Beyond the buzz – Real Usecases for AI in Banking

Banking should be automatic and should be intelligent. As a consumer or a business, banking should be something that ‘just happens’ when income lands in my account, with funds distributed optimally across deposit, credit or wealth management accounts. Upcoming bills, subscriptions and expenses should be automatically forecasted and scheduled. Most payments and expenses should be able reconcile themselves. This was the tone set at the Canadian Lenders Association Finance Summit this week, and I couldn’t have asked for a better segue into my panel’s session where we really got down to the good, the bad and the ugly of applied AI in banking and payments. Here’s some key takeaways:

1. Over-hyped: Superficial customer service AI chatbots.

Under-hyped: Transformationaly refactoring back-office processes like underwriting and loan servicing with AI

2. Over-hyped: LLMs can replace anything.

Under-hyped: Agentic designs with a mixture of generative AI, predictive AI and deterministic rules-based systems

3. Going the distance with AI projects from prototype to production requires whole new software and product development lifecycle

4. Perfection is the enemy of good. AI often *over* scrutinized for risk. Yes, AI will make mistakes. But your human-driven processes are also full of their own risks and human-centric errors. The bar to evaluate against is not zero risk tolerance, but how much can we materially improve performance, end-customer experience and de-risk operations vs status quo

5. Strategically, build an AI strategy that plays to the strengths of your organization. In theory, larger banks should have may advantages in leveraging AI, they have the resources of scale, they have deeper proprietary data to train from. However, larger banks may be held back by legacy tech, talent and organizational inertia. Tech partnerships may help them move faster. Smaller or newer banks/fintechs have the potential to be more nimble, more tech savvy but may lack resources and depth of data. Data-sharing partnerships or alliances may help them accelerate.

6. Open Banking and Rich Payment Data (as comes with standardized Real Time Rails payments) will be oxygen that fuels the next wave of killer AI usecases. BUT Canadian incumbent FIs have, for decades now, slow-walked the introduction of modern digital banking standards and APIs. In large part for competitive reasons. But in so doing, they may win a few battles but losing the war.

You can’t stop the march of technology. For those institutions that can’t keep up industrial change, the real risk is existential. The ever-widening gap what customers ‘should’ expect and vs what they receive today, cannot be sustained indefinitely.

Special hat tip to Hanna Zaidi at Wealthsimple and Rob Khazzam at Float who are both killing it these days in the Canadian market. I cribbed ‘automatic and intelligent’ from Hanna.

And huge thanks to my awesome panelists: Janet Lin, Rob Dunlap and Simon (Haoyu) Sun

Posted in Uncategorized | Leave a comment

Canadian Bankers Gather to Party like it’s 2015

This Canadian payments conference had me checking my Uber for a flux capacitor. Something felt like an accidental time warp, listening to sessions at the Payments Canada Summit in Toronto last week, coming from the level of payment discourse in San Francisco.

In Toronto, the big news is Canada’s RTR (real-time payment rails) finally approaching technical readiness, as well as Canadian Banks still-pending steps towards basic Open Banking. Collectively, digital banking standards that have been table stakes in other global markets for near a decade or more.

Now, warp back 10 days ago in San Francisco where the major card networks and the likes of Stripe, PayPal, OpenAI, Claude, Perplexity were product dropping breakthrough specifications on intelligent commerce and for payments handled by agentic AIs. These would be the kinds of standards that will become table stakes over the coming decade of payments and banking innovation.

Of course, when we say ‘technical readiness’ for Canada’s RTR is scheduled this fall, that’s still long before public availability sometime in 2026, though no specific date is agreed upon yet. And I get it, there’s a lot of work, testing and certification left for all participants to do. Meanwhile however, in San Francisco, Visa dropped their announcement of AI-ready payment services, and that the APIs are now available to developers on the same day.

Oh Canada, it wasn’t always like this. 20 years ago, Canada was a world leader in banking tech. First market in North America to launch chip cards and contactless payments. Among the highest in adoption of online banking, cards and electronic payments over cash and check. Canadian banks backed a world-first realtime money transfer service with CertaPay Email Money Transfers (now Interac ETransfers), beating the US-equivalent Zelle to market by 14 years.

Now it’s not a bad thing (it’s great!) that Canada is finally moving towards realtime payments that technically support the rich data standards of ISO20022. And yes, Canada’s version adds a few welcome layers missing from bare-bones RTR/ISO standards (like centralized fraud monitoring, and common end-user experiences via Interac E-transfer overlays). But there’s nothing here that couldn’t have been implemented more than a decade ago.

Of course the culprit here are Canada’s big banks. But on the surface, understandably. Why should incumbent banks bear the high costs, risks and liabilities of opening up their data and money-movement systems, for a perceived one-sided benefit to challenger banks or fintechs stealing their hard-earned market share?

At the Payments Canada Summit, a few of the invited fintech leaders gave us glimpses of an alternate future. The realization is that are innumerable 10x opportunities out there. But Canada’s track towards read-only open banking, and narrow usecase targets of RTR won’t cut it. Not to mention the need for global and cross-border interoperability.

