AI Readiness for SMEs — They'd Already Tried AI. It Changed Nothing. Here's What We Did Instead.

"We've been using AI for months. We've tried a few tools. Nothing's really stuck — and I'm not sure why."

The operations director on the other end had done the right things, on paper. He'd attended the webinars. He'd bought the subscriptions. He'd pointed his team at the technology and said, 'Have a go. ' A handful of people were using it. Most weren't. The ones with different prompts, workflows, and results. Nobody could say with any confidence whether it was making a real difference.

The company — let's call them Meridian Building Services, a 55-person mechanical and electrical contractor working across the South and Midlands — was not short of problems that AI could, in theory, help with. Fourteen active projects are running concurrently.

Three people manage all estimating. Site engineers were buried in report writing, which was eating up 2 or 3 hours per job. A compliance function drowning in paperwork that had to exist, be accurate, and be produced consistently, whether anyone had the capacity for it or not.

The tools existed. The budget was there. The team weren't resistant.

So why wasn't it working?

The Problem With "Have a Go"

This is the point in the conversation where things get frank.

When we asked what specific problem the AI was meant to solve, there was a pause.

"It was supposed to help the team be more productive."

More productive on what, exactly? Measured how? On which tasks? Owned by whom?

Longer pause. Not a defensive one — a thinking one. Because no one had pushed quite that hard before.

This is the pattern we encounter repeatedly in SMEs that have approached AI sincerely and early.

They arrive at the technology before they arrive at the problem. And AI — capable, patient, infinitely accommodating — will happily do things for a team indefinitely without ever doing the right things.

"Have a go" is not a strategy. It is a starting point that only becomes useful if someone is paying careful attention to what the going actually reveals. At Meridian, nobody was. So the tools sat at the edges of the business — occasionally useful, never embedded, never trusted.

The problem wasn't the technology. The problem was three layers upstream of the technology.

What the AI Readiness Assessment Actually Surfaces for SMEs

Before any tool goes into a business we work with, we run an AI Readiness Assessment. It is not a technology audit. It is a business one. It's our structured approach to AI readiness for SMEs — and it almost always reveals the same patterns.

We spend time with the people who do the work — not just the director who brought us in, but the estimators, the site engineers, the compliance leads. We map where time actually goes. We ask what "done" looks like for each task. We look at how data moves through the business. And we look, critically, at whether it moves at all.

At Meridian, three things became visible almost immediately.

• The data wasn't usable. The estimating team was spending close to 40% of its time extracting information from tender packages — hundreds of pages of PDFs, specifications, drawings, appendices. An AI tool could, in principle, summarise and extract from those documents in minutes. But the tender packages were arriving in inconsistent formats, being filed inconsistently, and no one had a clear protocol for what to pull out or why. AI cannot fix that problem. It can only replicate the chaos faster.

  
  

• The engineers had no shared template. Site reports varied significantly in format, depth, and language depending on who wrote them. Clients had raised it twice in the past year — not about the content of the reports but the inconsistency of them. An AI tool can generate a report. It cannot generate a standard that does not yet exist.

  
  

• Accountability was unclear. When we asked who owned each of the three major admin processes — estimating, site reporting, compliance records — the answer was effectively everyone, which meant no one. AI requires a workflow to plug into. In the absence of one, it becomes another ad hoc tool in a world of ad hoc tools, and another layer of noise that the team quietly learns to work around.

  
  

None of these findings were about AI. All of them were blocking AI from working.

The Part Nobody Wants to Hear

Ouch, this is where it gets bumpy. The scenario that sits underneath almost every failed AI pilot we've ever examined is this:

The technology was never the bottleneck.

This is not a comfortable message for a market that has spent three years being sold AI as the answer. But the SMEs that are getting real, durable results from AI adoption are almost universally the ones that did the slow, unglamorous upstream work first. They defined what they were trying to do. They standardised how they did it. They named who was responsible. Then they brought in the tools.

The ones that didn't — that went tools-first and hoped for transformation — are the ones calling us twelve months later, articulate about what didn't land but uncertain about why.

