The procurement value gap: why transformation isn’t delivering
Most organisations invest in procurement technology at the top, but the value is at the bottom. Procurement has been having one of its regular identity crises. Only this time the vocabulary is more expensive.
A few years ago, the conversation was about digitisation. Then it became automation, orchestration and generative AI. Now, if you want to sound current, you talk about “agentic AI” with a straight face and a well-designed slide. Each of these ideas contains something real and none of them are entirely wrong. But the way they are being discussed often misses the point. The constraint in procurement is not, at its core, a technology problem, but an outcome problem. Stop asking what the technology is. Start asking what outcome you’re trying to achieve.
Procurement as a function is not naturally set up to do transformation, it is set up to do procurement and that distinction matters more than it appears. When procurement tries to transform itself, what it often ends up doing is digitising what already exists rather than rethinking what should exist. Existing processes are wrapped in new tools, workflows are automated without being redesigned, and technology is layered onto operating models that were never intended to support the level of complexity now being asked of them. The result is predictable; a more digital version of the same thing, rather than something fundamentally different.
Procurement does not lack ambition, it has an abundance of it in presentation form. Every strategy now points in roughly the same direction: autonomous workflows, intelligent sourcing, predictive supplier management, AI-supported negotiations. The language has converged and the destination appears aligned, but while the conversation has moved quickly, most procurement environments have not. Underneath the ambition, many organisations are still dealing with fragmented supplier data, inconsistent taxonomies, disconnected systems, unclear ownership, and a surprising amount of manual coordination. What emerges is a familiar pattern; intelligence is layered on top of noise, and when it fails to deliver, the conclusion is often that the technology is immature, rather than that the foundation was not ready. That tension becomes clearer when you look at the current wave of enthusiasm around AI.
McKinsey1 writes about procurement entering the era of agentic AI, while Deloitte2 frames the function as being at an inflection point where digital capability and talent are becoming decisive. The narrative is compelling and it is also directionally correct, but it assumes a level of structural readiness that, in many cases, simply does not exist. Deloitte’s own analysis3 is explicit on this point; many organisations are still struggling to move from pilots into production, with legacy systems and weak data infrastructure acting as persistent constraints. The issue is not whether AI will matter. It is whether the environment it is introduced into is capable of supporting it.
Moving from pilot to production is well understood in practice, but less often solved. In my own experience building what became the Digital Procurement Garage, the challenge was never a lack of ideas or pilot activity. The real challenge was getting those ideas into production in a way that actually changed how procurement operated. Moving from pilot to production requires more than proving a concept, it requires integration into live processes, alignment with data structures, adoption by users, and the ability to operate reliably at scale. Without that, organisations accumulate pilots but not progress. It is also why there has been sustained interest in the Digital Procurement Garage model, with several organisations subsequently adopting similar approaches and continuing to explore how to replicate it in practice.
Most organisations don’t have a technology problem. They have a ‘pilot to production’ problem.
A more subtle but more important shift is taking place. Procurement is not just adopting new technology, it is being rebuilt, often unevenly and sometimes unintentionally, as a system. For a long time, procurement has been organised around events: a sourcing, contract, renewal, or supplier onboarding event. That model made sense in a world where information moved slowly and processes were relatively linear. But at enterprise scale, it begins to break down with too many suppliers, decisions, and interdependency to manage through manual coordination alone. What starts to emerge is something closer to a platform with a structured combination of data, workflows, and decision-making logic that allows the organisation to operate coherently at scale.
The organisations making real progress tend to converge on a similar pattern: A structured and connected data layer, cross-system orchestration able to coordinate workflows and decisions and an intelligence layer that applies AI where it can actually drive outcomes. That architecture is relatively easy to describe, but significantly harder to build in practice. In my own experience, first at BT and then more extensively at Maersk, the ambition was exactly this: to move towards a coherent, system-driven procurement platform. At BT, much of that capability was assembled through best-of-breed tools, brought together to test what was possible. At Maersk, it became a more deliberate platform build, scaling from nothing into a global capability over several years, combining both internal development and external solutions, and forcing the organisation to confront the realities of data, integration, and adoption at scale. Getting there requires sustained effort across data, architecture, operating model, and adoption. Most organisations today have elements of this, but very few have a fully coherent operation.
The traditional procurement technology debate feels increasingly unhelpful, especially the long-running argument between suites and best-of-breed solutions because both are correct in isolation. BCG’s4 recent work captures the appeal of suites well: integrated tools, harmonised user experience, lower perceived risk. At the same time, the argument for best-of-breed remains compelling, particularly where specialist capability and adaptability are required. Buyers often recognise that best-of-breed solutions align more closely with evolving business needs, even if they come with greater complexity. The problem is not the choice, but the absence of a clear architectural view of how they should work together. Without that, technology decisions become local optimisations rather than part of a coherent system design.
This is where orchestration becomes the more meaningful concept. Not as another tool, but as a way of connecting data, decisions, and actions across an increasingly fragmented landscape. For years, integration has been treated as something deep, heavy, and expensive, but that assumption is beginning to shift.
