Gartner expects worldwide AI spending to reach $2.59 trillion in 2026, up 47% year over year. [1] In the same forecast, its own analysts flag the uncomfortable part: CIOs are struggling to prove the value of those investments and demonstrate tangible business outcomes. Spending is compounding. Proof of value isn’t.
I don’t read that gap as a technology failure. The models work. It’s a data problem. There’s a widening gap between the workforce data we collect and the decisions leaders now need to make from it.
CEO
ProHance
What workforce data was built to tell us
We are not short of workforce data, and it’s a mistake to think today’s tools only measure time. The stack is mature. We have decades of productivity data — systems that capture output and quality, process mining that reconstructs how work flows, and analytics that benchmark teams against each other. On its own terms, it works well.
The problem is that every one of those tools was built for a human workforce, and most sit in their own silo. Utilisation lives in one system, output in another, process maps in a third. Each was designed around a person doing the work, and each assumes a manager who can join the dots. That held up when “workforce” meant the people on payroll. It doesn’t anymore. Work today is a blend of human workers (employees, contractors, vendor teams) and a fast-growing layer of digital workers: the copilots embedded in everyday tools, and the autonomous agents that pick up a task, act on it, and hand it back with little human involvement in between. Gartner expects 40% of enterprise applications to embed task-specific AI agents by the end of 2026, up from under 5% at the start of the year.
None of that human-shaped stack extends intuitively to digital workers. We’re running a hybrid workforce on instrumentation designed for a human-only one.
The visibility tax
Here is what those costs in practice. Every time work crosses a boundary, from an employee to a vendor, a person to a copilot, a copilot to an agent, something gets lost along the way: accountability, and a clean trail of who did what and to what effect. I’ve come to think of it as a visibility tax, and it climbs with every handoff. As agents proliferate, it compounds.
The tax is paid in three currencies. The first is a blind spot on the digital half of the workforce, the work now happening inside tools and agents rather than in front of a person. The second is a broken link between activity and outcome, because logins and licences confirm that a tool was opened, not that the work changed. The third is a loss of context. The same number means something very different for a senior engineer, a new hire, and an autonomous agent, and siloed, human-shaped data flattens all three into one.
This is why programmes stall at the same point. Gartner projects that more than 40% of agentic AI projects will be cancelled by the end of 2027, not because of model limitations, but because organizations can’t operationalize, govern, or demonstrate measurable value.The tax also has a sharper edge. Unsanctioned, ungoverned AI use is already a live liability in most large organizations, and it isn’t a future concern.
What leaders actually need from it
The instinct is to buy another dashboard. That’s the wrong reflex. More activity metrics only widen the distance between a signal appearing in the data and someone acting on it, and that distance, often a full review cycle, is where the return leaks away.
Leaders need workforce data to do three things it doesn’t do today. It has to measure AI as behaviour rather than deployment: breadth, depth, and whether workflows genuinely shifted against a pre-AI baseline, with shadow usage surfaced as something you can govern instead of guess at. It has to carry context, through an intelligence layer over the whole system that reads the data you already generate, understands the specific company, function, and role, and moves a manager from data to insight to recommendation to a timely action. And it must resolve into a decision, working less like a report to interpret and more like an analyst sitting beside every manager, one that begins to anticipate rather than only explain wherever the data supports it.
Whoever can see and govern all the work wins
Here’s what it comes down to. Return on AI shows up at the system level, not tool by tool. A copilot’s licence cost tells you nothing on its own. The number worth taking to a board is what a change did to throughput, quality, and cycle time across an entire process. And it’s the autonomous end of the workforce, far more than the copilots, that will carry that financial accountability over time.
The race won’t be won by whoever deploys the most. It will be won by the organizations that can see, govern, and improve productivity across all their work, human and digital alike, and prove what it was worth.
Authored by Ankur Dhingra, CEO, ProHance