First the data, then the agent
Jonny Winn is inside businesses right now, helping them get ready for AI. What he is seeing is not a tools problem.
Jonny Winn is currently working with businesses, guiding them through the process of preparing for AI adoption. I was interested to discover that the challenges he sees organisations facing are ‘data related’.
‘No surprise there’ you might say - after all he is a data strategist!
But read on because Jonny Winn’s job is to help customers solve the data challenge and introduce AI agents into their businesses so they can reap the rewards.
There’s a video walkthrough accompanying this article so if you want to get hands on - you’ll find the link below.
Meeting Businesses Where They Are
Business leaders want AI. They have been reading about it, hearing about it in the boardroom, watching competitors talk about it at conferences. They want an agent. They want automation. They want the efficiency gains they have been promised.
Jonny Winn is meeting them where they are. As AI & Data Strategy Consultant at The One Point, a Yorkshire-based managed service provider working with SMBs, enterprise, and public sector organisations, he is the person who helps them understand what they can build, and when.
And what he is finding, consistently, is that the queue is forming in the wrong place.
“AI is the shiny, shiny tool,” he says.
“What is more important, is the data foundations.”
The cupboard nobody wants to open
There is a moment in almost every AI readiness conversation, Jonny says, when the energy changes. It happens when you stop talking about what AI could do for the business and start talking about what the business’s data actually looks like.
“As soon as you start to talk about ‘you might not be quite there, we need to look at your data,’ that’s when people start to get nervous.”
The nervousness is understandable. Data is the thing nobody has properly sorted. It is in spreadsheets on people’s desktops. It’s in inconsistent formats across a dozen different documents. It’s in third-party tools that don’t talk cleanly to the Microsoft suite. It’s, as I put it to Jonny, ‘the scary beast in the cupboard’.
He didn’t disagree. But he was careful not to make it doom and gloom either. Because the point is not to scare anyone away from the work. The point is to do it in the right order.
“Without the data foundations, when you progress onto the AI side of things, it’s just going to become expensive experimentation.”
That phrase landed for me. ‘Expensive experimentation.’ Not transformation. Not efficiency. Just money spent on something that cannot perform because it has nothing solid to stand on.
It reminded me that my stepson bought us a robot vacuum for Christmas a couple of years ago. I was so looking forward to using it. I spent a long time setting it up and building the map of the house - but our connectivity wasn’t strong enough to support it.
After many fruitless trial runs watching it waddle half way down our long hall, halt with a confused wriggle - turn right and start banging it’s head relentlessly on the dresser instead of turning left into the lounge. I admitted defeat and gave it away.
The many hours invested in setting it up were lost. The vacuum technology was reputationally, excellent. But the infrastructure just wasn’t there to support it.
Jonny is describing a similar experience at a business scale - except the price tag is considerably higher, and it is a lot harder to give away!
What readiness actually means
Before any tool is discussed, Jonny runs a readiness assessment. A conversation about infrastructure, data governance, existing systems, what is clean and what is not. He works with clients to build a roadmap. Explaining to them where they are, what needs to change, and in what order. A client might be amber: nearly there, with specific things to tweak before moving forward.
This is unglamorous work. It doesn’t make headlines. It doesn’t get presented at conferences as a breakthrough. But it is the work that makes everything else possible.
(Fast forward one year - It wasn’t glamorous work installing the superfast broadband infrastructure we needed for a robot vacuum, but it made a big difference to the tech’s performance and the new one we purchased actually works now!)
The pattern Jonny is observing in SMBs is similar to that my week 04 guest, Kate Madison-Greenwell, spoke of. She said it wasn’t unusual for organisations of one-thousand to ten-thousand people to find that their data isn’t ready. The infrastructure isn’t aligned. And the pressure from above to just get on with it is real.
The smartest organisations, Jonny says, are the ones that invest in sorting the foundations now. They may not look like they are moving as fast as the ones rushing to implement. But when things break - which they are very likely to do if the work was skipped - those who sorted the foundations first will already be ahead.