Integration does not always need to be deep. It needs to be appropriate.
Often, the most valuable integration is not full system consolidation, but something much simpler: the ability to pass context, trigger an action, and move a process forward. Regularly, the most useful solutions are also the least complex. For example, the ability to press a button in one system and send the right data to another, at the moment it is needed, can remove more friction than a fully engineered, end-to-end integration that attempts to solve everything and often solves little. These are rarely discovered in architecture diagrams, but are discovered by spending time with users, understanding how work actually happens, and where the friction really sits. In practice, that means examining the task, doing the Gemba, observing it directly, and designing around it. Without that, integration becomes an abstract exercise. With it, orchestration becomes something practical and genuinely valuable.
You can’t orchestrate what you don’t understand.
Yet, while much of the conversation has moved towards orchestration and AI, a common problem persists. Many organisations are still missing the basics and the fundamentals of procurement have not changed: understanding spend, suppliers, contracts, and risk. Those are not advanced capabilities, but in many environments, they are not reliably in place.
For procurement, that gap is not just operational, it is financial. Transformation has become a significant investment for many organisations, often running into tens of millions across platforms, integrations, and change programmes. Yet, in too many cases, the return on that investment is unclear. Systems are deployed but not used, data is not trusted and digitised processes are not improved. This results in a digital ghost town, where capability exists on paper but not in practice. For a CPO, this creates a difficult position. For a CFO or COO, it raises a more direct question. Not whether more technology is needed, but whether the organisation is extracting any real value from what it has already bought. The risk is not just overspending, but scaling inefficiency, and losing visibility of it at the same time.
A system that isn’t used isn’t transformation. It’s just an expensive digital ghost town.
This is obvious for tasks that should be straightforward but are not. For example, identifying how much the company spends with its largest supplier, questions about contract expiry, creating a supplier briefing pack ahead of a meeting, all are slow, inconsistent, incomplete, or manually stitched together. These are not edge cases and not an easy conversation to have with your CEO or board member requesting the information. If those questions cannot be answered quickly, consistently, and with confidence, it becomes very difficult to build anything more sophisticated on top.
This is the part of transformation that receives the least attention, precisely because it is not particularly exciting. Clean, structured, connected data on spend, suppliers, contracts, and risk does not make for a compelling conference presentation. But it is the layer on which everything else depends. AI does not fix its absence and orchestration does not compensate for it.
That brings the conversation back to AI itself, which is both overhyped in its current application and underappreciated in its potential. Generative AI has largely been about answers; assistance, prompting and pulling data, improving how information is accessed and presented. Agentic AI moves into action: systems are no longer just responding to prompts, but working towards defined goals, observing, reasoning, acting, and iterating. The distinction matters, but it also risks being misunderstood. An agent operating in a well-structured environment can deliver meaningful value. The same agent placed into a fragmented, inconsistent, poorly governed landscape will simply expose those weaknesses more quickly. This is not a limitation of AI; it is a limitation of design.
The implications extend beyond technology into how procurement itself is organised. Historically, the function has been defined by process execution, supplier management, and governance. Increasingly, the more strategic role lies in designing how systems work, structuring data, and governing how decisions flow across the organisation. This is a different capability set and requires a level of comfort with systems thinking, architecture, and change that has not traditionally been central to procurement. Deloitte’s5 work on procurement talent reflects this shift, highlighting the growing importance of digital capability alongside traditional commercial expertise. Technology alone does not deliver transformation, but the ability to design, adopt, and operate it does.
If you extend these trends forward, the shape of procurement begins to change. As knowledge becomes more accessible, workflows become more automated, and execution more distributed, the need for a large, centralised function diminishes. What emerges is a smaller core focused on strategy, governance, and system design, with execution pushed closer to the business. Technology enables this and AI accelerates it, but it also places greater importance on getting the underlying system right.
There is a tendency to view all of this as a technology shift in how procurement operates. From process-driven to system-driven, manual coordination to orchestrated decision-making and an execution function to one that designs and governs. The language may continue to evolve, but the underlying direction is more consistent.
The organisations that succeed in this next phase are unlikely to be the ones that adopt the most technology or move the fastest to label themselves as AI-enabled. They are more likely to be the ones that take a more deliberate approach. Those that focus on outcomes before tools, that invest in data foundations before layering intelligence, and that design how their ecosystem works as a whole rather than optimising individual components in isolation. AI will play a central role in what comes next; it will not fix the fundamentals, but it will simply make it more obvious where they are missing.
And in procurement, the fundamentals do not just matter. They determine whether any of this delivers value at all.
- McKinsey: Redefining procurement performance in the era of agentic AI
- Deloitte: 2025 Global CPO Report
- Deloitte Tech Trend #2: Unleashing agentic AI’s true potential
- BCG: Suite or Best of Breed: Seven Questions for a Smarter Applications Strategy
- Deloitte: 2025 Global CPO Report
First published on LinkedIn, 19 April 2026.

