Four global leaders reveal why AI investment is outpacing upskilling, and what closes the gap
New global research from Accenture has put a number on a gap HR leaders have been describing anecdotally for months: 86 per cent of organisations plan to increase their AI spending, but only just over 40 per cent plan to upskill their people on how to use it.
HRD Australia put that gap to four leaders working on the front line of AI and culture: Tenielle Colussi, managing director of talent and organisation at Accenture in Australia and New Zealand ANZ); Bryan Stallings, chief evangelist at Lucid Software, based in Salt Lake City, Utah; Jessica Guistolise, evangelist at Lucid Software, based in Minnesota; and Dave Garrison, CEO and co-founder of Garrison Growth, based in Washington, D.C.
Across four separate conversations, a consistent message emerged: the technology is rarely the reason AI programs fail. The reason is almost always how – or whether – organisations bring their people along.
The upskilling gap, in the numbers
The scale of the disconnect is significant. Workers are already feeling AI's benefits at the task level – 68 per cent say it saves them time on routine work – but only 19 per cent feel they actually have the skills to succeed with it, according to Accenture's research. That gap between time saved and confidence built is, in Colussi's words, exactly where organisations are leaving value on the table.
The organisations that close that gap are pulling ahead financially. Accenture's global research, drawing on close to 6,000 executives and employees, identifies just 18 per cent of organisations as genuine "Talent Reinventors" – companies with a talent strategy fully integrated with AI. That 18 per cent grew revenue 1.8 percentage points higher than their peers in 2025.
"Only about 18 per cent of the organisations kind of showed six key characteristics that would drive revenue increase and profit increase for them as a result of the AI programs they're putting in," Colussi said. "Which is kind of concerning when you flip it and go, well, what are 82 per cent of those organisations doing?"
Rebuilding the job around AI, not bolting it on top
Colussi said the imbalance between technology spend and skills investment is the core problem. "Organisations are investing in AI, but only 40 per cent are investing in the skills that come along with AI," she said.
"When you look at skills gap and you look at the return on investment, that's kind of where your ROI is, is in the application of the AI. We're not actually seeing the investment in that."
She was also sceptical that classroom-style training changes behaviour on the job. "When people get back onto their jobs, onto the tools, they've probably only got about 10 per cent retention of what they learned in an academy or an online course or like a face to face training," Colussi said.
Instead, she pointed to organisations building learning directly into daily workflows – for instance, giving a call centre worker AI-generated prompts to choose between in real time, with the system learning from each choice made.
Rather than layering AI on top of existing roles, Colussi argued the more effective approach is to pull jobs apart and rebuild them from scratch. "How do we decompose the jobs and then rebuild them with AI in the loop and human in the lead, and make sure that we're getting the best out of the employees and the best out of our tech investment as well," said Colussi.
Why visible experimentation from leaders builds trust
At Lucid Software, Stallings described how openness from senior leadership – rather than a formal mandate – drove grassroots adoption inside his own team. "I would say that the biggest goof-off with AI has been our boss. I mean, he tries everything," he said of the company's chief marketing officer. "He's sharing his experiments and he's talking about it frequently with everyone, and he's encouraging us to try things out as well."
That openness changed how comfortable employees felt experimenting on work time. "At first when we had access to the tools, it felt a little bit like cheating to check it out," Stallings said. "But then he comes in and he's like, here's what AI did for me this weekend and look at these results. I liked this, I didn't like this, here's the things I've learned."
Stallings also raised a caution for HR leaders about what culture loses when organisations cut layers in pursuit of AI-driven efficiency. "Culture requires an interconnectedness within an organisation," he said. "As we see organisations cutting out layers, oftentimes that's where the interconnectedness was … somebody who's there focusing on the development of human beings, professional development, developing high-performing teams, and suddenly we might say, well, we don't need those roles anymore."
Making experimentation and failure safe to share
Guistolise, also at Lucid Software, said the same openness has to extend to sharing what doesn't work, not just what does – including outside of work entirely.
"We had somebody come up with – they vibe-coded an app so that nobody would ever have to argue about what's for dinner anymore as a family," she said. "It's that kind of thing that makes us – experimentation safe and safe across the organisation … something that's happening over in sales might give me an idea about something that I never would have thought of as a way to use something over here in my role."
