Why Wait for Month End?

Statistics Canada reports that 19.2 per cent of Canadian businesses used AI in the second quarter of 2026, up from 6.1 per cent two years earlier, and 64 per cent of Canadian accounting firms have embedded AI in their strategy. Yet AI’s productivity advantage largely disappears without complementary investment in data and skills, and for the small businesses that make up 98.2 per cent of Canada’s employer firms, that work usually falls to their accountant.

Mitchal Derksen, a professional accountant and self-described “nerd with a capital N,” is Co-Founder and CEO of Agile Bookkeeping, a B.C.-based firm that runs a weekly accounting cycle on AI it built in-house with a Canadian-first approach.

He is mentored by Jim Balsillie, former Co-CEO of Research In Motion (BlackBerry) and Chair of the Council of Canadian Innovators. “Mitch recognized how AI was reshaping traditional industries and that accounting would be no exception,” Balsillie says. “Agile brings accounting expertise and AI together, creating new opportunities for Canadian small businesses to build trusted financial data at scale, and reduce friction across accounting, banking and lending.”

Derksen was interviewed by Umar Choudhry, Director of Education Outreach at RPA Canada, for Professional Accountant Magazine.

Umar: For readers who don’t know you, who are you and what are you trying to accomplish?

Mitchal: I’m a CPA, so not an RPA, but a designated accountant in the same space. I started at a Big Four firm and spent my first decade with large organizations across public practice, private industry and the public sector, and developed a skill set as a software engineer along the way. By 2018 I felt that, ultimately, I was just serving somebody else’s wallet. I left to bring the analytics and forecasting that large organizations spend millions on to small businesses at a rate they can afford.

Umar: You’ve said accountants tend to be late to technology. Why?

Mitchal: My first busy season, I was assigned a testing section with over 100,000 transactions. I pulled the list into Excel, wrote a simple algorithm to filter it, and used a random number generator to pick the 20 samples. It took 15 minutes. My senior said, “That’s not how we do it. We highlight everything in yellow and manually pick 20.” She made me redo it, and it took four hours. It was this mentality of “that’s the way it’s always been done.”

Net-net, I think the caution is a good thing; accountants are a careful, intelligent quality control check. The downside is that we get stuck in old ways of thinking. But the real value of an accountant isn’t in counting the numbers. It’s in translating them into something useful for business owners.

Umar: Walk us through the pivot into bookkeeping.

Mitchal: In 2020 a good friend bought a mountain bike and ski shop. He got the keys on March 9. On March 16, COVID was declared a pandemic and he was told to close. We jumped in, built dashboards and modernized the bookkeeping. Six months later they had freed up half a million dollars in cash simply because they could see what was going on in real time.

That’s when we realized we were solving the wrong problem. Everything runs on data, and the bookkeeping data underneath our models wasn’t timely, accurate or reconciled. When my friends first asked if we’d do their bookkeeping, my partner looked at them like, “We’re designated accountants, we’re not bookkeepers.” Thank God we got over our egos. The value of a trained accountant doing bookkeeping is immense, whether that’s a CPA or an RPA.

We’re called Agile because we run the books in weekly sprints, like the software methodology. Instead of 12 bites in the fiscal year, we get 52. If you won’t see July statements until August 20, how do you make decisions today?

Umar: Where does AI come into it, and how do you handle accuracy?

Mitchal: We build most of our own stuff. The AI everyone knows is generative. What we build is procedural AI, which is good at making procedural decisions autonomously. Feed it a Home Depot receipt: for a construction company that’s probably job supplies; for an office it’s probably repairs and maintenance. It gets complex quickly, and that’s where the frontier models don’t stand up. In the real world they’re only about 80 to 85 per cent accurate at this. And 80 per cent is not good enough, because if you start with data that’s 80 per cent accurate and forecast that out, you end up so far from the mark.

So, we built a three-tiered system. It starts with a review engine that is not AI, coded by accountants to think like a reviewing accountant, with no hallucinations. It doesn’t resolve issues; it highlights them for our human accountants. Then our bookkeeper brain handles categorization, sales taxes, invoices, bank feeds and reconciliation, and our accountant brain reviews that work in the context of the industry and the client and sends it back: “No, that’s not correct, try again.” Our people deal with the exceptions and, more importantly, provide the customer service.

“When Canadian firms own and control their IP and data, they capture more economic value, retain talent and build strategic advantage while generating benefits across the economy. To better support Canadian innovators, all roads lead to building technical capacity, and creating the conditions for them to scale and succeed at home.”

Jim Balsillie, former Co-CEO of Research In Motion (BlackBerry) and Chair of the Council of Canadian Innovators

Umar: Fast forward to 2030. What is an accountant doing?

Mitchal: Accountants are going to be more valuable than they are today, but the nature of that value will shift, and not everybody will get it right. Within five years AI will have automated most of the transactional work associated with a bookkeeper or junior accountant role.

Our biggest early mistake was assuming that fast, accurate, affordable books were a winning combination. We delivered perfect books incredibly fast and still lost customers, because we weren’t adding value by just doing that. Customers want a trusted advisor. We overhauled our service model around communication and regular check-ins, and since then we’ve had one of the lowest churn rates in this industry. There’s AI fatigue now; all you want is to talk to a person. People still see accountants as bean counters. What we actually are is translators.

Umar: What’s your advice for a new RPA or CPA worried about AI?

Mitchal: Ask the firm you’re applying to very specific questions about how it uses AI. If there’s strong hesitance to use it for transactional work, that firm will probably get stuck. But you also don’t want the ones where everything is AI-generated, because they’ll lose customers by not valuing the human connection. Look for the right mix of optimism and caution.

Umar: Jim Balsillie introduced us. What has he brought to how you think?

Mitchal: Through our local tech accelerator I got to know someone mutually acquainted with Jim, and he’s been a fantastic mentor for just over a year. His insights are especially sharp around maintaining sovereignty over our IP within Canada: building Canadian technology and actually retaining the rights to it here.

That’s one reason we’re almost exclusively focused on the Canadian market. It’s easier to go into the States; that market is softer and less price-sensitive. But Canada is often overlooked by innovative products, and we wanted a solution that’s Canadian first, with our IP developed in and for Canadian businesses.

Umar: Any final words for our readers?

Mitchal: Find a mentor. There are mentors everywhere, and all it requires is the boldness to ask. Ultimately, it’s all about relationships, and that’s the future of our industry too. AI will take away the menial work, and what’s left for accountants is the value-add: the advice, the translating. Anyone who finds this compelling, reach out and connect with me.

Derksen’s forecast puts the transactional layer of accounting on a five-year clock and moves the profession’s value toward the owner who needs numbers turned into decisions. That is territory the Registered Professional Accountant designation was built for: the RPA is designed around the practitioner who does bookkeeping, tax, advisory and compilation engagements for the small and medium-sized businesses that larger firms overlook. Recently, RPA Canada’s leadership committed to ensuring its members are future-ready accountants, with renewed emphasis on AI and emerging technologies, so the next generation of RPAs is trained for the work that will remain once the data entry is gone.

The shape of that future is visible in Derksen’s story: an accountant who went back to the least glamorous part of the job, built the tools to do it properly, and kept them here.