Go slow to go faster is an odd thing to say at a conference about the Age of Acceleration. But the more I reflect on our experience at SEGRO, the more useful it feels.
Next week I’m speaking at CMS’s annual conference, building on our shared conversations around digital and innovation. Preparing for it has brought me back to the messy middle of transformation: the gap between introducing something new and people making it useful in their own work.
A plan can make that journey look wonderfully tidy. Clear ambition. Promising technology. Then comes the less tidy business of changing how the organisation actually works.
At SEGRO, we spent time building curiosity about digital, confidence around change and excitement about the art of the possible. Alongside that came data skills, data management, governance and agreed principles for AI use. We then gave colleagues tools to look at their own work differently, and they went after it. Faster and faster.
Looking back, that preparation was not the pause before transformation. It was part of the transformation. AI makes the lesson particularly relevant now, but it did not invent it.
Give people something to move towards
We wanted colleagues to see themselves in the change, even in charge of it. Not simply to understand individual tools, but to become curious about what they might make possible. It is hard to build momentum around something that still feels as though it belongs to somebody else.
Our investment in learning helps bring that preparation to life: 636 learning hours, 35 sessions and 188 colleagues, alongside 89% adoption within the pilot. They offer a snapshot of the investment and engagement, rather than the whole story; the detail of how learning and adoption are measured matters too. For me, they make the commitment tangible. We were not simply making a tool available and hoping people would find their way.
Value gets stuck in translation when we explain what technology can do without connecting it to people’s working day. AI brings this into sharp focus, but the question is familiar: what could this help me do differently on Monday morning? Perhaps that is becoming a bigger part of the digital professional’s job: less explaining the tool, more helping people find a reason to use it.
Give the enthusiasm solid ground
There was another part of getting ready at SEGRO that deserves equal attention: data. Building data skills, strengthening data management and governance, and agreeing our principles for AI use were instrumental in opening up the opportunity. These were not housekeeping jobs while we waited for the interesting bit. They were part of making the interesting bit possible.
Data readiness starts with ordinary questions. What does this information tell us? Is it accurate and current enough for the task? What is missing? Data skills help colleagues challenge an answer rather than assume that because AI has produced something convincing, the information behind it must be sound. We do not need everybody to become a data specialist. We do need people to recognise when to trust, when to question and when to ask for help.
Data management and governance put responsibility around that information. Who looks after it? How is it kept useful? Who should have access, and for what purpose? Agreed principles for AI use help colleagues decide how to proceed. A tool making something possible does not automatically make it appropriate.
None of that means waiting for perfect data. It means understanding what is good enough for the intended use and what needs attention first. That is a more useful conversation than treating data as either ready or not ready, with nothing in between.
Then put the tools in their hands
With those foundations taking shape, we gave colleagues tools to look at what they do and consider where they could make a difference. Where is the friction? What takes longer than it should? What could be simpler? The opportunity became something they could shape, rather than something they were waiting to receive.
Colleagues applied the capability to understanding complex documents, developing content and roadmaps, challenging proposals and building prototypes around real constraints. We did not have to predict every use case upfront. Real work revealed possibilities we had not anticipated. The pace came from people applying the tools to work they understood, not simply adopting somebody else’s answer.
That brings the process into view too. Not just, can we do this task faster? But, does it need to happen this way at all? Automating a task may save effort. Asking why it exists can change the opportunity altogether. There is value in taking that pause before making the existing version of the work run more quickly.
And when somebody hesitates, we need to listen. Excitement does not remove every reasonable concern, and another demonstration is not always the answer. People need to understand where their judgement fits, what they remain responsible for and what support they can draw on.
We did not need every answer before colleagues began. We needed enough confidence, skills and direction to start, then room to learn from what happened. That is the distinction I want to hold on to: learn first, then scale. Not learn everything first. Learn enough through use to make a better decision about what comes next.
From a colleague’s idea to something we can share
At SEGRO, we talk about Little AI and Big AI. Little AI starts with a colleague’s own idea, often using their own data to make something easier for themselves or their team. Little describes where it starts, not how much it matters. It gives colleagues a practical way to explore a possibility close to work they understand.
Big AI works across the business. It brings together repeatable capability, AI agents that carry out tasks, and data from across SEGRO. The question moves from ‘How could this help me or my team?’ to ‘How could this help colleagues across SEGRO?’
Our AI Centre of Excellence provides the route between the two. When a colleague’s idea has wider potential, it gives us a way to explore what would make it useful beyond its original setting. Something can work well for one team and still need attention before others can depend on it. The task is to help it travel without losing the practical insight that started it.
