The Shift Report · Issue 1 · 6 April 2026

Why frontline AI investments fail without training the people using them

Every operations site buys new software — the spend is easy to justify. Training is harder, so when budgets tighten, training gets cut and the software stays. That’s how AI investments quietly run in reverse.

Every operations site I’ve been in over the past few years has bought new software. ERP upgrades, safety platforms, scheduling tools. The spend is easy to justify. There’s a vendor, a demo, an ROI calculation in a slide deck.

Training is harder to justify. It doesn’t come with a demo. The ROI is real but slow and difficult to measure. So when budgets get tight, training gets cut. The software stays.

McKinsey published research earlier this year on why frontline AI investments are failing to deliver. Their conclusion: companies deploying AI without building the human capability to use it are running the investment in reverse. The technology arrives before the people are ready for it. Productivity doesn’t improve. Leadership decides the tool didn’t work. The tool wasn’t the problem.

This is the pattern I keep seeing. AI is moving fast. The gap between what the technology can do and what teams are trained to do with it is widening, not closing. That gap is where you either win or lose.

Three things worth knowing

1. Your SOP is no longer a document. It’s becoming a workflow.

A useful piece from F7i.ai on how SOPs are evolving in manufacturing tracks a real shift: the best-performing operations are moving away from static SOP documents toward “digital guardrails” — interactive, adaptive workflows embedded directly into how work gets done. AI tools can now generate SOPs by observing workers performing tasks, compressing weeks of documentation into hours.

My take: most operations still run on PDFs in a shared folder. The gap between that and a live, AI-assisted workflow is large, but the direction of travel is clear. The question for operations managers isn’t whether this is coming. It’s how far behind you’ll let yourself fall before starting to close the distance.

2. What the top end looks like, and why it matters even if it’s not for you yet.

Voovio builds photographic, interactive replicas of industrial facilities. Operators practise procedures on-screen as if they’re standing in the plant. Their clients — petrochemical companies, refineries, large process industries — can send operators into the simulated plant before they’ve touched the live environment.

My take: I include it because it shows where this ends up. The principle isn’t exclusive to large corporations: train people in a simulated version of their actual environment before they touch the live one. It’s the direction lower-cost tools are moving in. Knowing what the frontier looks like helps you recognise the accessible version when it arrives.

3. KION is training AI in virtual warehouses to handle situations that almost never happen.

KION Group presented at NVIDIA’s GTC conference in March with a specific approach. To prepare autonomous warehouse trucks for rare but dangerous situations, they build digital twins of warehouses and run thousands of simulations without touching live operations. Their first autonomous truck is now operating at a GXO Logistics warehouse in France alongside 200 manual trucks.

My take: the synthetic training data idea is the part that transfers. You don’t need autonomous trucks to ask: what are the rare situations in my operation that we never practise because they don’t happen often enough? Near-misses, edge cases, the scenarios that catch new people out. Most operations have no systematic way to prepare people for those. That’s a solvable problem.

One thing to try this fortnight

Look at the last person who joined your team. How long before they were genuinely confident doing the job without checking with someone? Write that number down.

Now ask yourself: what specifically caused the delay? Not “the learning curve.” The actual thing. Was it a process they struggled to understand? A document that didn’t match what actually happens on the floor? A reluctance to ask questions? That specific friction point is where the opportunity sits.

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.