Two numbers published this month should stop every chief procurement officer mid-budget cycle.
Sixty-three percent of senior procurement leaders expect AI to significantly improve procurement performance over the next three years. Only 25% have undertaken job and role redesign as a direct result of AI implementation.
The figures come from two Gartner surveys covering 213 senior procurement leaders and 134 procurement respondents within a broader C-suite study. They describe a function that believes in the technology and has not yet touched the organization chart.
That gap has a name. Job chaos. It happens when AI hollows out standardized, middle-skilled work while responsibilities, workflows and required skills shift underneath people who were never told the shift was coming.
Where the pressure lands first
Picture your category management team. Somewhere in the middle of that structure sits a cluster of roles built around standardized output. Running the RFQ process. Normalizing supplier responses into a comparison grid. Updating the spend cube. Chasing contract renewal dates. Preparing the quarterly business review deck.
None of that work is trivial. All of it is structured, repetitive and well documented, which makes it exactly what AI absorbs first.
The demand curve is shifting in two directions at once. Toward lower-skilled roles that are heavily AI-augmented, where a less experienced person supported by good tooling produces what used to require years of category familiarity. And toward higher-skilled positions requiring creativity and complex problem solving, the work of negotiating a difficult supplier relationship or restructuring a category strategy under tariff pressure.
The middle is where the squeeze happens. And the middle is where most procurement organizations have concentrated their headcount for twenty years.
What redesign actually means
Fareen Mehrzai, Senior Director Analyst in the research firm’s supply chain practice, frames the core issue plainly. Successful AI implementation at the functional level is fundamentally a role redesign challenge. Responsibilities and skills across procurement are already changing, and CPOs who delay redesigning roles risk leaving employees without a clear path forward and their organizations without the talent needed to capture the full value from AI.
The practical starting point is an audit conducted with employees, not about them. Take each role and separate the responsibilities that should stay with people from those that should move to AI.
Run this on a sourcing manager and the split becomes obvious quickly. Drafting the initial RFQ document, moves. Deciding which suppliers to invite, stays. Consolidating bid responses, moves. Reading the room when a strategic supplier signals capacity constraints, stays. Monitoring contract compliance against agreed terms, moves. Deciding whether to enforce a penalty clause with a supplier you need next quarter, stays.
Michelin offers a useful proof point from the adjacent planning function. For its automotive division, roughly 50% of forecast volumes now run without human intervention, which freed planners to concentrate on new product launches and the references the model consistently handles badly. The organization did not shrink its planning team. It moved people up the value chain and documented what they were now responsible for.
That documentation step is the one most procurement functions will skip, and it is the one that prevents chaos.
The skills that appreciate
As routine work automates, the premium shifts to capabilities AI cannot readily replicate. Creativity. Influence. Agility. Ethical judgment.
These sound soft until you attach them to procurement situations. Influence is what gets an internal stakeholder to accept a supplier they did not choose. Ethical judgment is what stops a cost-saving sourcing decision that would expose the company to forced labor risk in a tier-three supplier. Agility is what reorganizes a category strategy in six weeks when a tariff announcement invalidates the assumptions underneath it.
The uncomfortable truth is that these capabilities erode when they go unused, and many procurement professionals have spent years in roles that exercised process discipline rather than judgment. Assessing them honestly, then building targeted development before the erosion sets in, is a different exercise from sending the team on a generative AI training course.
Experimentation as operating model
The third move is structural rather than individual. Create a workforce sandbox, a defined space where employees can test AI tools and workflows safely, learn continuously, and treat failed experiments as part of development rather than as mistakes on a performance record.
This works only with transparency about the AI agenda. If your team suspects the real objective is a headcount reduction nobody has announced, no sandbox will produce honest experimentation. People do not volunteer the automation of their own job to a leader they do not trust.
Communities of practice and peer networks accelerate the learning considerably, because the useful discoveries in procurement AI are specific. The prompt that reliably extracts obligations from a contract in your template format. The workflow that handles supplier onboarding exceptions without breaking. These travel between colleagues faster than they travel through formal training.
Mehrzai is clear that this is not a project with an end date. CPOs must give employees agency in shaping how their roles evolve, and role redesign cannot be a one-time exercise. It needs continuous adjustment as AI advances and as employees learn what actually works.
The question has changed
Most procurement leadership teams are still asking how they can use AI. That question produces a list of use cases, a pilot, and a slow fade.
The more useful question is how procurement should be redesigned to create more value in an AI-enabled world. That question produces an organization design, a skills plan and a conversation with HR.
It is a significantly larger conversation. It is also the one that determines whether the 63% who expect performance gains actually get them.
Two takeaways
First, run the responsibility audit this quarter, with your team in the room. Take your three largest role types and split every responsibility into stays with people or moves to AI. The exercise takes two workshops. It gives every person in the function a documented view of where their job is heading, which is the single most effective defense against the morale collapse that follows uncertainty. Doing it with employees rather than to them also surfaces the undocumented exceptions that would have broken your automation anyway.
Second, rewrite your hiring profile before your next requisition. If you are still screening primarily for ERP fluency and process execution, you are recruiting into the part of the function that is contracting. Screen instead for negotiation under ambiguity, stakeholder influence and the ability to reason about a supplier decision with incomplete information. Those capabilities take years to build and will be scarce precisely when everyone wants them.
The technology question in procurement is largely settled. The organization question has barely started.
Which responsibilities in your team would you move to AI tomorrow, and which would you refuse to move at any price? And has anyone in your function been told honestly what their role is expected to look like in three years?
Continue the discussion on Chain.NET, the global supply chain community, https://www.chain.net/c/ask. Ask your questions, join events and debates, and see the full member benefits at https://gscc.co/benefits.