62% of Firms Are Stuck in AI Pilot Purgatory
Quick answer
Most UK SMEs have AI somewhere but very few have moved agents into production. This article explains why the gap exists, what it costs, and the three practical steps to get past pilot stage and into real business value.
Most UK SMEs have AI somewhere but very few have moved agents into production. This article explains why the gap exists, what it costs, and the three practical steps to get past pilot stage and into real business value.
I see this all the time with UK SMEs. Somebody goes to a conference, sees a demo of an AI agent handling invoices or writing reports, comes back fired up. They buy the tool. They run a pilot. It works fine in the demo environment. And then... nothing. Six months later the pilot is still a pilot. The agent is doing one narrow thing for three people, and the board is asking what they actually got for their money.
Sound familiar?
The numbers are brutal
McKinsey's latest State of AI survey found that 88% of organisations now use AI in at least one business function, up from 78% the year before. PwC found that 79% of senior executives say AI agents are already being adopted in their companies.
But here's the kicker. McKinsey also found that only 23% of organisations are actually scaling an agentic AI system in a real business function. Another 39% are experimenting. That means roughly 62% of firms using AI are stuck somewhere between "we've got a pilot" and "this is actually part of how we work."
They're not failing. They're just not arriving anywhere.
Why pilots don't become production
In my experience, and I've seen this dozens of times, the reasons are nearly always the same:
- The data underneath is a mess. The pilot worked because someone manually cleaned a dataset. In production, the agent pulls from live systems with duplicates, missing fields and inconsistent formats. It falls over.
- Nobody owns the workflow. The pilot was run by an enthusiast. Production needs someone accountable for the process the agent is touching. If that person doesn't exist, the agent becomes orphaned tech.
- There's no governance. McKinsey's trust research found only 30% of firms have proper AI governance for agentic systems. Without clear rules on who can change what, what happens when the agent gets it wrong, and where the guardrails are, nobody wants to switch it on for real.
It's not a technology problem. The tech works. It's an operational readiness problem.
What to actually do about it
If you're stuck in pilot purgatory, or about to start an AI agent project, here are three things that actually move the needle:
1. Fix the data the agent will touch. Before you even think about production, audit the data sources. Customer records, product data, document stores, whatever the agent reads from. Get it clean enough to trust. Not perfect. Good enough. 80% is fine. 40% with a prayer is not.
2. Pick one workflow and own it end-to-end. Don't try to deploy five agents at once. Pick one painful, repetitive process. Invoice processing, report generation, customer triage, whatever. Assign a real owner. Define what success looks like. Define what happens when the agent gets it wrong. Run it for 90 days, measure it, then decide whether to expand.
3. Put basic governance in place before you switch it on. That means: who has access, what data can the agent see, what's the escalation path when something looks wrong, and how often do you review it. A one-page policy. Not a 40-page framework. Just enough that everyone knows the rules.
The bottom line
The firms getting value from AI agents aren't the ones with the best technology. They're the ones who did the boring groundwork first. Clean data, clear ownership, simple governance. That's the stuff that turns a pilot into production.
If your AI agent is still living in a sandbox after six months, it's probably not the agent that needs fixing. It's everything underneath it.
References & Further Reading
- McKinsey, The State of AI (Latest Global Survey)
- PwC, AI Agent Survey of Senior Executives
- Maven AGI, AI Agent Adoption Statistics for 2026
- Anthropic, Building Effective AI Agents
Stuck in pilot purgatory?
I help UK SMEs get their data and workflows ready for AI agents to actually work in production. Not just demos. Real business value.
Let's have a chatMatty Hatton is the founder of Digital Adaption, an ERP and data consultancy based on the Wirral. He has spent 15 years delivering ERP transformations for manufacturers, including leading the data migration on a £4.5m consolidation of four legacy systems onto a single Infor LN cloud instance for a 220-user group. He holds an MSc in Digital Transformation and IT Strategy from Manchester Metropolitan University and is Microsoft PL-200 certified.
Start with a 30-minute data risk call