The Shift Report · Issue 3 · 4 May 2026
Shadow AI in operations: your team is already using it
Your team is probably already using AI — to rewrite emails, summarise documents, explain error messages, translate instructions. Most of it is invisible. Once AI is interpreting procedures, it has crossed into how work gets done.
For the past year, most companies have talked about AI adoption as if it is still waiting for a project plan. It is not. People are already using it — asking ChatGPT to rewrite emails, summarise documents, explain error messages, translate instructions, draft training material. Some of that is helpful. Some of it is risky. Most of it is invisible.
Lenovo published research that puts a number on this: more than 70% of employees are using AI weekly, and up to one third are using it outside IT oversight. The phrase for this is “shadow AI,” which sounds dramatic, but the reality is simple: people are reaching for tools that help them get through the work faster.
This is not really an IT story. It is an operations story. Because once AI is being used to interpret procedures, draft shift notes, translate safety instructions, or answer process questions, it has crossed into the way work gets done. At that point the question is no longer “should we use AI?” It is “are we giving people a safe and useful way to use it?”
Three things worth knowing
1. Shadow AI is a symptom, not just a risk.
Lenovo’s research says usage is accelerating faster than governance. Employees are using AI with or without formal approval, creating gaps in security, cost control and visibility. The tools people reach for in the moment are rarely the ones IT selected.
My take: when people bypass the official system, it usually means the official system isn’t helping them quickly enough. You can ban that behaviour, but you probably won’t stop it. The better question is: what problem are they trying to solve? That answer tells you more about your real AI needs than any strategy exercise.
2. Frontline workers want answers in the flow of work.
eduMe’s latest AI in L&D report surveyed more than 1,100 frontline workers, managers, training creators and senior leaders. 84% of frontline workers would prefer to self-serve answers using AI-powered support, and 92% want guidance that fits directly into their workflow.
My take: this is exactly why traditional training often fails. The worker doesn’t need another module three weeks before the problem happens. They need a clear answer at the moment the problem appears. That’s a different design brief entirely — and most operations haven’t started thinking about it yet.
3. AI training is becoming basic workforce infrastructure.
The UK Government has expanded free AI training with the aim of reaching 10 million workers by 2030. Skills England has launched an AI and automation practitioner apprenticeship focused on real workplace problems, productivity, safe use and responsible adoption.
My take: this is the signal that AI literacy is becoming like email or basic safety awareness. Not everyone needs to become technical. But everyone will need enough understanding to use the tool without creating new risk — and operations managers will be the ones deciding what that looks like on their sites before any central guidance arrives.
One thing to try this fortnight
Ask three people on your team: “Have you used AI for anything work-related in the last month?” Don’t start with policy. Don’t start with risk. Just listen. What did they use it for? What problem were they trying to solve? Did it save time? Was the output checked by anyone?
That conversation will tell you more about your real AI adoption than any strategy slide. And it’s the only way to find out what’s actually happening before it becomes a problem you find out about differently.
SOPwise turns a Standard Operating Procedure into a training package: task-based steps, critical mistakes, and scenario questions. £35 per SOP, no subscription. Built for UK manufacturing and chemical distribution.