reinforce this aggressively through bundling discounts, loyalty programs, and the sheer friction of moving anything. The implicit message: it’s easier to stay. The unspoken truth: it’s more profitable for us when you do.

The third belief is that trust lives inside institutions. People trust RBC or TD or Scotiabank not because they’ve evaluated these companies’ risk management practices, but because the buildings are large, the logos are familiar, and the implicit government backstop (deposit insurance) makes failure feel impossible. This trust is real. It measures at 78% for banks versus 32% for third-party financial apps. But it’s institutional trust, not relational trust. People trust the building. They don’t trust the advice.

These three beliefs keep the current structure intact. They also happen to be increasingly wrong.

What’s actually happening

The supply side of personal finance is already modular. People just haven’t named it that way.

A typical financially engaged Canadian household in 2026 looks like this: chequing at one of the Big Five. A high-interest savings account at EQ Bank or Wealthsimple Cash, because the rate is materially better. A credit card from a different issuer, often Amex, chosen for rewards unrelated to the primary bank. Investments at a robo-advisor or self-directed brokerage. A mortgage from whichever lender won the rate negotiation at renewal time. Maybe a TFSA or RRSP parked somewhere else entirely.

This person already uses five or six financial providers. They already comparison-shopped their way into this arrangement. They already implicitly made the decision that no single institution deserves all of their business.

But the demand side hasn’t caught up. Every one of those relationships is managed separately. Separate logins. Separate apps. Separate statements. Separate customer service numbers. When a better savings rate appears, the person has to discover it, evaluate it, open a new account, verify their identity again, transfer the money manually, and remember to update their mental map of where everything sits. The cognitive and operational burden of managing a modular financial life falls entirely on the individual.

The banks benefit from this friction. Every inconvenience is a retention mechanism. The harder it is to leave, the fewer people leave. According to Accenture’s 2025 Global Banking Consumer Study, nearly three-quarters of banking customers globally have a relationship with at least one competing bank. The study’s conclusion was blunt: functional digital parity has pushed the industry into a state where every bank’s app is adequate, and no bank’s app is indispensable. The differentiation game on features is essentially over. Everyone is good enough. Nobody is great enough to justify consolidation.

Meanwhile, all five major Canadian banks are investing heavily in AI. RBC’s NOMI has delivered over two billion personalized insights and helped clients save $3.6 billion through automated savings features. TD is committed to generating C$1 billion annually in AI value and has 2,500 people working on it. Scotiabank’s AIDox system processes 90% of commercial client emails autonomously. BMO ranks number one globally in AI talent development. CIBC’s internal platform has saved over 600,000 hours since launch.

Every one of these investments points inward. The AI analyses the bank’s own data. The insights are about the bank’s own products. The recommendations lead to the bank’s own offerings. Not one of these banks has built anything that shows you a competitor’s product, even when the competitor’s product would serve you better. The walls are getting smarter, but they’re still walls.

Canada’s Consumer-Driven Banking Act, now progressing through Parliament, will change the plumbing. Phase 1, targeted for 2026, mandates API-based read access. Any accredited entity can pull a consumer’s financial data from multiple institutions with explicit consent. Phase 2, expected mid-2027, adds write access and payment initiation. That’s the moment when “switch my chequing to EQ” becomes something a system can execute, not just something a person can wish for.

What “open banking” actually means in practice

Banks don’t hand out passwords. But the effect is close. Under the Consumer-Driven Banking Act, every Schedule I bank in Canada will be legally required to build and maintain standardized APIs that let accredited third parties access customer data. The customer gives explicit consent. The bank provides the data. It can’t block it, slow-walk it, or make it painful.

What flows through those APIs in Phase 1: deposit account balances and transaction history; credit card balances, limits, and transactions; investment account holdings and activity; and loan details, including mortgages. Basically, everything a customer can see when they log into their own banking app. A third party, with permission, sees the same thing.

Phase 2 adds the ability to act. Initiate payments. Move money. Fund new accounts. Redirect direct deposits. Execute switching between institutions. Not just read the data. Write to it.

Not everyone can plug in. Third parties must be accredited through a regime overseen by the Bank of Canada and meet security standards, liability requirements, and consumer protection thresholds. Once the framework is operational, screen scraping gets banned entirely. The fragile, insecure workaround that current aggregators rely on disappears. Replaced by something regulated, standardized, and mandatory.

The UK built this seven years ago. It now has 13.3 million active users, 31 million payments monthly, and API availability averaging 99.88%. It works. The question in Canada is not whether the infrastructure functions. It’s who builds the intelligence on top of it.

The infrastructure companies are already in place.

Flinks, 80% owned by National Bank of Canada, provides connectivity to 15,000 North American institutions. Plaid, at a $6.1 billion valuation, has signed an API-based data access agreement with RBC covering its 14 million digital clients. SnapTrade offers something even further ahead: read-and-write access to investment accounts, enabling not just data retrieval but trade execution through APIs.

