The AI productivity mirage: Executives see eight hours saved. Workers see almost nothing.

Ask C-suite executives how they feel about AI and nearly three-quarters say excited. The people doing the actual work? Different answer entirely. Anxious. Overwhelmed. That's what almost 70% of non-management employees told researchers. We know which group is closer to reality.

You sense this if you work in a large company. Same building, different weather depending on the floor. Up in the corner offices, AI saves time. Down where the actual work happens, AI creates more of it. Two new surveys, one from Section covering 5,000 white-collar workers and another from Workday polling 1,600, confirm what cubicle dwellers already suspected: the productivity miracle lives mostly in executive imagination.

The numbers don't lie

Two-thirds of non-management staff say AI saves them less than two hours per week. Some say it saves nothing at all. Executives tell a different story. More than 40% claim eight hours of weekly savings. A full workday that nobody downstairs is experiencing.

The gap goes beyond hours. Section, an AI training vendor and OpenAI services partner, surveyed workers and found that 97% are either using AI badly or avoiding it entirely. Just 15% have found what researchers call "value-driving use cases," applications that might actually affect the bottom line. What's the most common use? Replacing Google searches. Worth noting: Section sells AI training, so they benefit when companies invest more. But Workday and PwC data tells the same story.

Steve McGarvey designs user experiences in Raleigh, North Carolina. He's 53, and part of what he does involves making websites work for visually impaired users. He's tried getting help from large language models.

"I can't count the number of times that I've sought a solution for a problem, asked an LLM, and it gave me a solution to an accessibility problem that was completely wrong," McGarvey said.

So now he spends hours arguing with the AI, walking it through why its suggestions break accessibility standards. The tool meant to speed him up has slowed him down. That's the pattern nobody talks about in the pitch meetings.

The estimation problem

Here's what rarely makes it into the AI pitch decks: unpredictability is its own kind of tax.

Dan Hiester works as a user-experience engineer in Seattle. He's 44. One summer afternoon, he asked an LLM to fix a coding problem. Should have taken thirty minutes. Took his entire afternoon instead. But a different task, something that would have eaten days before AI? Done in twenty minutes.

"It's done a complete reset of my understanding of how to estimate the time it takes to do something," Hiester said. That line deserves a pause. Project management runs on estimation. So do deadlines. So does your credibility with your boss. AI doesn't just change how long things take. It wrecks your ability to guess. Thirty minutes might stay thirty minutes. Might become four hours. You won't know which until you're already committed.

Workday calls the broader problem the "AI tax." Their survey found 85% of employees claiming one to seven hours of weekly time savings. Sounds good until you look closer. Much of that time went straight back into fixing AI mistakes and rewriting AI drafts. Savings on paper. Gone in practice.

The use case desert

Section's research found a problem that more training won't fix. They named it the "Use Case Desert."

You've learned to prompt. You understand the interface. You sat through the training, watched the slides, maybe got a certificate. Then you go back to your desk and realize you have no idea what to actually use AI for. Twenty-six percent of workers in the survey admitted they don't have a single work-related use case. Another 60% said their applications stay at beginner level.

The data shows the consequences. Almost a quarter of workers, 24%, save zero time with AI. Another 21% save less than two hours weekly. Add those up: 45% of the workforce getting nothing or close to it.

Most organizations need ten-plus hours saved per employee per week to justify their AI spending. Almost nobody is getting there.

The view from the top floor

Walk into an executive AI briefing and you'll see confidence. Charts pointing up. Stories from teams that got it working. Buzzwords about transformation. Easy to feel optimistic when that's the air you breathe.

Section asked C-suite respondents about their AI strategies. Eighty-one percent said their company has a formal one. Eighty percent said tools exist with clear access. Seventy-one percent described widespread adoption and open sharing of what works.

Individual contributors answered differently. Only 28% thought their company has a formal AI strategy. Just 32% said tools exist with clear access. And 8%, not a typo, said they've seen widespread adoption and best-practice sharing.

Fifty-three points of difference. Same company, same AI tools, completely different understanding of what's happening.

The emotional split runs deeper. Ninety-four percent of executives trust what AI produces. They use it daily. Only 2% avoid it for work. From the top floor, AI is an obvious success.

On the lower floors, people are struggling. Forty percent of all respondents said they'd be fine never touching AI again. That's not fear of the new. That's a judgment about whether the thing works. And when these numbers finally reach the boardroom, watch for defensiveness. Watch for doubling down. Organizations that bet this big on transformation don't reverse course easily.

