AI Transformation: What We Learned Moving SupplierGateway to AI-First
Earlier this year, we decided to move SupplierGateway to a fully AI-led operating model. The strategy was clear. What wasn't clear was whether we were prepared to execute it honestly. This is what we learned.
Time was the first thing that broke
Early on, I was asked for a deadline and, without thinking, I said a week. Not because the work needed a week, but because that's the shape of time we're used to giving. Safe. Familiar. Defensible.
I was challenged: Why a week? What actually happens if it's done in a day?
That question wasn't about feasibility. It was about habit. We didn't compress the timeline by working harder. We did it by removing assumptions — the meetings that existed by default and handoffs that existed for reassurance.
Before scaling execution, we needed guardrails. Clear rules for how work enters the system, how decisions are made, and where human judgment genuinely adds value. Because moving fast without governance just shifts risk. Moving fast with it changes how an organisation actually runs.
We kept designing for today
Again and again in planning sessions, we caught ourselves ideating from today's reality. Designing flows, interfaces, and interactions based on how people currently engage with the platform — even while talking about an AI-led future.
Agentic systems don't just execute faster. They change who the primary consumer is, how information is accessed, and when decisions are made. That requires a different design posture — one that assumes AI is not an assistant at the edge, but a first-class participant in how the platform operates.
We had to keep stopping ourselves mid-design and asking: are we solving for today's workflows or tomorrow's? Are we optimising for human navigation, or machine-driven orchestration? Those pauses were uncomfortable, but necessary.
Your tolerance for friction drops fast
A deliverable planned for two days slipped to three and I was genuinely frustrated. That work would have taken months not long ago. And here I was, annoyed it wasn't done in 48 hours.
When you start operating AI-first, small delays stand out. Dependencies surface earlier. You experience time differently. What once felt fast now feels slow. That shift in baseline expectations is one of the clearest signs the transformation is real.
Uneven speed creates real tension
Some parts of the business are moving very quickly while others, for good reasons, are moving more deliberately. Left unchecked, fast areas get frustrated while slower areas feel pressure, and the organisation stops moving forward together.
The temptation is to wait for perfection before pulling the trigger. The problem is that in a transformation like this, waiting for perfect often means waiting too long. We're learning to make bolder calls earlier, even when they feel uncomfortable, and trust the system to absorb iteration rather than delay progress.
That doesn't mean being reckless. It means accepting that alignment and clarity sometimes come after action, not before it.
The team leaned in
When you're fundamentally changing how a business operates, you don't really know how it's going to land. What stood out at our culture workshop was the energy in the room. Yes, there were honest questions. Yes, there's uncertainty. But more than anything, there was genuine excitement about what we're building.
Transformation isn't just about technology or governance. It's about whether people feel part of something evolving — not something being done to them. This isn't something you can impose. It's something you earn through clarity, transparency, and shared direction.
It's still early and we're still learning. But I'm encouraged by what we're building and how we're building it.