But no one can forestall the march of innovation forever. The larger the innovation gap becomes, the more painful the disruption is likely to be. Even as an incumbent banker, I’d rather be at the table actively building and pushing for forward-visioned infrastructure investments and standards. Rather than still fighting rearguard actions against the last wave, while the next wave is already at the shore. The coming AI wave will be as big a channel-switch in how our industry serves customers as mobile, the internet or the ATM was.

It’s a trap to just think of AI as just relevant to worker productivity. Instead, realize that your end customers are adopting AI too. When it comes to purchase decisions or financial advice, consumers are turning to ChatGPT first for knowledge and advice. Increasingly for any brand, the growth in traffic to your website isn’t humans, it’s coming from bots. Ready or not, Sam Altman and his ilk are inserting their apps between you and your customers. The next logical step will be the customer’s AI provider actioning those recommendations as an agent. This is called agentic AI, or agentic commerce, and it’s already happening. AI progression is inevitable and as an industry, we need to be prepared together. That’s readiness is what global fintech leaders in San Francisco are already building for right now. Bay Street needs to catch up.

Need further evidence? consider the latest resurgence of crypto. Notwithstanding all the high profile failures and burst bubbles, huge levels of renewed investment flood into stablecoin payments and borderless cash-management solutions. Though even ‘solution’ here is the wrong word. Crypto was never the first choice of solution. More like a symptom of how encumbered and closed our banking industry has become. There is so much legitimate pent up demand for truly global and effectively programmable money. So much so, that standing up whole parallel financial systems based on crypto-tech, actually sounds like a good idea to many otherwise rational people. Ultimately, the market will find its way to reward those that can solve unmet needs.. We are at a critical fork in the road as to whether the purveyors of traditional banking ledgers are part of the future or not.

But here’s the chance for Canada to turn delay into opportunity. The opportunity is to not aim just to catch up to the world, but rather to take a tiger leap. It will take joint coordination, but the relative concentration of big banks is actually an asset here, and Canada has a past history of effective industry collaboration. The multinational diversity of Canada is an incredible asset in the global market. This should be Canada’s moment to take a leap for global leadership, esp while Canada’s neighbor to the south persists in self-destructive policies of both offending and trade-sanctioning itself from the rest of the world.

On both sides of the border, there’s a sudden window for industry to take the lead. Section 1033, America’s open banking rule is on pause, as is temporarily Canada’s Open Banking legislation (which, for now, died on the books with the last Liberal government). Why not have working, de facto standards gaining traction in the market, before policy makers feel compelled to swing the regulation hammer? There’s still a chance here for banks, fintechs, networks and user groups to get in front with a market-driven vision for the future. A vision that could bring in future-ready open banking in payments, but also more-evenly balances investment costs and risks with business models for all concerned.

Toronto’s next big industry event is coming up on June 2nd, the CLA’s Canadian Finance Summit. A great opportunity to re-elevate the conversation.

Posted in fintech | Tagged | Leave a comment

Visa drops the mother of all AI payment demos

Visa just dropped a mother of all AI payment demos. Visa gets agentic commerce. The card network dropped their 2025 announcements this morning and the implications are significant. Surprised to see pretty much everything on my wishlist and more.
Here’s the thing. AI services are already disrupting how people discover, evaluate and make decisions for what they buy, particularly for high-consideration and high-intention purposes. And this trend is only going to accelerate as agentic capabilities get stronger. But you can’t buy something through ChatGPT and you can’t (and shouldn’t!) just hand chatGPT your raw credit card info and hope for the best.

For better or worse, AI flows are non-deterministic, they can make mistakes, and there are real security and data privacy concerns when commingling different LLM plugins (aka MCPs) from different sources into an agentic workflow. Like the common example of researching and booking a travel itinerary.

The solution we need is that you want to give the AI a more tightly-controlled payment credential, not a blank check.

  1. start with a trusted AI provider, then provision that AI provider with a tokenized card number in lieu of your real one. Visa demonstrated this step with consumer tapping a card to their phone, just like adding a card to Apple Pay wallet.
  2. Once provisioned, the card only activates through express consumers approval, enabling that token for limited time, merchant scope, payment limits etc.
  3. Then have the payment network itself is able to enforce that any transactions to that token expressly match the purchase intent expressly approved by the consumer for that session.
It’s these out-of-band controls that provide key guardrails to any AI usecase. Really, this is everything we needed to start building solutions. Apparently, Visa’s APIs are available starting today and MCP components later in June. Attached is just a min demo video, highly recommend watching the hour long webcast with all the details (I’ll add an update if that’s posted later).

Honestly, I was not expecting the card networks to be this forward thinking yet, it feels like somebody cooked here. Maybe partnerships with OpenAi, Anthropic, Perplexity etc. will do that 🙂

Posted in Uncategorized | Leave a comment