When the pitch meets payroll

Cautionary tales have piled up. Klarna told the world in 2024 that AI would handle customer service. Hundreds of outsourced agents, gone. Press releases celebrated. Then the future got messy. Routine-looking customer questions kept needing human judgment. Klarna brought back about a dozen gig workers, quietly, to handle what the machines couldn't.

Duolingo's CEO Luis von Ahn sent a memo: the company would "gradually stop using contractors to do work that AI can handle." A year later, headcount was up 14%. The automation meant to cut labor ended up working alongside more humans. Von Ahn adjusted his message on LinkedIn, saying AI would "accelerate what we do" instead of replace who does it.

Call these failures if you want. I'd call them corrections. Companies learning that AI handles some things brilliantly and other things terribly, with the line between those categories moving around constantly.

The bottom line stays flat

PwC showed up to Davos this week with sobering numbers. They'd surveyed 4,500 CEOs across the globe and asked about AI results. Twelve percent, that's all, said they'd gotten both cost savings and revenue growth from AI.

More than half? No significant financial benefit at all.

Billions invested. Transformation promised. Still not showing up where it counts: margins, efficiency gains, the numbers accountants actually put in reports. Your company keeps announcing AI initiatives while your workload stays the same? That's why.

The training trap

Companies have a default response to AI disappointment: train more people. Section found that 44% of employees have received company AI training. Trained workers do score higher on proficiency tests. But even after training, average scores sit at 40 out of 100. These employees remain stuck in the "experimenter" category, using AI for basics without moving to anything more substantial.

Organizations react to these numbers with frustration, then more of the same. Another module, another workshop. What they haven't faced: the problem isn't prompting. Training programs focus on safety rules and prompt structure when workers actually need help figuring out which parts of their specific jobs AI can improve.

Fifty-four percent of engineers don't use AI to write or debug code. Eighty-seven percent of product managers don't use it for prototypes. The obvious applications for each role sit untouched. People learn the tool and then have nowhere to point it.

Individual contributors left behind

The people who would gain most from AI productivity, individual contributors grinding through repetitive tasks, have the least support for using it. Only 32% have clear tool access, versus 80% of executives. Only 27% have received company training, versus 81% of the C-suite. Only 7% get reimbursed for AI tools, versus 63% at the top.

In most companies, resources go where the work is. AI inverts that. The people doing tasks that AI might actually automate get the least help. The people in strategy meetings get premium access to tools they barely need.

Individual contributors have noticed. Manager support for AI experimentation dropped 11% since May 2025. Only a third get encouragement to try new things. The organizational message, intended or not, comes through clearly: AI is for leadership.

What happens next

Senator Mark Kelly spoke at a workforce event in Washington recently and said it plainly: "People are skeptical."

He argued for broad-based training to build trust. Maybe that works. But the data points somewhere else. Workers have tried AI. They've seen its limits up close. Watched it produce wrong accessibility guidance and broken code and drafts that need complete rewrites.

Their skepticism comes from experience. It's earned.

The companies that solve this will stop counting adoption rates, the number that flatters executives, and start counting time saved per person and actual business results. They'll build use-case libraries specific to each role instead of generic prompt courses. They'll give individual contributors the access that currently flows upward.

The companies that don't will keep announcing AI transformations while workers quietly return to doing things the old way. Leadership stays warm with optimism. Everyone else stays cold with the same workload they had before. And the productivity revolution remains what it's always been: a forecast that reads differently depending on which floor you're standing on.

Related Stories

Why Everyone Uses AI But Nobody Gets More Done

AI usage doubled but productivity flatlined. MIT reveals 95% see zero returns while 'workslop' costs companies $9M annually. Learn why AI fails.

The Implicator

AI Profits Rise While Training Budgets Fall

Enterprise AI delivers 74% positive ROI as companies slash training budgets. Wharton study reveals the extraction model threatening tomorrow's capability.

The Implicator

Big firms hit the brakes on AI

Large company AI adoption drops for first time as Census data shows 14% to 12% decline. MIT finds 95% of enterprise AI pilots fail to deliver ROI.

The Implicator

Harkaram Grewal

New Delhi

Freelance correspondent reporting on the India-U.S.-Europe AI corridor and how AI models, capital, and policy decisions move across borders. Covers enterprise adoption, supply chains, and AI infrastructure deployment. Based in New Delhi.