Workers sort and bag cereal at the Farm Girl warehouse in Mississauga, Ont. The company uses AI to keep track of its product as many other companies are still experimenting with AI and have yet to integrate it into core operations.Galit Rodan/The Globe and Mail
It’s often said that Canadian companies have been slow to adopt artificial intelligence. But at Safe Software Inc., adoption might be going a little too well.
Six months ago, the Surrey, B.C., company was spending about $20,000 a month on AI tools such as Claude Code and ChatGPT for its 370 or so employees. But as usage has grown, so too have the costs. It’s now spending about $100,000 a month – a five-fold increase in just a few months.
“We really want to get this under control, because that’s over a million a year,” said co-founder and chief executive officer Don Murray. These costs wouldn’t be so much of an issue, he said, if Safe Software had a clear picture of the return-on-investment, but that’s fuzzy in some cases.
It’s one of the reasons Safe Software recently hired a director of AI to better determine how employees should be using these tools and nail down the benefits, financial or otherwise, of all that spending. “I want to make sure there’s an ROI,” Mr. Murray said.
It’s not the only Canadian company hiring people to bring more rigour and metrics to generative AI adoption. Browse job postings these days and you’re likely to see many “AI enablement” positions. EY Canada, Air Canada, Bank of Nova Scotia, Ontario Teachers’ Pension Plan, Coveo and Sage Recruiting have all sought folks to lead or work on wringing value from AI tools and find the right use cases.
A spokesperson for job search site Indeed in Canada said that terms such as “AI enablement” are showing up far more often than a year ago, though it remains a very niche role.
Meanwhile, some companies that make AI software are embedding staff with customers to build custom applications, which is typically called forward-deployed engineering. OpenAI purchased a consulting and engineering firm this year and launched a new business called the OpenAI Deployment Company to work with clients on bespoke solutions. Ottawa-based supply chain software company Kinaxis Inc. is building teams to install with its clients, too. Manik Sharma, chief of agentic solutions at Kinaxis, described it as a new way to work with customers, “rather than selling software and walking away.”
The practice of forward-deployed engineering is not just about helping companies deal with technological issues around AI. “You’re adopting the need to shift the culture and change the way work is done,” said Charbel Safadi, CEO of Zafin in Toronto, which makes AI tools for regulated industries. “If you don’t reimagine the work, you’re going to see what’s happening in the marketplace right now, which is significant cost.”
Studies have been mixed on whether companies are experiencing the promised benefits of AI yet. A KPMG poll last fall of more than 750 executives in Canada found that just 2 per cent said they were seeing results in the form of productivity or cost savings. That’s partly because many companies are still experimenting with AI and have yet to integrate it into core operations.
Such futzing around can come at a cost, however. Uber Technologies Inc. blew through its AI budget just a few months into the year because employees were using Claude Code so much. The company’s president said later there is no clear link between AI usage and better service for customers. Meta Platforms Inc., meanwhile, is taking steps to rein in usage.
The most intense corporate users of AI tools, about 1 per cent of companies, are spending US$7,450 per employee per month on the technology, according to data from U.S. financial software company Ramp. That’s about half the monthly salary of a software engineer. The gap in spending is large, however. The top 10 per cent of firms spend US$610 per employee per month, while the median amount is just US$11.
Canadian employers are paying the price after AI proves unable to replace laid off staff
Some companies are comfortable with rising spend. Shopify Inc. noted increased costs during its May earnings call as merchants made use of the company’s AI tools. But chief financial officer Jeff Hoffmeister said “we’re seeing significant benefits,” while another executive said more than 50 per cent of the code at the company is now written by AI.
Still, other companies are taking a more targeted approach, either to deal with rising costs or prevent them from spiralling out of control in the first place. After the release of ChatGPT in late 2022, many companies dispatched their IT departments to secure enterprise subscriptions for these tools to let controlled experiments flourish – with mixed results. “The practical application of AI ended up being more complicated than people thought,” said Louis Tetu, executive chairman at Montreal-based software company Coveo Solutions Inc. “Boards and management teams are getting impatient.”
Coveo is essentially an AI company, and sells AI-powered search and product recommendation tools to companies. It’s the kind of place where managers have to justify why AI can’t do a job before hiring someone new. Even so, the company recently posted a job for an AI enablement lead to “guide leaders toward the most meaningful AI opportunities while ensuring responsible, aligned adoption.”
Coveo had to crack down on AI use, too. “We did put a stop on everybody firing up every model any time to do whatever, because we saw the bills come in,” Mr. Tetu said. “Not everybody needs the latest version of Claude for every workload.”
Safe Software has not had to dial back access, though Mr. Murray said it’s an option if costs spiral. With the arrival of the company’s new AI director in June, Mehdi Ghotboddini, that might not be necessary, however. He previously worked in data science and analytics roles at a health care provider, and over the next while at Safe, he’ll identify the highest-priority uses and come up with the metrics to judge success. Productivity might not be the only one. “It’s not as simple as saying how much of an employee’s time is saved,” Mr. Ghotboddini said. “It’s how we can evaluate quality, measure speed and customer value.”
Can AI be steered by anything but profit? OpenAI trial offers clues, but no verdict
At EY Canada, chief technology officer Biren Agnihotri has been adding adoption and enablement specialists to help scale internal tools to employees. “They are very important,” he said. “They help turn AI potential into real productivity and client value.”
The firm has created a few roles that didn’t exist a few years ago, in fact. In addition to adoption specialists, which is more of a change management role, EY Canada has been seeking people to create training materials for AI tools, others to design and monitor workflows for agents, and finance professionals to keep tabs on the costs and benefits.
Given how quickly AI is changing and the uncertainty around how the technology will affect labour, it’s possible some of these new roles might not be necessary in a few years. Mr. Agnihotri pointed out that EY has some 100 AI use cases in production and more in development. “It’s giving us that level of confidence that these roles are going to be long-standing,” he said.
Earlier this year, Scotiabank unveiled a custom-built AI platform called Scotia Intelligence that relies on a variety of models and allows employees to build agents to take care of some office-related tasks, among other features. Now, it’s hiring AI enablement engineers to work on the platform and help employees make use of it.
“We basically hold their hand as they come along this journey, and then we have a better shot of consistency and success,” said Valli Musti, senior vice-president of enterprise data and AI platforms. Employees have been submitting ideas for use cases, and every week her team meets to discuss what can be acted on.
Jennifer Boville, founder and CEO of Farm Girl, says her company hired a head of AI enablement who led the development of a tool that pulls together information from across the company and replaced some existing software it was using.Galit Rodan/The Globe and Mail
It’s not only companies steeped in traditional office work seeking AI enablement help. Farm Girl, which manufacturers sugar-free cereals from a factory in Mississauga, hired one a few months ago. AI adoption had been ad hoc, with the company using ChatGPT and Claude at the office. “It definitely saved time, but we were quickly realizing we needed to go a level deeper,” said CEO Jennifer Boville.
Farm Girl brought in a head of AI enablement a few months ago, who led the development of an AI platform that pulls together information from across the company, including the factory. “It gives us this dashboard that has visibility into all of the critical metrics we’re watching in real time, because our factory runs 24 hours a day,” she said. It’s also replaced some existing software Farm Girl was using.
But how would the company quantify the benefits? That’s still a bit tricky. Ms. Boville thinks about it like this: Without real-time information from the factory, staff might not be able to intervene if something is amiss, leading to a spoiled batch of cereal. “It’s a savings of time, it’s a savings of money,” she said. “Some of these things are hard to quantify, right?”
With reports from Sean Silcoff