AI adopters lost 9% of focus time. Three advisors ask where the work went
A change consultant in Australia, an e-learning founder in Greece and a behavioral scientist in the US on why faster AI output often just moves the repair work elsewhere.
AI adopters more than doubled their time on email and messaging.
They also lost 9% of their focused time.
Three advisors, from Australia, Greece and the US, think faster output often just moves the repair work elsewhere.
We asked which parts of a job AI has actually taken, and which still need a person. These three answered with a better ledger: count the cleanup, not just the first draft, because a business case that stops at speed is only half a business case.
The task got automated. The work around it didn't.
Mark Winter, Author, Pragmatic Change. Change and transformation consultant at Pragmatic Change (Victor Harbor, Australia)
“The costliest mistake is automating the task and leaving the work around it untouched.”
“An ActivTrak analysis of 164,000 workers across 443 million hours found that people who adopted AI more than doubled their time on email and messaging, lifted business software use 94 per cent, and lost 9 per cent of their focused time.”
Winter is a change consultant, and he brings receipts. Automating a task is the easy half; the reviews, handoffs and follow-ups around it are where the bill turns up.
When output goes up, so do the escalations.
Christopher Pappas, Founder of eLearning Industry Inc (Cholargós, Greece)
“A clear warning sign is when a system produces more output but also triggers more escalations.”
“Savings matter only when the process remains reliable, because employees repairing automated decisions are not reducing work but simply moving it elsewhere.”
Pappas says to document why experienced people make exceptions before automating anything, then measure quality, not just time. His test is blunt: if people are fixing what the system decided, nothing got cheaper.
Compare the whole workflow, not the first draft.
Dr. Gleb Tsipursky, behavioral scientist and author of The Psychology of AI Adoption at Work, Disaster Avoidance Experts, LLC (Columbus, United States)
“A practical test is to compare the full workflow, including correction time and escalation costs, rather than celebrating faster initial output.”
Tsipursky votes for both. AI compiles, a person checks the sources, challenges the assumptions and owns the decision. Cut the person because the draft arrived fast, and the safeguards leave with them.
Where they split
Winter sees a system problem: "We've swapped the motor. We have not redesigned the factory."
Pappas sees a measurement problem. More output plus more escalations means the work moved, not shrank.
Tsipursky sees a design problem. Keep a person where judgment and accountability remain, and keep an eye on the talent pipeline too.
My take
Before the next role goes to an agent, I'd ask one question: who fixes what it gets wrong, and does that time show up on any report? If nobody can answer, the savings are a guess.
Your turn
After AI sped up the output on your team, where did the cleanup land, who does it now, and does anyone count it?
Next: Google Business posts ran fully on AI, but autopilot review replies drifted off-brand, so an SEO agency and an invoicing founder keep a human check on anything a customer sees.
