31% of Firms Run AI Agents Now — Most Aren't Ready
Quick answer
MIT Sloan and BCG research shows AI agents hit 35% adoption in two years and 31% of enterprises now run them in production. But most SMEs are skipping the data foundations and governance that make agents actually work — here's what to fix first.
MIT Sloan and BCG research shows AI agents hit 35% adoption in two years and 31% of enterprises now run them in production. But most SMEs are skipping the data foundations and governance that make agents actually work — here's what to fix first.
I've been talking to a lot of SME owners lately who've seen the headlines about AI agents and are proper excited. "Matty, we need to get some of these AI agents sorted." Fair enough. I get it. The hype is real.
But here's the thing — a joint report from MIT Sloan Management Review and Boston Consulting Group, published in late 2025, found that agentic AI reached 35% adoption in just two years. For context, that's a speed that took traditional AI eight years to match. And as of mid-2026, an estimated 31% of enterprises now run at least one AI agent in production, according to S&P Global and McKinsey data.
That's fast. Proper fast. But here's what keeps me up at night — most of the SMEs I work with aren't even close to ready.
What the research actually says
The MIT Sloan/BCG report — "The Emerging Agentic Enterprise" — tracked companies like Microsoft, Chevron, Capital One and SAP. These are organisations with deep pockets, dedicated data teams, and years of cleanup behind them.
The finding that jumped out at me wasn't the adoption number. It was this: the companies succeeding with AI agents aren't the ones buying the best tools. They're the ones who sorted their governance first.
AI is moving from a passive tool to an active participant. You give it a goal, and it figures out the steps to get there.
That sounds brilliant until you realise what it means in practice. If you give an agent a goal and your data's a mess, it'll hit the wrong target at speed. It's like giving a fast car to someone who can't read a map.
Why SMEs are different
The big firms in the MIT Sloan study have something most SMEs don't — clean, governed, well-documented data flowing through systems people actually trust.
I see this all the time. A manufacturing SME comes to me wanting AI automation. I look at their ERP and find three different part numbers for the same component across two sites. Customer records haven't been deduplicated since 2019. The finance team has a spreadsheet that disagrees with the ERP by £40k every month.
You can't layer an AI agent on top of that and expect magic. You'll get confident, fast, automated wrong.
The Cloud Security Alliance has also been sounding the alarm. Their view is that enterprise security now hinges on governing these agents like you'd govern human employees — monitoring what they can access, what they can do, and having a way to pull the plug when they go sideways.
Three things to actually do
If you're an SME owner reading the AI agent headlines and feeling FOMO, slow down. Do these three things first:
- Sort your master data. Get customer records, product codes and supplier details clean and consistent. One version of the truth. This is the foundation everything else sits on — and it's where I spend most of my time with clients.
- Document your workflows. AI agents need clear process definitions. If your purchasing process lives in someone's head and varies by site, no agent can automate it. Map it out first.
- Start with one bounded use case. Don't try to agent-ify the whole business. Pick one repeatable, well-understood process — invoice chasing, report generation, data validation — and prove it works there before expanding.
The companies winning with AI agents right now aren't the ones who moved fastest. They're the ones who laid the groundwork first and then deployed with confidence. There's a reason banking and insurance lead the adoption stats at 47% — they're heavily regulated industries that already have tight data governance and documented processes.
If your data house isn't in order, no AI agent is going to save you. It'll just help you make bad decisions faster. Sort the foundations, then layer the clever stuff on top. That's how you actually win.
References & Further Reading
- MIT Sloan Management Review & BCG, The Emerging Agentic Enterprise
- S&P Global / McKinsey, The State of AI 2025–2026
- Cloud Security Alliance, Enterprise AI Security Starts With AI Agents
- Gartner, via Boston Institute of Analytics, AI Agents Explained: Business Transformation in 2026
Sorting your data out before going AI?
I help UK SMEs get their data house in order before throwing AI at the problem. ERP data, master data, reporting trust — the unglamorous stuff that actually makes the shiny stuff work.
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