Why net value, not hours saved, should guide HR’s AI playbook: experts
Artificial intelligence is threaded through everyday work, but the actual gains may not always be clear to organizations, according to a new study.
The global study from Workday found that about 37 per cent of the time saved through AI is being erased by rework, as employees go back to correct or rewrite low quality output. That means that for many organizations, every 10 hours of apparent efficiency may be dampened by nearly four hours that disappear into fixing mistakes and filling gaps. Only 14 per cent of employees consistently achieve net positive outcomes from AI use, according to the study.
Another survey by Section AI of 5,000 knowledge workers across large companies in Canada, the U.S. and the U.K., found that a quarter of workers report saving no time with AI and another 44 per cent save less than four hours per week — well below what most organizations would need to see to justify the cost and disruption of large-scale AI adoption. Only 32 per cent report saving four hours or more weekly, with a mere six per cent hitting the 12-plus-hour threshold that would signal transformative change.
A notable example of disconnect: 81 per cent of C-suite members believe their company has a clear, actionable policy that effectively guides AI use, but only 28 per cent of individual contributors agree, according to the knowledge workers survey.
Speeding up isn’t always improving performance
It’s a reminder for HR leaders and their organizations that speeding up tasks isn’t the same as improving performance, according to Fredrik Odegaard, associate professor of management science, at Western University in London, Ont. “Companies certainly so far haven't seen these huge benefits that were promised by AI, and these efficiency gains don’t materialize on the bottom line,” says Odegaard.
However, Odegaard notes that he hasn’t necessarily seen the specific circumstances where employees have to deal with a significant amount of rework.
“I think the bigger problem that I experience in my interaction with leaders and businesses is that even though these tools may produce time savings in terms of certain tasks and that even though in principle it's supposed to enable workers to spend more time on higher value-added activities, you still don't see a bottom line on that,” he says. “So if it doesn't generate either additional revenue or a measurable cost savings, it's not quite clear where the return is showing up.”
Workday’s data show how quickly time savings can be diluted. Nearly 87 per cent of employees use AI at least a few times a week and 77 per cent say they are more productive because of it, yet more than one third of the time they save is quietly lost to rework and verification, according to the study.
In addition, heavy daily users, often aged 25 to 34, are the most likely to scrutinize AI output, with 77 per cent checking it as rigorously or more rigorously than human work. At the same time, just 37 per cent of this group report better access to skills training, despite their high exposure to AI driven tasks.
Redesigning roles around AI
The Workday study also highlights how rarely organizations redesign work around AI. Across the sample, nearly nine in 10 respondents say fewer than half of roles have been updated to include AI-related skills. In organizations struggling to achieve net gains, less than 25 per cent of roles are considered AI ready.
The fundamental reason organizations are leaving value on the table is that they’re treating AI as a “plug-and-play” replacement for tasks rather than a catalyst for redesigning work, according to Aashna Kircher, group general manager for CHRO products at Workday. “Many companies have simply layered advanced technology over rigid job structures, which creates friction when the speed of the tool exceeds the capacity of the existing workflow,” says Kircher.
Odegaard sees that disconnect in sectors that have embraced digital tools, including Canadian health care. He points to AI scribes that automatically generate clinical notes from patient visits. In theory, these tools free physicians from administrative work, he says.
“So these AI scribes can make a doctor be much more efficient in meeting with the patient, and they make fewer mistakes, perhaps, in transcribing their notes,” says Odegaard. However, he says that the expected productivity dividend doesn’t show up in system-level metrics. Instead, physicians’ days simply end closer to on-time with less paperwork spilling into evenings rather than more patients being seen, says Odegaard.
For HR leaders, that example underlines a critical point: efficiency on individual tasks doesn’t automatically translate into organizational performance. Without redesigning roles or service targets, or properly assessing what AI is doing for an organization, saved minutes can simply evaporate into lower stress and shorter days rather than better results, according to Odegaard.
Measuring net value, not just hours saved
The Workday study suggests that organizations should track the net value of AI rather than focusing only on gross efficiency. It also urges leaders to shift from speed-based KPIs to outcome-based metrics. According to Odegaard, each organization has to align its AI initiatives with the formal metrics it uses to see how savings are manifested.
“There are different efficiency measurement tools — for example, there's one mathematical model called data development analysis that measures these productivity measures a certain set of inputs along with some process, and then that converts that into output,” says Odegaard. “So it’s how efficient are you in creating these outputs from these inputs — which variables to measure is highly dependent on the context of the industry.”
Kircher believes that HR leaders need to stop measuring how much time AI saves and start measuring the quality of what it actually produces, and that means tying AI metrics to existing scorecards. “Traditional metrics like hours saved can be very misleading because they don't account for the hidden time employees spend fixing AI mistakes,” she says “We should look at the entire lifecycle of a task to see the real impact — for example, in recruiting, it doesn’t matter if AI can draft a job description in seconds if the resulting interviews don't lead to better candidates, so by shifting the goal from ‘how fast can we fill this role’ to ‘how successful is this new hire,’ leaders can see the true cost of their AI strategy.”
Reinvesting AI gains in people, not just technology
Workday’s research also finds a clear link between reinvestment choices and net AI value. Organizations that direct a greater share of AI-related savings into skills and workforce development — rather than solely into more technology and infrastructure – are more likely to report durable performance gains. Yet, on average, leaders still allocate 39 per cent of cost savings to technology and only 30 per cent to workforce development, and just 29 per cent of respondents say they reinvest time savings into employee AI upskilling, according to the study.
The Section AI survey found that 97 per cent of the workforce are using AI poorly or not at all, and only 2.7 per cent qualify as "AI practitioners" — people who integrate AI into their workflows and see significant productivity gains — and a mere 0.08 per cent are considered "AI experts."
The same survey revealed that just 27 per cent of individual contributors have received company AI training, versus 48 per cent of managers, 68 per cent of directors, 70 per cent of VPs, and 81 per cent of the C-suite.
Odegaard believes that many employers underestimate the depth of change required when they introduce AI tools into established workflows. “It means that people have to completely change the way they work, their behavior, and the way they’ve done things. That's a big ask,” he says. “It's not like an instant switch that. just because you put the system in that all of a sudden it's going to be great — training is absolutely going to be crucial and being able get the workers to align their processes with how the technology works and its capabilities.”
Training gap can lead to ‘AI tax’
On training, the study highlights a misalignment. While two-thirds of leaders cite skills training as a top investment priority, only 37 per cent of the heaviest rework group report increased access to training. Without closing that gap, HR risks amplifying the “AI tax” — pushing more responsibility for quality assurance onto employees who have not been given the time or tools to adapt.
“Organizations are rushing to produce more output without developing the new skills and processes needed to turn it into a high-quality finished product,” says Kircher. “This means employees have to do the extra work themselves, wasting the time the new technology was supposed to save.”
According to Kircher, the organizations getting the most value from AI are the ones pairing the tools with a plan for how people use them. “In our research, leading organizations treat the time AI frees up as something to reinvest in their people, not just as extra capacity to check off more tasks,” she says.
This means carving out time for targeted skill development, updating roles and expectations so managers an HR are evaluated on their decisions and people outcomes, and protecting some of the time AI saves for “human connections” such as one-on-one meeetings and cross-functional problem-solving, according to Kircher.
