By Kapil Raval · June 2026 · 5 min read
The promise was productivity. For too many organizations, the reality has been rework. Is it a lack of imagination — or simply a fear of missing out? Here is what the data says, and what better leadership looks like.
There is a phrase that has stayed with me.
Demis Hassabis, CEO of Google DeepMind, recently said that using AI primarily to cut jobs reflects a “lack of imagination.”
He is right. And the data is beginning to prove it.
When AI arrived at scale, many leaders faced immediate pressure: justify the investment, show results fast.
Cost reduction was the obvious answer. Fewer headcount. Leaner teams. Faster outputs with less overhead.
It made sense on paper. In some situations, it still does.
But for many organizations, this framing led to a consequential mistake. AI was deployed not to build more — but simply to replace.
That is where things started to go wrong.
The first consequence has been well documented, though not always honestly acknowledged.
Early-career professionals have been disproportionately affected. Entry-level roles that once served as the on-ramp to organizational knowledge — and to the next generation of capable leaders — have been quietly eliminated.
This is not just a social concern. It is a strategic one.
The talent pipeline matters. Institutional knowledge, built over years, matters. When we remove the foundation, we do not always notice the cracks right away.
We will.
Here is where it gets more concrete for business leaders.
2026 was supposed to be the year of the AI Dividend. Productivity gains would finally hit the bottom line. AI would move from promise to proof.
For a segment of organizations, that is exactly what is happening. Anthropic recently built its own productivity tool, Cowork, in just ten days — using AI-assisted development. That is the real promise: expert teams building at a pace that was previously impossible.
But that is the lab reality.
The enterprise reality is different.
Workday reports that 40% of the time saved by AI is being lost to rework — correcting errors, fixing inaccurate outputs, and editing generic content that misses context. Forrester finds that only 15% of companies can actually link their AI investment to a measurable profit increase. Gartner’s data tells a similar story.
The AI Dividend, for many, has quietly become an AI Rework Tax.
The difference between the lab and the enterprise is not the technology. The technology is largely the same.
The difference is preparation.
Teams that use AI well have done the structural work first. They understand what AI can do reliably — and where it still needs human judgment. They have trained their people. They have redesigned workflows, not just bolted AI onto existing ones.
Most enterprises have not done this yet.
Consider autonomous AI agents in e-commerce. Billions have been invested. The technology can locate a product in milliseconds. Yet large-scale deployments remain elusive. Why? Because an agent operating in messy, real-world data — without the nuance a human naturally applies — still generates more risk than reward per transaction.
Fast in the wrong direction still costs you.
Some executives moved quickly. They saw the headlines about AI speed and capability. They reduced headcount before structural changes were in place. They assumed AI would fill the gap.
In many cases, it did not.
What they got instead was a more burdened remaining team — now responsible for verifying, correcting, and managing AI output on top of their existing workload. The skilled people who once caught problems early, who carried institutional knowledge, who mentored junior colleagues — they were gone.
The Rework Tax fell on those who stayed.
This is not a criticism of AI. It is a reflection of how difficult organizational change actually is — and how rarely it gets the investment it deserves.
So what should leaders be measuring?
Not headcount reduction alone. Not raw speed. These are incomplete indicators.
The more useful question is: what is the net value created?
That means accounting for time saved, minus time lost to rework. It means measuring output quality, not just output volume. It means asking whether decisions made with AI support are actually better — not just faster.
The companies winning with AI in 2026 are not those with the most tools or the most agents. They are those who have closed the gap between AI capability and human readiness.
Their people are not janitors of AI output. They are architects — setting direction, applying judgment, and using AI to execute at scale.
Forrester recently noted that AI is trading its tiara for a hard hat.
The hype cycle is softening. The real work of making AI operational has begun.
That is actually encouraging. It means the conversation can shift from what AI might do to what your organization is genuinely prepared to do with it.
That shift requires leaders willing to invest in people — in training, in redesigning roles, in rethinking how value gets created. It requires a longer view than the next quarterly result.
It requires imagination.
Not the imagination of science fiction. The practical kind: imagining what becomes possible when your most capable people are freed from repetitive work and empowered to solve harder problems.
That is where the real AI Dividend lives.