The AI companies are watching all of this closely. Anthropic launched Claude for Financial Services in July 2025, the most comprehensive dedicated financial vertical from any major AI company. It includes financial modelling tools and integrations with S&P Global, Moody’s, FactSet, and Morningstar. OpenAI took a different route: major bank partnerships rather than direct products. BBVA deployed ChatGPT Enterprise to all 120,000 employees across 25 countries. Santander rolled it out to 15,000.

Perhaps the most telling move: Block, Anthropic, and OpenAI co-founded the Agentic AI Foundation with the Linux Foundation, specifically to establish open standards for AI agents operating across financial systems. That’s infrastructure-level coordination for a future where AI agents autonomously manage money across institutions. Nobody builds standards for a future they don’t expect to arrive.

The pipes exist. The regulatory permission is arriving. The question is what sits on top.

The behaviours that prove this is coming

People don’t wait for permission to solve their own problems. They use duct tape. The duct tape is the evidence.

Wealthica, a Canadian platform, aggregates accounts from over 20,000 institutions into a single dashboard. It’s read-only. It can’t move money, can’t optimize, can’t recommend. It just shows you the picture. People use it anyway. The demand for “show me everything in one place” already exists. The product that satisfies it is crude. People accept the crudeness because the need is real.

Borrowell compares financial products from 50+ institutions and makes recommendations based on your credit profile. It earns referral fees from providers, which means its economic incentives aren’t fully aligned with yours. People use it anyway. They want someone, or something, to tell them what’s better. They’ll accept a conflicted intermediary rather than do the comparison themselves.

Rate-chasing behaviour is now mainstream. EQ Bank built a $10 billion deposit base largely on one proposition: a better interest rate than the Big Five. No branches. No relationship managers. Just a number that’s higher than the number at your current bank. Canadians opened accounts there not because they wanted a new bank, but because they wanted their idle cash to stop losing value. The behaviour is the signal: people will move money to a better product when the friction is low enough.

Wealthsimple’s trajectory is the clearest proof. Three million users. $100 billion in assets under administration. A $10 billion valuation. The company spans investing, trading, saving, spending, tax filing, and crypto. It acquired a Montreal AI research platform in August 2025. Its stated ambition is to integrate more of its clients’ financial lives under one roof. It even built a voice AI assistant that handles portfolio and transfer queries.

But Wealthsimple is itself a financial provider with products to sell. Its AI tools carry an explicit legal disclaimer: informational only, not personalized financial advice. It cannot be the neutral coordination layer because it has a structural conflict. It can’t serve you the best chequing account if the best chequing account is at EQ Bank, and Wealthsimple also offers chequing accounts. This is the same structural problem that limited every previous attempt at financial aggregation. The company that sells the products cannot also be the one that neutrally evaluates them.

The most telling recent development arrived in March 2026. Perplexity AI launched Perplexity Portfolio in partnership with Plaid, allowing users to connect brokerage accounts and ask natural-language questions about their holdings. It’s read-only. It covers only investments. It doesn’t touch banking, credit, mortgages, or insurance. But it’s the first major AI company to integrate directly with financial account data via open banking infrastructure. The pattern it establishes is clean: AI interface on top, financial data infrastructure below. The person asks a question in plain language. The system answers using their actual data.

That pattern will expand because consumers are already assembling a modular financial life and bearing the coordination costs alone. The behaviour precedes the product. Every person who manually moves money between a Big Five chequing account and an EQ Bank savings account, every person who checks Borrowell before opening a new credit card, every person who logs into Wealthica to see their full picture, is performing the job that the coordination layer will eventually automate.

What has to be true for this to work

Four conditions. All of them are hard. None of them are optional.

Trust has to transfer from institutions to the layer. This is the existential challenge. FCAC research found that only 9% of Canadians had heard of open banking. After being given a definition, only 15% said they’d participate. More than half said no. The protections people demanded were specific: full protection from financial losses, the ability to revoke consent at any time, mandatory breach reporting, and standard security requirements.

Trust in AI for financial services is even worse. A 2025 YouGov survey found just 19% of people trust AI in finance, while 48% actively distrust it. If an AI financial assistant makes a single mistake, 58% of consumers say they’d abandon it permanently.

There’s also what might be called the trust-under-stress problem. In calm markets, a pretty interface with smart suggestions can build a loyal user base. But stress exposes everything. A frozen account. A fraud incident. A market crash. A disputed charge. In those moments, people don’t ask, “Which app is smartest?” They ask, “Who stands behind my money?” The answer has to be clear and immediate. Wealthsimple’s own customer service data reveals that satisfaction collapses precisely in these moments. The complaints are consistent: accounts are frozen with no explanation, holds are placed with no escalation path, and automated responses are provided when a human is needed. If the coordination layer inherits those failure modes, it inherits the distrust that comes with them.

