Andrew Forde has been inside the Bay Street rooms where corporations are hatching their AI plans. The problem, he says, is a lot of those conversations are focused on the wrong priorities. 

KPMG Canada’s head of AI research said some companies are still applying AI to “everything blindly.” But he said many firms are missing a more fundamental shift in how work is organised. 

“When you make a worker more productive, you don’t actually reduce jobs, you just create more work,” he said in an interview, describing a Bay Street that is too busy handing out AI software licences to pursue new lines of businesses.

“There’s a faction, I think, of folks who are still treating AI like a conventional IT project… I think that conversation is starting to happen, but definitely not widely enough.” 

Forde does not shy away from highlighting the fragilities in AI systems. The accounting and consulting firm said it promoted Forde to lead its AI research at the end of March in part because clients needed AI advice grounded in evidence instead of hype—hoping that rigour would set it apart from other Canadian professional services firms. 

Rival firms, meanwhile, are betting big on the technology as they face new competition from AI developers. Accenture, for example, grew its AI-related workforce from 40,000 in 2023 to 77,000 last year. Though 60 per cent of its annual sales came from technology partnerships, generative AI accounted for US$2.7 billion of its US$69.7 billion in revenue. OpenAI has also launched a consulting arm with partners Bain, McKinsey and financial giants like Brookfield. 

KPMG last week announced a partnership with Anthropic to provide tax and private equity tools for firms and their portfolio companies. While the private equity initiative will start in the U.S., KPMG Canada CEO Benjie Thomas expects the partnership to be significant here, because of the firm’s focus on private equity clients. 

Forde’s core responsibility is to ensure companies understand AI’s nuances, he said. In 2009 and 2010 he worked at NASA, where engineers are trained to question underlying assumptions in technology like the Mars Climate Orbiter, which was destroyed in 1999 after a mismatch between imperial and metric units caused a navigation failure. 

Similarly, Forde said, users often ask an AI system to, say, pick a number between 1 and 7 without understanding whether the output is random or shaped by statistical patterns in human preferences. 

Consultants will need to gain client trust by proving they can address such problems with ethics and precision, he said. “Do you care enough, and pay attention enough, and are you curious enough?” 

It’s a concept consultants have struggled with, including KPMG Australia, which faced an internal AI-cheating scandal. EY Canada retracted a paper this month that included made-up information, while Deloitte’s nearly $1.6 million consulting project in Newfoundland and Labrador was marred by a report with fake citations. 

These gaffes come as the consulting industry is already under pressure to prove its worth. The federal government was taken to task over its McKinsey contracts in 2024, and Deloitte’s AI-related errors have resurrected questions about Ottawa’s spending on consultants.

Forde, who has a PhD in engineering from the University of Toronto and still teaches there, is not anti-AI. He has argued that Canada needs to catch up to global peers on AI commercialization, and hopes to see clients benefit if the federal government’s new AI strategy focuses on a handful of sizable bets. 

In Canada, economic leaders have been particularly focused on using AI to boost productivity, as the country’s output has lagged G7 peers. So far, most of Canada doesn’t seem to have fallen too deep in what Forde calls the “hype cycle.” Data from job site Indeed suggests AI saves workers hours per day, and Statistics Canada found last year that about 90 per cent of businesses did not change staffing levels after implementing AI. 

The consulting industry is grappling with these questions internally as well. Last year, KPMG began recruiting junior workers to manage AI agents. It has started to use dashboards to track its workers’ AI usage in its U.S. advisory business after laying off around 400 people in the division. Forde said he doesn’t see tracking logins to AI systems as the best way to motivate change, and has urged business leaders and consultants to look beyond measuring worker productivity and counting AI token usage. 

“The conversation around productivity is an interesting conversation, but it’s not the conversation,” said Forde. “I’m a firm believer in happier employees just do better work. So how do you use these technologies to create better environments for your employees to be creative?”