As the AI revolution continues apace, data has confirmed that generative AI is making some workers more productive and some companies more profitable, though this does not mean it's a good idea to start cutting staff.
During a virtual roundtable hosted by KPMG last week, Pär Edin, the U.S. AI go-to-market leader for the Big Four firm, said there is hard data showing that, for at least some workers, generative AI has been paying dividends in terms of productivity, referencing research from last year finding that, on average, the technology has introduced
He referenced KPMG's own research on top of this, which found that—after analyzing 10,000 companies—generative AI has an EBITDA impact ranging from 3 to 17%, which is calculated as time freed up multiplied by the labor cost of that time, which he felt was a highly significant impact. Effectively, he said, generative AI has created an entirely new driver for productivity.
"It varies by sector and company, but those are really, really huge numbers. This is an additional lever that didn't exist 18 months ago. Now any company can pursue single-digit or low double-digit percentage points of improvement. Not overnight, but within a 12-36 month period using existing tools," he said.
That implies companies can do more with less, but Edin warned it does not mean companies should start reducing headcount. In fact, he said, generative AI is pretty terrible at fully replacing people, at least right now. While AI is often touted for its automation capabilities, he said over the past few years companies have found this was a flawed conception. The promise of generative AI isn't so much in replacing people but augmenting them.
"It's not a headcount-reduction tool in the sense some may have thought about," he said. "[Instead, it's] really a task augmentation tool. We talked about how to get those numbers. You need to break down the entire workforce. I don't mean headcount but tasks and activities. For every one of those, there are some pretty interesting benchmarks on how much time could be freed up by using better tools. Think of it more as a power tool for the mind than an automation factory."
He understands this might not be what certain business leaders want to hear. Edin noted that he has had many conversations with finance and accounting leaders that basically come down to ROI. This isn't always the easiest to measure, especially when it comes to AI tools, so sometimes it can be difficult to communicate the benefits. If it's not reducing the cost of labor, some wonder, what's the point? Edin, though, felt that focusing on the cost of labor was missing the point entirely.
"The most likely case we discussed was not labor cost or headcount reduction but gradual market expansion," he said. "So, think of it as companies continuing to grow at the same or greater pace on the top line while not growing labor costs and headcount at the same rate—or even keeping them steady."
Given that, by definition, this is more about supporting future growth than directly creating it, he conceded it can be difficult to quickly make back the investment. This has led to a push and pull for accounting and finance leaders between wanting to implement AI for its productivity benefits while, at the same time, wanting to spend only on that which has a direct business case.
"There is a tug of war between wanting to fund this as much as possible, because it does drive productivity, but at the same time not being too overblown about what it will do when explaining this to the board or an investor. This is a balancing act between wanting to do it and being fiscally responsible," he said.
It may be easier to directly communicate the need to adopt AI in the future. Edin broke AI development down into three phases: retooling, reengineering and reimagining. The first phase, retooling, is about doing the same job with the same person and role but just more efficiently than before. He noted most companies are in this phase, rolling out pilots and training their staff. The second phase, reengineering, is where workflows themselves are changed to include AI, which he said serves to free up time and enhance efficiency by not just doing the same job faster but doing a better job overall. Some companies, he said, are just entering this phase. Finally, reimagining is something few to no companies are doing now: thinking about AI as it applies to the entire business model.
"This is when you think about disruption. Will your entire business model be wiped out? Or will you disrupt others? You might go lower in the value stack, or even enter a different market entirely using this technology," he said. "These phases are somewhat sequential but are happening in parallel depending on the company. Most companies sit somewhere between the first two phases."
Agentic AI—where bots are given limited autonomy and initiative—may place companies between the second and third phases, but even then he said it will not mean the end of human involvement.
"There will be many types of tools," he said. "Even in an automated factory, you still have wrenches and screwdrivers. It will be an ecosystem. We'll continue to use many different tools. The AIs are great because they're flexible—they can do things they weren't originally designed to do, and they can get better."
