ChatGPT and Gemini may be slowing you down: 3 signs AI is hurting productivity

New research on OpenAI’s ChatGPT and Google’s Gemini, cited in McKinsey & Company’s Technology Trends Outlook 2026, points to an awkward possibility: AI may be costing you productivity, not adding. A lot of the time saved on paper gets burned up in checking outputs, fixing errors, and rewriting what the tool gave you.

McKinsey’s Technology Trends Outlook 2026 found that productivity actually dropped in 30% of companies after they rolled out agentic AI tools. The slide shows up most clearly when people use chatbots to produce first drafts of emails, reports, summaries, and presentations. McKinsey’s point is pretty simple. These systems can help with wording, structure, and framing, but when they generate a full draft, cleaning it up can take longer than just writing the thing yourself.

That oversight has its own name now, “botsitting,” and McKinsey says it averages 6.4 hours a week, or 37% of total AI time. Nearly half of that goes to correcting mistakes and re-prompting. Only 35% counts as immediately productive. While 76% of users save at least an hour a week, just 26% put that time into more work. And even though 80% say they feel faster personally, only 37% of companies report EBIT gains. McKinsey also says 45% spend an hour or more fixing output generated by someone else.

If you lean on OpenAI’s ChatGPT or Google’s Gemini, that warning shouldn’t be brushed aside. The Massachusetts Institute of Technology, MIT, describes a “productivity J-curve,” where output gets worse before it gets better. But that improvement only shows up when companies put real effort into training, review, task selection, and cleaner data. Without that, you’re not streamlining work. You’re adding admin.

You can use OpenAI’s ChatGPT and Google’s Gemini on the web and on mobile. But if they leave you spending more time editing than writing, the productivity gain has already started to flip the other way.