It is not a failure of the team. It is not a failure of ambition. It is a sequencing problem.

What Actually Happened Next

The first four weeks at Meridian contained no AI whatsoever.

We worked with the team to do something that sounds simple and isn't: define the processes as they actually exist, not as they appear on the org chart. The informal handoffs. The workarounds everyone knows about, but nobody has documented. The unspoken knowledge that lives in one estimator's head or in a site engineer's personal system of colour-coded folders.

This is the work that most AI implementations skip. It is also the work that determines whether those implementations take root or quietly die.

By week five, Meridian had three things they hadn't had before: a standardised tender intake protocol, an agreed site report template, and a named process owner for each function. Small changes. Undramatic. Nothing that would appear in a press release.

Week six was when the AI came in.

And it worked. Almost immediately. Because when tender documents arrived consistently formatted and filed, the AI could extract the right information and surface it in a useful way.

Because when the site report template existed, AI could generate a first draft in minutes that matched the standard rather than inventing its own. Because when process owners were named, there was someone with authority to iterate, improve, and hold the team to the discipline over time.

The technology hadn't changed. The readiness had.

What Changed — In Numbers, and in Feel

By the end of three months, Meridian's estimating team had reduced their tender extraction time by around a third. The site engineers were reclaiming four to five hours per week — time that went back into client relationships, technical problem-solving, and the kind of on-site presence that builds long-term confidence.

The compliance lead — who had been the most sceptical voice in the room at the start — became, unprompted, the strongest internal advocate. "I just needed to understand what it was actually for," she said. "When it had a job, it did the job. Before, it just felt like noise."

But the more meaningful change was in the operations director himself.

At the start of the engagement, he described AI as "supposed to" help. By the end, he could say precisely what it did, which parts of the business it had changed, what had produced that change, and what would need to happen for AI to take the next step with them.

That shift — from vague aspiration to operational clarity — is what we are actually offering. Not the tools. The clarity.

What Meridian Teaches Every Other SME

The Meridian story is not unusual. It is, in fact, close to the majority experience among SMEs that have engaged with AI before engaging with the question of why.

A few things that transfer directly to your business, whatever sector you're in:

• AI amplifies what's already there. If your processes are clear, AI makes them faster. If your processes are unclear, AI can amplify the confusion. There is no shortcut past the upstream work.

  
  

• The most resistant person in the room is often the most important. The compliance lead at Meridian was sceptical because she understood the detail of the work better than anyone else in the building. That understanding, once genuinely enrolled in the process, became the engine of a better implementation. Don't manage sceptics into compliance. Listen to what they are actually telling you.

  
  

• Ambiguity about ownership is the silent killer of AI projects. Not resistance. Not capability gaps. Not budget. The absence of someone who owns the process and can hold the standard. Name the owners before the tools go in.

  
  

• The real ROI isn't in the tool. It's in the clarity that the readiness process forces. Meridian didn't just get faster. They got legible — to themselves, to their clients, to their own people. That is a competitive advantage that compounds.

  
  

Where to Start If This Sounds Familiar

If you read the Meridian story and felt quiet recognition — yes, we've got tools, but it hasn't really landed — that recognition is worth taking seriously.

It is almost never a technology problem. It is almost always a readiness problem.

Talisman's AI Readiness Assessment exists precisely for this moment. It is not a sales process. It is a structured diagnostic—a way to understand where your business actually is before making decisions about where AI fits and how. What we find almost always surprises people, and almost always explains what has felt confusing.

The businesses that will use AI well over the next five years are not the ones that adopted fastest. They are the ones that asked the right questions before they bought anything.

One of those questions is simple: are you actually ready?

Take Talisman's AI Readiness Assessment →https://www.talisman-ai.co.uk/ai-readiness-assessment

Talisman works with SMEs at the point where AI ambition meets operational reality — helping business owners and their teams get from "we're trying AI" to "AI is working, and we know why." If the Meridian story sounds familiar, the conversation worth having is probably the next one.