If Canadian insurers get their AI transformation right over the next two to three years, the results should be visible well beyond the technology itself, according to Sonia Sood (pictured), national human capital insurance leader at Deloitte. She points to productivity, customer experience, and leaner, more supervisory teams as the clearest signals of success.

"I would expect organizations to be more productive," Sood said. "We have to see that as an outcome."

She said a properly executed transformation should also show up in how customers experience their insurer, not just in internal efficiency metrics.

"I would expect insurers to have better customer experiences, so their NPS scores or CSAT scores are going up, because their employees are equipped with better tools," Sood said. "There's more automation where it matters. I would see less administrative transactional tasks being performed."

Read more: AI job elimination fears are 'a bit of a myth,' Deloitte's Sood says

That shift, she said, will change what roles inside insurance organizations actually look like day to day, with a noticeable move toward oversight-style work rather than the transactional tasks that dominate today.

"You would see changes in roles and responsibilities. So you'd see a lot more supervisory roles, in this case overseeing what AI does," Sood said. "I think you'll always need humans in this, but what they will do and their ability to now use AI in their day-to-day is going to feel very different in the future."

Right now, she said, most organizations are still in an earlier, more experimental phase of that shift.

"Right now I would say generally the workforces are experimenting day to day," Sood said. "In the future it's going to be so blended into the way that we work every day that it'll become the new norm. But right now we're still not quite there. We're evolving to that state."

Asked directly whether she expects insurance teams to become leaner as a result, Sood said yes, though not uniformly across every function.

"I would expect teams to be leaner in the future state, given all of the investments that are being made here," she said. "Every CEO, CFO wants to know how can I do more with less, or how can I at least stay steady and not have my FTE base continue to rise."

That doesn't mean every team shrinks at the same rate, she added. Some functions, particularly those tied directly to AI and technology, may actually grow even as others become leaner through automation.

"Some teams may stay lean, and actually some teams may grow, as AI evolves. Your technology teams, your AI teams, may actually grow in the future state, and other teams may become leaner because much of the work gets automated," Sood said. "But I'd say on the whole, I would expect teams to be leaner in the future, as administrative transactional tasks reduce and we're able to increase productivity overall."

Getting to that future state, however, depends on insurers closing several gaps that Sood said still exist across the industry today. The first is reskilling. As AI takes over parts of roles like underwriting, the people who previously did that work are left with time and capacity that needs to be redirected somewhere deliberate, rather than left undefined.

"If we're going to eliminate components of a job, let's say you take an underwriter and some is going to be completely shifted to either it's going to be automated or you're going to introduce AI to components of it, then that arguably creates some capacity," Sood said. "So what do we want them to do with that? Is it that we want them to focus on more complex work? Is it that we want them to spend more time with customers? Is it that we want them to do different kinds of analysis?"

That question, she said, only gets answered properly if insurers invest in building AI fluency across their workforce, not just among leadership.

"We need our employees within the insurance sector across all functions to actually increase their AI fluency," Sood said. "Insurers need to invest more here to build that baseline, so that we're doing things in a safe but also responsible way in the organization, but it's also yielding positive outcomes."

Read more: Access to AI is not the same as being ready for it

The second gap is structural: most insurers are still running AI initiatives as a separate track from their core, day-to-day operations, rather than genuinely folding the two together.

"Right now we're at this kind of inflection point where use cases were happening over here, business as usual is happening over here," Sood said. "And now we actually need to get these two interlocked together. And the way to do that is actually reimagining the end-to-end workflow."

The third gap is governance. Sood said the frameworks insurers have historically relied on for data and analytics oversight need to expand to account for AI specifically, and that expansion brings its own tension between control and accessibility.

"Up until now the focus has been around data and analytics, and now it's data, analytics and AI," she said. "What does that governance structure need to look like? How do we keep enough at the centre, but how do we democratize the capability and the usage so that it can actually be utilized on a day-to-day basis, but also protecting the organization from all sorts of risks that are associated with it?"

The fourth gap, according to Sood, is measurement. Despite ambitious claims made about what AI can do for insurance organizations, she said many companies still lack a clear framework for evaluating whether their AI investments are actually delivering results.

"There's a lot of big statements made around the promise of AI, so understanding what the baseline is," Sood said. "But then once you actually implement some of these tools and capabilities, what is it actually yielding for the workforce and the organization? Is it improving your customer experience? Is it actually improving the margins on your work? Are your employees actually feeling fulfilled in this type of work now that they're working with AI?"

Closing these four gaps, Sood said, is what will ultimately separate insurers that get lasting value from AI from those that simply layer new tools onto old ways of working, and determine whether the future she describes actually materializes on schedule.