Organisational foundations key to converting AI promise into procurement value

Despite advancements in AI, procurement organisations struggle with margin erosion due to organisational and process shortcomings. Experts emphasise that aligning people, data, processes, and governance is essential for AI to deliver sustainable value and efficiency gains.

A stark contradiction now defines the procurement landscape: vendors and consultants promise that artificial intelligence can shoulder the bulk of procurement work, yet many organisations still report...

persistent margin erosion and disappointing returns from automation spending. The reason, evidence suggests, lies less in the algorithms than in the organisational bedrock those systems are asked to sit upon.

At the centre of the problem is a set of non-technical shortcomings that routinely undermine AI roll-outs. Recent industry research points to four interlocking deficits that prevent procurement functions moving beyond pilots and proofs of concept.

First, process and data fragility. Large portions of procurement remain manually intensive and poorly digitised. A study by Ivalua found that 53% of procurement and supplier-management processes are not yet digital, with teams estimating they spend roughly 22% of their time on manual paperwork. McKinsey research adds that many organisations have only limited data infrastructure maturity, with a substantial share of spend data not held in a central, clean repository. The practical consequence is simple: feeding AI with inconsistent, incomplete or uncategorised data amplifies existing waste rather than eliminating it.

Second, fuzzy strategy and divergent objectives. Procurement leaders frequently set broad ambitions such as “improve efficiency” without breaking them down into measurable, staged goals. Analysis by ProcureInsights and other industry commentators shows that a large proportion of procurement-focused AI initiatives , some surveys put the failure or abandonment rate as high as 60–70% , fail because of scope creep, budget overruns or misalignment between ambitions and capability. The result is diffusion of effort and scarce leadership attention.

Third, a yawning talent and literacy gap. While automation can take on repetitive tasks, the residual work , supervising models, resolving exceptions, embedding outcomes into supplier relationships , demands hybrid capabilities that many teams lack. Thought leadership on procurement transformation stresses that people and change management typically account for the majority of success in digital programmes; yet upskilling and creation of new roles such as data stewards and prompt engineers remain uneven.

Fourth, governance and risk shortfalls. The emergence of more autonomous “agentic” AI introduces fresh risks around security, accuracy, ethics and regulatory compliance. Industry authorities emphasise the need for robust human-in-the-loop controls and explicit governance frameworks; building those safeguards requires time and investment that can delay near-term margin gains but are essential to avoid costly failures.

Technology choice itself is another frequent misstep. Instead of matching tools to task complexity, some organisations leap to sophisticated generative AI where simpler automation would suffice. Analysts note that robotic process automation remains highly effective for high-volume, rule-based transactions such as procure-to-pay, and that an overreliance on complex AI can add unnecessary expense and brittleness. Conversely, when the right level of automation is employed against streamlined processes and good data, results can be substantial.

There are, however, practical exemplars of how to extract value. Industry case studies and consultancy reports show companies that re-engineered processes first and then layered AI on top achieving material gains. McKinsey describes examples where AI-enabled agents delivered double-digit reductions in sourcing costs and significant efficiency uplifts in negotiation and analysis. Independently reported accounts credit firms such as Bosch with dramatic improvements in agentic procure-to-pay workflows after redesigning the underlying process rather than simply deploying new software.

The losses from failed or abandoned projects are tangible. ProcureTech surveys and market analyses estimate global waste from unsuccessful procurement AI initiatives runs into the tens of billions annually, driven by sunk project spending and ongoing operational inefficiency. At the same time, research from automation advocates shows that successful AI adoption can deliver sharp improvements in visibility, cycle times and strategic focus when foundations are sound.

The policy lesson for procurement leaders is clear: treat AI adoption as an organisational transformation first and a technology implementation second. Prioritise process simplification, data governance and staged, measurable objectives; invest in upskilling and new hybrid roles; select the right automation modality for each task; and establish governance that balances autonomy with oversight. Doing so sacrifices some immediate ease of deployment but substantially raises the odds that automation will translate into lasting margin relief.

In short, procurement’s future will be won not by chasing the largest or most hyped models, but by rebuilding the work those models are asked to perform. When organisations align people, processes, data and governance, AI shifts from a speculative promise to an engine of durable value.

Source: Noah Wire Services