AI Is Freeing Up Hours, but CEOs Cannot Find Them on the Bottom Line

The productivity story executives were promised has arrived, at least on paper. CEOs now rank AI ahead of process redesign as the main reason their organizations got more efficient this year, and workers say the tools hand them back the better part of a working day each week. The problem is what happens next. For most companies, the freed hours are not showing up as revenue, margin or growth, and very few leaders can see clearly where the time actually went.

The Gains Are Real. The Results Are Missing.

EY’s latest CEO Outlook, released on October 1, surveyed 1,200 CEOs of large companies across 21 countries between August and September. Half named AI technologies as the single largest contributor to their productivity gains over the past 12 months, narrowly ahead of business process redesign at 46%. Nearly half, 48%, say they are reinvesting the extra capacity into innovation and transformation.

But 23% say AI-driven gains are often absorbed before they reach bottom-line results. Measurement is the deeper gap: 56% of respondents have at least some reliable visibility into AI costs and returns, yet only 16% have clear, real-time visibility of ROI, and 24% can track some AI costs while having limited ability to judge the overall return. Andrea Guerzoni, EY-Parthenon’s global vice chair, put the problem plainly: “AI is creating real productivity gains, but productivity alone is not a strategy.”

The reinvestment pattern also tilts away from people. Only 15% of respondents say productivity gains are being put back into workforce development, even though 47% say their organizations are not developing AI skills quickly enough and 72% expect skills shortages to become a bigger barrier to growth than access to capital within three years.

Where the Freed Hours Actually Go

The worker view explains a good part of the leak. Glean’s Work AI Index, based on a survey of 6,000 full-time digital workers in the U.S., U.K. and Australia, finds that workers estimate AI automation saves them about 11 hours a week. Only 13% say their organization is performing significantly better because of AI.

Some of the difference is spent making the tools work. Glean finds workers spend 6.4 hours a week botsitting: 2.3 hours feeding context, 2.2 hours supervising outputs and 1.7 hours debugging. That is more time than they spend using AI to produce work. Thirty-six percent of AI sessions fail outright and need a restart or substantial rework, and 77% of workers have corrected or redone AI-assisted work in the past month.

Fragmentation makes it worse. Seventy-seven percent of AI users work with multiple AI tools each week and a third use four or more, while 60% rerun the same prompt across tools because the first output was not good enough. Workers juggling several tools are 35% more likely to report frequent botsitting. An hour saved drafting a document is easily spent reconciling three versions of it across three assistants.

Expectations Are Rising Faster Than Headcount

Leaders are planning as if the gains will compound regardless. Chief Executive’s October CEO Confidence Index, based on 321 U.S. CEOs polled October 6 and 7, found 85% expect higher revenue in 2027 and 80% expect higher profits. Yet only 37% expect to end 2026 with more employees than they had in 2025, down from 45% in September, and hiring is the weakest of their growth forecasts.

The pattern reaches well below the large enterprise. The SBE Council’s October Small Business Check Up, a survey of 516 owners of firms with 2 to 99 employees, found 58% use AI agents regularly or occasionally, and 28% of owners reporting stable or improved results credit adopting technologies such as AI. Growth targets are going up, staffing plans are flat, and AI is expected to cover the gap. That only works if the time it frees is pointed at specific work.

Glean’s data suggests what separates the organizations that do get there. Those reporting transformative impact track an average of five AI metrics rather than three, with revenue growth among them 43% of the time versus 24% elsewhere. Where quality is tracked, 83% of workers say AI improved work quality, compared with 68% where only productivity is measured. Organizations that count only hours saved tend to get hours saved, and not much else.

A Capacity Conversion Playbook

• Assign freed hours before they appear: When a workflow is automated, decide in advance which backlog, customer or revenue work the recovered time goes to. Unassigned capacity is how 23% of CEOs end up seeing gains absorbed before the bottom line.
• Measure output, not time saved: Add quality and revenue measures to AI dashboards alongside productivity. Transformative organizations in Glean’s data track five metrics, not three.
• Budget for botsitting: At 6.4 hours a week, the work of feeding context and checking output is a real cost. Count it when you calculate net savings, and design workflows that give AI the context it needs up front.
• Cut the tool count: A third of AI users run four or more tools, and multi-tool users botsit more. Consolidate onto fewer platforms connected to the systems where work and data already live.
• Reinvest some of the gain in people: Only 15% of CEOs are putting productivity gains back into workforce development, while 47% say AI skills are not developing fast enough. Fund the skills that turn saved time into better work.
• Close the ROI visibility gap: With only 16% of CEOs seeing AI returns in real time, attach a cost and outcome owner to every AI deployment so results can be read monthly, not annually.

AI has done its part of the job. It has given organizations time back. What happens to that time is an operating decision, not a technology one, and the companies that decide on purpose where the hours go will be the ones whose productivity gains finally show up in the numbers.