Picture the scene. You're at your desk, deadline looming, and you decide to let AI handle the first draft. You type a prompt. The result is wrong. You try again. Still not right. Ten minutes later you're no closer, and the clock is ticking. Eventually you close Copilot and do it the old way, the way that takes longer, but at least you know it works.
If that sounds familiar, you're not alone. And according to Forrester VP and Principal Analyst JP Gownder, it's becoming one of the main productivity problems in the modern workplace.
"I want you to put yourself into the position, I know I have personally, of: I try a prompt and it fails. I try another prompt and it fails," he told UC Today. "At that moment, I have a decision to make. Either I can keep messing around with Copilot with no actual guarantee that I'm going to get it to do what I want, or I can give up and do it the old way. What we're seeing is a lot of abandonment behaviour, because people are either wasting time and never getting an answer, or simply abandoning the tool. And when they abandon the tool, they fall off the learning curve entirely."
Watch the full interview: Why AI Literacy Is Hurting Productivity: Forrester’s JP Gownder
The numbers behind the problem
Gownder's comments come alongside Forrester's second AIQ report. AIQ stands for Artificial Intelligence Quotient, a measure of employee readiness to succeed with AI tools at work. The findings make uncomfortable reading for any organisation that has invested heavily in enterprise AI.
Despite more than 80% of companies having deployed at least some AI tools, just 16% of employees across the US, UK, Germany, France and Australia achieved a high AIQ score in 2025, up from 12% in 2024. Gownder is clear that the pace of progress is nowhere near matching the pace of deployment.
Only 51% of organisations train non-technical staff on generative AI at all. Just 23% teach prompt engineering. And only 37% of employees feel confident adapting to AI-driven ways of working, a figure that has barely shifted year on year. As UC Today has previously reported, nearly half of all AI licences go unused, costing large enterprises an average of $80.6 million annually — and the AIQ data helps explain why.
"For most employees, the cost to that individual of using a tool like Copilot or Gemini is often higher than the time savings they receive on the other end," Gownder explains. "Because they are learning by doing, and that learning is slow, painful, and happening without nearly enough support."
A new problem: AI slop
Beyond the abandonment cycle, Gownder identifies a second productivity drain emerging in workplaces. He calls it AI slop.
