Inside the Canadian midmarket CFOs’ struggle to make AI work

Listen to the article

7 min

This audio is auto-generated. Please let us know if you have feedback.

Though Canadian finance teams are experimenting with artificial intelligence, some midmarket CFOs there say turning those experiments into dependable financial processes is proving much harder.

At a recent CFO Alliance roundtable in Toronto, finance leaders quickly took the opportunity of a shared space to discuss unreliable data-sharing systems and technology spending that cannot be allocated properly. 

Another topic of discussion, according to Nick Araco Jr., founder and CEO of the CFO Alliance, was AI use happening through employees’ personal accounts, a phenomenon colloquially known as “shadow AI” that Deloitte quantified in recent research overseas.

Araco said the finance leaders raised issues that had not surfaced during the organization’s earlier U.S. roundtables. The discussion offered a look at how Canadian CFOs are approaching the same technology from a different point in the adoption cycle. Some said they are watching companies south of the border move first and are applying some of those lessons.

Vanessa Galarneau was not at the roundtable, but has corporate finance experience on both sides of the Canadian-U.S. divide. The Canadian finance executive is currently working in San Francisco building Pluvo, which the company describes as an “AI-native layer built for strategic finance.” Galarneau is the Canadian technology company’s co-founder, CFO and COO.

In her view, Canadian finance teams have the same appetite for AI with fewer resources available to move experiments into production. Companies in Silicon Valley often have enough growth capital to support research and development throughout the business, she told CFO.com in a recent interview, adding that Canadian CFOs usually face a higher bar for spending without a clear return.

“There’s less margin for error [in Canadian markets],” Galarneau said. That affects how quickly finance teams can hire technical talent and build the infrastructure around AI. It also raises the stakes for every experiment that receives funding.

The invisible AI leak

Inside the Toronto roundtable, the first concern involved AI use happening beyond the company’s approved tools. “The leak isn’t an enterprise tool. It’s the personal account,” Araco said. In an unprecedented fashion, employees today can move company information outside a controlled environment within seconds.

“At that moment, the company’s zero-data-retention contract with its vendor means nothing,” Araco said.

Existing customer agreements add another layer of risk, because many employees may know their company has signed a nondisclosure agreement without connecting it to the spreadsheet they paste into a consumer AI tool. Araco said that contract angle was particularly more prominent in Toronto than at any of the recent U.S. roundtables.

One company represented at the event now blocks browser-based AI and requires employees to use a desktop application restricted to certain drives. The goal, the CFO said, is to keep the output local and limit the information the tool can reach. That policy grew out of concerns about employees relying on settings they did not fully understand.

Another CFO said they had identified themselves as the CFO in the settings and instructed the tool to treat everything as confidential. Araco recalled another participant responding that approach quite candidly, saying “I wouldn’t trust that as far as I can throw it.”

Galarneau assumes some employees at most companies are already using personal AI accounts. The absence of reliable data on that activity makes the exposure harder to manage. “I don’t think anyone has a clean number, and that’s part of the problem,” she said.

When an AI cash forecast loses the ERP

A century-old manufacturer at the roundtable offered one of the clearest examples of AI moving too close to a core finance process. The company started by using ChatGPT to draft financial statement notes for its first bank loan. It later connected Claude to Oracle NetSuite to produce KPIs.

Its controller eventually created an AI skill to build the cash forecast. The connection to the ERP dropped without anyone noticing, and the tool rebuilt the forecast using incorrect numbers. “Nobody caught it until they went looking,” Araco said.

That experience produced a new operating rule for the company. Now, finance can use AI to determine what a report should include and how it should look, but a developer then builds the report directly inside the system.

“AI is for design, and systems are for production,” Araco said. He called the approach “the most practical thing” he had heard across the roundtable series because it gives finance teams a repeatable process for human involvement.

Galarneau said finance must establish which data is authoritative and decide where approvals belong. She also said that the ability to trace an answer back to its source is pivotal. “A lot of pilots prove the model can do something useful,” she explained. “Far fewer prove that it is worth operationalizing and safe enough for finance to rely on.”

The required talent is difficult to find in either Canada or the U.S., Galarneau also said. She explained how companies need people who understand how finance works and can translate those processes into technical infrastructure. This talent gap, she said, is creating an international demand for great finance engineers.

“You cannot just bolt [model context protocol] onto old finance infrastructure and expect it to become AI-native,” she said.

Counting the cost of doing something new

Spending created another problem for the Toronto CFOs. “Token is the cost that nobody can see,” Araco said. One part of the bill a finance leader in Toronto could see was an engineer’s $5,000 spend on AI across 30 days, without knowing which project produced the expense.

Customers were also driving up costs by submitting broad requests to AI agents currently in operation, some CFOs said. Those CFOs and their teams have begun assigning models according to the work being performed and found that a newer low-cost model matched the flagship model on one job. That change helped reduce AI spending by around 30% for that group.

“If I’m driving to the corner store, I don’t need a Ferrari,” one CFO said during the discussion. Another participant summed up the ownership question: “We don’t need a chief token officer.” Araco said one line captured the broader adoption dilemma: “Do I want to be customer 46, or customer 46,001?”

Araco also said he found the Toronto finance leaders more deliberate about governance as they worked through many of the same questions. “Toronto is having the U.S. conversation with the benefit of watching first,” he said.

One finance leader at the roundtable also said AI had expanded what the company could do without creating an obvious efficiency gain. “We’re not more productive. We’re doing things we never did before. How do you price that?” the executive asked Araco and the other CFOs in attendance.

During efficiency gains, accountability remains critical as AI takes on more of the routine work. One participant said they could increase annual report production from 25 to 40 using AI, though every figure would need to be checked line by line. Araco said the finance leader told the room: “I can’t tell a judge the AI said so.”