This isn’t a product problem. It’s a trust-market fit problem. The product can be flawless and still fail because people don’t believe it’s acting in their interest. The trust gap between banks (78%) and third-party apps (32%) suggests that the most viable path might be a layer embedded within or endorsed by a trusted institution, rather than a challenger brand starting from scratch. The UK’s emerging model, in which banks like HSBC display competitors’ accounts within their own AI-powered apps, may be the template.

The Synapse Financial Technologies bankruptcy provides the essential warning. At its peak, Synapse supported 18 million users and $9 billion in assets through banking-as-a-service partnerships. When it collapsed, $265 million in customer funds were frozen, and a shortfall of up to $96 million was identified. Customer money was effectively missing. Any coordination layer occupies the same structural position Synapse did. Powerful but fragile.

The business model has to prove alignment, not claim it. This is where most visions of the future get lazy. “We’ll recommend the best product for you” sounds great until you ask who pays for the recommendation. Every financial comparison site that has ever existed makes money by directing users toward providers who pay referral fees. Ratehub. NerdWallet. Borrowell. The user’s interest and the platform’s economic interest are not the same thing.

Three models can work. A subscription model where the user pays, and there are no referral economics. A clearly disclosed marketplace where referral fees are visible, and the user can see when a recommendation has economic incentives attached. Or an advisor-like retainer where the layer charges for the quality of guidance, the way a financial planner does.

The subscription model is the cleanest. It’s also the hardest to sell. People expect financial tools to be free. But the moment the layer routes users toward whoever pays the most instead of whoever serves them best, the core asset is gone. The core asset is belief. Belief that this thing is working for you, not extracting from you. Business model is a trust model. They’re the same thing.

The AI has to earn trust through accuracy, not claim it through promises. Financial AI cannot afford to be wrong. Not sometimes wrong. Not occasionally wrong. The tolerance for error in a system that touches your savings, your mortgage, and your retirement funds is effectively zero.

Origin Financial, the closest global analog to the full vision, runs a multi-agent architecture that passes every output through 100+ fiduciary compliance checks. It uses deterministic engines for math and LLMs only for scenario interpretation. This hybrid architecture, not pure AI, is the minimum viable approach. OSFI’s Guideline E-23, effective 2027, will formalize these requirements for federally regulated entities in Canada, including explainability and validation standards for AI models.

The accuracy bar is high. It should be. People’s financial lives aren’t test environments.

Liability has to be accepted, not avoided. This is the condition nobody wants to talk about. When a coordination layer acts on behalf of a user, moving money, switching accounts, redirecting payroll, someone has to be responsible if something goes wrong. Under Canada’s consumer-driven banking framework, liability allocation between data holders, data recipients, and service providers is a core design element.

The company that accepts meaningful liability for its recommendations sends a signal that no marketing claim can match. It’s saying: we believe in this enough to stand behind it financially. That acceptance of risk is a costly signal. It’s expensive to fake and impossible to copy cheaply. In positioning terms, it’s Gravity. It’s the structural decision that proves the position rather than claiming it.

Any company that builds the coordination layer but structures itself to avoid liability when things go wrong is telling you everything you need to know about how much it trusts its own product.

Where this ends up

The person in this future doesn’t choose a bank. They don’t need to. They have one financial home. A single interface that sees everything, understands the full picture, and acts across providers on their behalf.

They say “my mortgage renews in two months, what should I do?” and get an answer grounded in their actual balance, their actual income, their actual risk tolerance, and the actual rates available from every lender in the market. Not a generic article about mortgage renewal. Not a comparison chart, they have to interpret themselves. A specific recommendation with the math shown.

They say, “Move my idle cash to whoever pays the best rate,” and it happens. No forms. No new login. No identity verification for the fourth time this year. The layer handles the complexity. The person handles the decision.

The banks still exist. They still hold deposits, issue mortgages, and underwrite risk. But they sit behind the layer, competing on product quality the way hotels compete on Booking.com or airlines compete on Google Flights. The relationship moves from the institution to the interface.

This is not a new pattern. It’s the oldest pattern in digital markets. When a better aggregation interface emerges, the underlying providers become infrastructure, whether they choose to or not. The airlines that invested in direct-booking technology retained more value than those that didn’t. The music labels that took equity in Spotify captured billions. The providers that fight the layer simply delay the inevitable while losing the opportunity to shape it.

The question for every player in Canadian financial services is not whether this coordination layer arrives. The behaviours already prove the demand. The infrastructure already exists. The regulation is on its way.

The question is whether you’ll be the interface or the infrastructure behind it. And whether the person managing their money will think of you as their financial home, or as a provider they barely remember the name of, somewhere behind the screen.