Then the agent
Here is what is waiting on the other side of doing it right.
In Jonny’s view, agentic AI is the biggest opportunity available to businesses right now. And the thing that makes it genuinely exciting is that it does not require a technical team to build.
“It’s no code, really,” he says. “Which is fantastic for people who aren’t from a technical background.”
He demonstrated this with a mock company he created especially for the demo using Claude: Acme Solutions Limited. He built a customer onboarding agent connected to a knowledge base (a simple PDF containing the company’s onboarding process). The agent knows its role. It knows the tone and voice it should use. And it knows that when a question falls outside its knowledge it must triage it to a human, cleanly, without making anything up.
That last point matters more than it might seem. A well-built agent with a clear knowledge boundary is not just useful. It’s trustworthy. And trustworthiness is what allows a business to deploy it with confidence.
Most businesses are already paying for the Microsoft enterprise suite. The ability to build agents is already inside it - through Copilot, or at a more advanced level through Copilot Studio.
The tools are there. The barrier is not access to technology. It is having the data foundations in place to make that technology work.
A different kind of expert
Jonny came to this work via fifteen years in creative production, a business degree, and a mid-career realisation at forty that the thread running through everything he had done was data. Not AI, but Data. He was able to tell stories with the data and was intrigued the way it changed when it was translated for different audiences dependent on the context.
He went back to do an MSc in AI and data science at Hull University, graduated recently with a distinction, and now he does the work he has longed to do in the place he and his family call home.
He calls himself a curious experimenter and places himself at six out of ten on the AI spectrum. As you can imagine, I pushed back on this because he has a deeper technical understanding than most people advising organisations on this subject. But his point was sound. Claiming nine or ten means claiming expertise at the rocket scientist level and in this field, that’s exactly the kind of overclaiming that gets organisations into trouble. Wasn’t it Mark Twain who said something elegant which translates to ‘the more you learn the more you realise how little you know.’
What made this conversation with Jonny so useful was never going to be that he was a lofty 10 on the AI spectrum, it was much more about the combination of him knowing enough about the technical foundation combined with his willingness to stay honest about what he doesn’t yet know. That builds trust.
The Jobscaping® thread
Jonny’s vision of the ideal AI-driven society is one where AI handles enough of the administrative weight of work and life that people get time back for what actually matters. For him personally, that means family. For him professionally it means giving his clients the time back to do the job they are capable of doing, rather than drowning in the admin and the busy work.
This is the Zone One and Zone Two work of Jobscaping®. These are the tasks that can be delegated or handled with AI, cleared away so that Zone Three remains fully resourced. The relationships, the judgement, the decisions that require a human who has had the space and the conditions to think ‘clearly’.
But none of that is accessible if the data foundation is not there. The agent cannot perform the busy work for you, if it has nothing solid to draw from. The time cannot be reclaimed if the infrastructure to reclaim it has not been built.
First the data. Then the agent. Then the time.
This week’s resource
Jonny talks enthusiastically about the opportunities offered by AI and has produced a video walkthrough showing exactly how to build a Microsoft Copilot agent from scratch. It uses the Acme Solutions example from this article and it shows you the version with and without a knowledge base so you can see the difference a connected knowledge base makes.
QED.
Take care for now and stay Sovereign.
Watch Jonny’s video walkthrough of building an AI agent for practical guidance.
Take care for now and stay Sovereign!
Thanks Carolyn. The robot vacuum analogy is the one I'll be borrowing when I'm next in front of a board, with full credit.
It does something my version doesn't. It moves the data conversation out of homework territory and into something everyone has lived: a piece of kit that should have worked, sitting in a cupboard. Same lesson, less defensiveness.
The Jobscaping connection at the end is the part I keep coming back to. Clean data is what makes the Zone 3 time genuinely reclaimable. Without it, the agent runs and the hours stay missing.
Looking forward to the next one.