She said naming ambiguity directly, rather than glossing over it, is what makes AI transformation land. "AI transformations are not going to work if we are not paying attention to the people who are in the midst of them," Guistolise said.
"I think providing clarity and providing as much guidance along the way as we can, knowing that everything's changing really quickly and being clear about that too … let's document it, what we know for now, and let's work together on that."
Guistolise also questioned whether workforce cuts should be labelled a productivity gain at all. "I think there are some organisations that talk about it in terms of, we were able to cut 10 per cent of our workforce," she said.
"Well, is that really a productivity gain, or did you miss the point? Are you figuring out how to use all of that human energy to put towards accomplishing your mission and vision with 10 per cent more human fuel to do that?"
Positioning AI as a purpose tool, not a mandate
Garrison, whose firm advises companies ranging from sub-billion-dollar businesses to organisations with revenues in the tens of billions, said the pattern he sees repeatedly is executives treating AI adoption as the objective itself, rather than a tool serving a broader business purpose.
"The issue is not how do we implement AI," he said. "The issue is how do we position AI within our firm as it relates to our compelling purpose, our values, and our business objective? How do we allow people to see AI as a tool to accomplish things we already know, as opposed to a new thing?"
He was direct about where he believes accountability sits for AI-related layoffs, describing a comment he had seen from one executive criticising cuts made elsewhere. "It was cowardice," Garrison said.
"You could have gone to your board and said, I'm going to take these saves and I'm going to apply them to redevelop the skills of our employees for the future. But instead you went and cut all of those people and delivered some savings to the bottom line. That's very short-term savings."
Garrison said HR's role is to surface the anxiety employees are already feeling rather than let it go unaddressed. "There's really an opportunity for HR to partner and be the voice of, here's what people's suspicions are," he said.
"If we can position this as just another tool in our arsenal that people can learn how to use and discover how to apply themselves, it will go a lot further than people feeling like it's being forced down their throat."
He also pointed to a practical reframe for time-poor employees – the same barrier Accenture's Australian data flags as the biggest one locally. "If I say to them, hey, do you have any great ideas about how to do things better or serve customers better, 100 per cent of the hands go up," Garrison said.
"And then if I say, how many of you have time to get to those things, it's like 10 per cent of the hands go up. So our challenge is time … if we start to think about AI as a way to give us time to get to more important things, it suddenly is not a technology tool at all. It's a freedom tool."
What does successful AI adoption actually look like in practice?
Garrison described a multi-billion-dollar company that initially appointed an AI advocate in every department and asked each one to find a problem AI could solve. Six months on, the results hadn't matched expectations.
"They are retreating to say, okay, let's break into small groups instead of meeting as a large group, let's break into small groups and share ideas on what's working," Garrison said, describing a shift toward low-risk, informal experimentation over top-down directives.
Stallings offered a manufacturing-era analogy for where he believes most organisations currently sit. He described factories that initially replaced a single steam engine with a single electrical engine, keeping the same long drive shaft running through the workshop – missing real efficiency gains until smaller motors were distributed throughout the floor.
"It feels like we're at that stage where we still have a lot of individual use, and we're trying to help companies understand how to adopt it as an institution," he said. Guistolise summarised the implication bluntly: "Essentially, we have to redesign the factory floor again."
Colussi tied the same idea back to Accenture's research, noting that Talent Reinventors were more likely to build talent mobility into their strategy than to rely on external hiring.
She also referenced a case she recalled reading about, of an organisation that let go of its entire junior workforce to bank productivity gains from AI, only to face a capability shortfall years later. "It's a nice example to kind of complement what we've been talking about," she said.
The throughline across all four conversations – and the new Accenture data – is that culture and communication, not the sophistication of the tools themselves, will determine which organisations convert AI investment into real performance gains.
As Garrison put it, technology adopted without context becomes just another item on an already overloaded to-do list: "The risk is it fails not because the technology doesn't work, the technology is great. It's because it's one more item on a long list of to-dos, and we're under pressure, and I'll get to it when I get to it."
That reframe – from mandate to invitation, and from productivity extraction to reinvestment in people – is one HR leaders are likely to keep returning to as AI adoption accelerates through 2026.
