Copilot licences sitting idle. Data platforms that could run agents today, sitting untouched because nobody was empowered to start. A survey published this week found that 59% of companies are spending over $1 million a year on AI, but only 29% are seeing real returns. That gap has a name, and the name isn't technology.
The organisations we work with that are genuinely ahead have one thing in common. They started before they were ready. Imperfectly, with limited resources, learning as they went. The ones still waiting for the right moment, the right framework, the right sign-off are falling behind people who were never more qualified than them- they just started.
The dust is not settling
There is a version of this story that ends questionably for procurement. Leaders see the pace of AI development and decide to wait. Wait for the technology to mature. Wait for best practice to emerge. Wait until things stabilise.
The problem is that the pace is not slowing. Agentic AI- systems that can reason, plan, and execute tasks autonomously- has moved from experiment to production deployment across industries faster than most enterprise technology roadmaps anticipated. What was a niche concept eighteen months ago is now running in financial services, logistics, and supply chain functions at serious scale.
Waiting for the landscape to settle is not a neutral position. It is a decision to let the gap widen.
The adoption gap is real, and it is widening
The organisations struggling are not short of investment. They have the licences, the platforms, the vendor relationships. What they are short of is momentum.
We see this consistently. Teams using Copilot as a slightly better search engine. Procurement functions sitting on data infrastructure that agents could be working on today, where nothing has been built because no-one has been given the authority to start. The investment is there. The capability is there. The activation is missing.
The new Writer enterprise AI survey captures this precisely: 75% of executives admit their AI strategy is more for appearance than actual guidance. That is not a technology problem. That is a leadership one.
The barrier to entry has never been lower
What has changed fundamentally over the past two years is access. A procurement team with the right skills and curiosity can build agents, automate workflows, and redesign processes without waiting for IT sign-off, without a seven-figure implementation, without a three-year roadmap. That was not true in 2021. It is emphatically true now.
This is one of the most significant shifts in enterprise technology in a generation. The limiting factor used to be budget or vendor capability. Neither is the constraint anymore.
The constraint is mindset. Specifically, whether leaders are willing to give their teams permission to build, and whether those teams believe it is their job to try.
What the leaders who are ahead are doing differently
They are not doing AI perfectly. Nobody is. What they are doing is treating AI as something their function owns, not something that happens to them.
They are identifying one workflow- supplier onboarding, spend analysis, contract review, market intelligence- and asking what it looks like if AI handles the parts AI is good at. Then they build it, learn from it, and move to the next one. They are not waiting for a centre of excellence to hand them a blueprint. They are becoming the blueprint.
The organisations achieving real returns share a pattern: they connect AI directly to outcomes, they assign ownership, and they give the people closest to the work the authority to act. That last part is the one most procurement functions have not figured out yet.
The question that matters
The leaders who will define what procurement looks like in five years are asking a simple question right now: if AI handles everything it is capable of, what do my people actually do?
That is not a technology question. It is a question about what a procurement function is for, what value it genuinely creates, and what human judgment is actually worth in that context. The functions that answer it well- and act on the answer- will look completely different from the ones that did not. The functions that wait for someone else to answer it first will spend years catching up.
The gap between those two groups is opening now. The time to close it is also now.
© Hudson & Hayes | Privacy policy
Website by Polar
