An Inquiry into Open Banking, Personal Finance, and Agentic AI

The argument: Canada is still building the rules for how third parties can access consumer financial data. As AI-powered personal finance tools become more capable, regulators may soon have to answer a more difficult question: how much financial discretion should consumers be able to delegate to AI agents acting on their behalf?

Like many other ChatGPT users, I received a promotional email on September 19th with the subject line, ‘Are you still paying for that?’ 

The subject line was relatively self explanatory – prompting users to use ChatGPT to ‘find forgotten subscriptions in one place with ChatGPT’. The body of the email itself explained how users could do this – revealing that users could use the Plaid Integration to review their financial activity and extract information about their recurring spend. 

Separately, a thought occurred to me in the afternoon – Did I miss an announcement revealing the long awaited regulatory framework for open banking in Canada?

Open Banking Could Shift Power Away From Canada’s Largest Banks

Fintechs and lobbyists alike have been patiently awaiting the implementation of Canada’s long-awaited open banking framework – or, as the federal government now refers to it, consumer-driven banking. Canada has now legislated the foundations of the regime, but the system itself is not yet operational for consumers. Open banking refers to consumers’ ability to securely share their financial activity and banking data with third parties. While many other developed economies are further ahead in this ‘race’, the Canadian banking system is unique given the oligopolistic forces at play. As such, open banking is expected to de-centralize some power from the ‘big 5’ Canadian banks and create more healthy competition with fintech companies.

Importantly, Canadian banks are limited to working with the consumer data from the customers accounts held within that institution. Companies like Plaid can aggregate financial activity across sources to provide a much more robust, complete view of the customer’s financial habits and extract insights accordingly – e.g., Customer A’s cash flow if she collects hairdressing income via e-transfer to her EQ Bank account, she gets paid from her social work job through her RBC account direct deposit, and occasionally does UberEats. Access to customer financial data at this latitude, in combination with advanced machine learning techniques, enables companies like Plaid to do what Canadian FIs aim to do, but at a higher quality – better granularity on spend (instead of spend on ‘transit’, viewing spend for Ride Sharing versus Public transportation).

Canadians Are Already Sharing Financial Data Without Open Banking

Currently, Canada has legislated the foundations of an open banking framework, but the system itself is not yet operational. In the meantime, Canadians are still engaging in practices that are less standardized and potentially less secure to access the same types of services – meaning the demand is already there. Screen scraping is one example. The Financial Consumer Agency of Canada defines screen scraping as a process where a fintech app requires you to provide your online banking username and password, uses those credentials to automatically log into your bank account as if it were you, and then retrieves the relevant financial data. This is not ideal – under an open banking framework, the customer would instead authorize their FI to share specific banking information with a third party, and the bank would securely share that data through an API.

AI Is Arriving Before Canada’s Open Banking System Is Fully Operational

Interestingly,  in countries like the UK and Brazil, open banking infrastructure is already more mature, meaning there are clearer rules governing how third parties can access and use customer financial data. In contrast, Canada is still formalizing those rules at the same time that AI-powered personal finance tools are becoming more sophisticated. While these tools may initially be limited to reviewing customer data and extracting insights, the longer-term use case is likely to include agents that can recommend  and execute actions on the consumer’s behalf.

While ChatGPT has apparently only rolled out the Plaid integration in the US, it begs the question –what happens when AI-powered personal finance tools arrive here while the country’s open-banking infrastructure is still being built?  

The Regulatory Question Is Moving From Access to Action

In the future state, the intersection of personal finance, open banking, and agentic AI may evolve such that agents are given consent to autonomously execute activities based on guidelines and parameters defined by the user. While the human user may still be involved in the ultimate decision, we are not far from a world where agents are given authority to automatically execute relatively simple tasks – for example, discontinuing a subscription when the monthly fee increases. 

Public evidence of regulatory developments suggests that Canada is beginning to think about this evolution. The first phase of the consumer-driven banking framework focuses on ‘read access’: establishing the rules around how third parties can access and use consumer financial data. The government has separately indicated that future work will consider ‘write access’, or the ability to initiate an action from an account, including making payments or managing enrollment in a product or service.

Agentic AI Raises a New Question Around Financial Discretion

However, agentic AI adds another layer to this conversation. The question may no longer simply be who can access a consumer’s financial data, but what authority can be delegated once they have access to it. If an AI agent can review my transactions, should it also be able to cancel a subscription on my behalf? What about moving money between accounts? Switching financial products? And if an agent executes the wrong action, where does liability ultimately sit? As such, defining the appropriate level of financial discretion that can be delegated becomes increasingly important as AI, especially agentic AI, interacts with the personal finance world. 

Canada has spent years working through the question of how consumers should safely share their financial data with third parties. As AI-powered personal finance tools become increasingly capable, it may have to answer the next question much sooner than expected:  What happens when those third parties are no longer just reading our financial lives, but exercising discretion within them?

Sources:

• Financial Consumer Agency of Canada. “Open Banking.” Government of Canada. https://www.canada.ca/en/financial-consumer-agency/services/banking/open-banking.html
• Department of Finance Canada. “Budget 2025: Canada’s Framework for Consumer-Driven Banking.” Government of Canada. https://www.canada.ca/en/department-finance/programs/financial-sector-policy/open-banking-implementation/budget-2025-canadas-framework-for-consumer-driven-banking.html
• Department of Finance Canada. “Government Pre-Publishes Regulations to Prevent Fraud and Facilitate the Next Phase of Consumer-Driven Banking.” Government of Canada, June 2026. https://www.canada.ca/en/department-finance/news/2026/06/government-pre-publishes-regulations-to-prevent-fraud-and-facilitate-the-next-phase-of-consumer-driven-banking.html