The agentic library makes shared capability tangible. We have 24 live agents, nine public and 15 private, with access depending on the agent, and have recorded 25,000+ agent conversations. Across our wider AI activity, we have recorded 300+ ideas. Colleagues have also submitted 500,000+ prompts in the preceding 12 months. The 12-month timeframe applies to prompts; the ideas and conversations are separate activity totals. Behind those numbers are familiar tasks: improving a prompt, challenging a proposal or making writing clearer. They give us a sense of the curiosity at work. The next question is what difference it makes.
Not every Little AI idea needs to become Big AI. Making one colleague’s working day easier is worthwhile in its own right. We might adapt an idea, take it further or decide it should stay where it is. Learning has to be allowed to change the plan, otherwise we are only looking for confirmation.
Going bigger is not just adding more users, either. Is the data understood in the same way by another team? Is it kept up to date? Should everyone have access? This is where those earlier investments in data and principles are tested in practice. The Centre of Excellence gives us a route to work through the questions, rather than leaving each team to solve them alone.
Connect the work, not just the tools
Sharing useful capability is one thing. Connecting it into the way the business operates is another. The broader opportunity is not simply more agents doing more tasks. It is connecting people, processes, data and agents so that useful work flows between them. That is the value of orchestration: connecting the operation, rather than adding another tool alongside it.
SAMA, our SEGRO Asset Management Application, gave us a practical lesson in this. Bringing portfolio activity and performance into one connected view depended on connecting previously fragmented data, processes and people. The starting point was the decisions asset management colleagues needed to make, not the technology we wanted to introduce. That experience showed us what wider orchestration asks of the data, processes and people around an application. The application matters, but so does what building it taught us about the organisation around it.
This is where the messy middle becomes particularly visible. An individual task can work beautifully while the hand-off to the next person or process still needs attention. Has the right information arrived? Is somebody clear about what happens next? Where does human judgement come in? A faster first step is not much help if the work then waits somewhere else.
Learning before we scale therefore means looking at the connections, not just the performance of each tool. It brings us back to why colleagues’ understanding of the work matters so much. The technology may complete a task. People can help us see whether the work as a whole has become better.
Prove the difference in the work
Sustainability reporting makes the point rather more concrete. One of our earliest substantial use cases covered 11 markets and 35 different data types. Information was being translated, extracted, re-keyed and consolidated manually through spreadsheets. The difficulty was not a shortage of information. It was the effort required to make it usable.
Before applying AI, we looked at where the information came from, where effort was duplicated and where expert judgement was needed. That opened up a more controlled approach: AI supporting the combining, cleaning and validation of data, with people focusing on exceptions, assurance and judgement. Our proof of value showed the potential to move the process from one month to one hour. That gives us something worth pursuing. It is not yet a benefit delivered at scale; the test of that potential is how it holds up in routine use, including how the timings compare. For me, this is learning before we scale in practice: an encouraging result that informs the decision about whether and how to go further.
Our internal 2026 figures put time saved at 27 minutes per user per day, with 67% reporting improved quality. What interests me is seeing time and quality together. A quicker answer is useful only if it helps the work. These figures offer a useful snapshot, with more to understand about the group measured, the timeframe and how the time savings were calculated. I see them as a starting point for understanding the difference AI makes, rather than a promise for every colleague or a cash-saving calculation. As capability grows, the level of autonomy should fit the task, its risk and its maturity. More AI does not automatically mean less human judgement.
Go slow to go faster
I would love to say we had planned the whole sequence at SEGRO. We hadn’t. Some of it was lucky happenstance, but it worked, and we are now seeing the benefits. Looking back helps me recognise which lessons I would carry into another transformation.
It would be easy to turn the experience into a tidy formula after the event. That would rather miss the point of writing about the messy middle. We did not find a perfect sequence for transformation. We learnt that time spent on people’s readiness and the data behind their work could help them move with greater confidence. Some of the work towards pace does not look much like pace while you are doing it.
There is a responsibility for those of us leading this work too. We cannot encourage curiosity and then behave as though we already have every answer. Knowing a little about a lot helps: enough about AI, data, risk and adoption to ask useful questions, and enough humility to listen to the people applying it.
The encouraging part of SEGRO’s story is not just what we put in place. It is what colleagues chose to do with it. That is the bit I want to keep noticing as the technology changes and the possibilities grow.
Perhaps the place to start is not with another question about what AI can do. Ask a colleague what gets in the way of their work. Explore what could be different. Look together at the information they would need and what would give them the confidence to try.
That might stay a Little AI idea. It might become something others can use too. Either way, it starts with somebody seeing a possibility in their own working day. Perhaps that is how we find our way through the messy middle: not by making the plan look tidier, but by helping people take the next step.
Perhaps the Age of Acceleration makes that preparation more important, not less. Go slow enough to get ready, then give people the confidence and tools to move. What encourages me most is the confidence colleagues have built to ask better questions and shape what happens next.