If you work in accounting, bookkeeping, or tax, you’ve probably asked yourself some version of this question: is AI coming for my job? On a recent episode of Your Business Unleashed, we sat down with Peter McCarroll, CPA, CA, owner of Fuel Accountants in Toronto, to talk about exactly that. Peter has spent the last couple of years helping accounting firms figure out how to use AI the right way, and the conversation ended up covering everything from security risks to what accounting careers will look like in ten years.

Here’s what stood out most.

Most Firms Don’t Have an AI Problem, They Have a “What Are We Even Trying to Solve” Problem

Peter’s biggest observation from working with firms across the industry is that almost nobody struggles with a lack of AI tools. There are hundreds of vendors, dozens of platforms, and no shortage of demos promising to change everything. The real issue is that most firms jump into AI without ever answering a basic question: what problem are we trying to fix?

Without that answer, even the best tool only delivers small, marginal improvements. It’s a leadership issue long before it’s a technology issue, and it applies just as much to law firms, engineering firms, or any business built on knowledge and expertise.

What are the Three Levels of Using AI in Your Business (Most People Only Use Level One)

One of the most useful frameworks from the conversation was Peter’s breakdown of the three levels of AI use:

Level one: efficiency. This is using AI to write emails, draft letters, or speed up routine work. It’s helpful, but the research shows it only saves about 15 to 20 percent in cost, and there’s a ceiling. Push too hard on efficiency and you start hurting the client experience.

Level two: strategic thinking. This is where AI starts to earn its keep. Peter records his client and sales calls, then has AI review the conversation and point out what he missed or could have handled differently. He’s even set a standing instruction telling the AI not to flatter him, but to critique him honestly, because by default, most AI tools are people-pleasers.

Level three: growth. This is where firms create entirely new services and revenue streams instead of just doing old work faster. Because there’s no existing benchmark to compare against, the upside is much bigger than squeezing out a bit more efficiently.

Most businesses never get past level one. The firms that will pull ahead are the ones willing to go further.

Why is Canadian Bookkeeping Harder Than American Bookkeeping (and Why That Matters for AI)

A lot of AI accounting tools are built for the U.S. market, and that creates a real gap for Canadian firms. In the U.S., cash-basis filing and simpler tax rules mean bookkeeping can often be done quickly with minimal detail. In Canada, GST and provincial sales tax requirements mean every transaction needs more context, and in provinces like B.C. with multiple sales taxes, it gets even more layered.

This is part of why Canadian accountants can’t always adopt the same shortcuts that work south of the border, and why source documents and proper coding still matter so much here.

What are the Risks of Using AI With Client Data

This part of the conversation is essential for reading for anyone handling client information. Peter breaks security down into three levels every firm needs to understand:

First, make sure your AI subscription doesn’t train your data. Anything you type into a model that learns from your inputs could theoretically resurface later in someone else’s conversation. Peter’s example: enter a client’s tax return into a free, untrained data model, and months later that same information could show up if someone asks about that person’s income. The fix is simple: only use paid, business-grade plans where your data isn’t used to train the model.

Second, understand how long your data is retained. You want enough retention to pick up a conversation where you left off, but not indefinite storage of sensitive files.

Third, and this is the trickiest one, make sure information doesn’t bleed between clients. If a tool has “memory” turned on, details from one client’s file can accidentally surface while you’re working on someone else’s. This is especially important when connecting outside systems, since those connections often apply to your whole account rather than one specific client folder.

Will AI Take Accounting Jobs?

Rather than a simple yes or no, Peter frames this as a shift rather than a straight loss. The traditional path into accounting has always involved years of repetitive, detail-heavy work, learning the basics by doing them repeatedly. That kind of work is disappearing fast.

What’s replacing it is a much bigger role for review and judgment. As AI takes over the data entry and first-pass preparation, the people who used to spend years grinding through files will move earlier into reviewing exceptions and anomalies, the part of the job where real expertise and client value live.

The comparison Peter uses are a good one: think of it like training airline pilots on a simulator before they ever fly with real passengers. Firms will need to find new ways to teach judgment and client instincts without relying on years of repetitive transaction work to build it.

What Does the Future of Accounting Profession Look Like?

Here’s the part of the conversation that feels most hopeful. As AI takes over more of the technical and repetitive work, the accountants and firms who succeed won’t be the ones with the fastest software. They’ll be the ones who double down on real relationships.

Peter talks about this in return to something closer to how business used to work: more phone calls, more coffee meetings, more actual conversations with clients. When technical work becomes commoditized, human connection becomes the thing that keeps clients loyal.

He describes the future of client service using a simple analogy: accountants have spent 200 years looking in the rear view mirror, reporting on what has already happened. AI is quickly turning that into a dashboard, giving real-time information. But what clients really want is a co-pilot, someone who can look through the windshield with them and help them see what’s coming next.

That’s the shift worth paying attention to. Not “AI versus accountants,” but a chance for accountants to spend less time on data entry and more time doing the part of the job that builds long-term client relationships.

Want to hear the full conversation with Peter McCarroll, including his take on Claude Cowork, prompt injection risks, and why he’s still cautious about rolling out AI agents across his team? Listen to the full episode of Your Business Unleashed.