Procurement's AI adoption stalls due to organisational bottlenecks, not technology

The argument that procurement should fix its workflow before splashing out on AI is no longer just a consultant’s refrain; in 2026 it is showing up in the numbers. Boston Consulting Group’s May 2026 supply-chain report found procurement had the lowest AI adoption rate of the 13 business functions it tracked, at 35%, even though overall AI adoption across supply-chain activities was about 44%. The same study, which drew on BCG’s Build for the Future 2025 Global Study and a supply...

-chain planning survey, said only about 30% of companies were seeing measurable AI value in planning use cases. Meanwhile, Vertice said in July 2026 that software procurement cycles were getting longer, not shorter: contracts below $10,000 took 54 days on average in Q2, while deals above $100,000 took 93 days. (bcg.com)

That matters because the drag appears to be less about drafting speed or analytics than about governance. Vertice’s breakdown points to the same culprits procurement teams complain about privately: fragmented approval chains, sequential compliance checks and a growing list of stakeholders as contract values rise. Larger deals now pull in Finance, IT Security and Legal, and the company said AI-specific supplier reviews are adding more diligence time where businesses have not yet built mature rules around data handling and model training. But even smaller purchases, which rarely trigger heavy legal or security scrutiny, also slowed in Q2 2026, suggesting the deeper problem is how work is routed through organisations rather than the absence of another digital assistant. (vertice.one)

A pre-AI case often cited by procurement advisers comes from A1 Telekom Austria Group. In a July 2020 interview with McKinsey, Michaela Mayrhofer, the company’s director of supply chain, described how A1 abandoned category silos and the traditional pyramid structure in a procurement operation handling around €2 billion of annual spend. Instead of expecting a single “perfect” buyer to master every capability, the company built a talent pool and formed small teams around individual sourcing projects, using a kanban-style scheduling system to allocate work. A1 said the shift improved its customer satisfaction score from five or six to nine, while savings achieved per strategic purchaser rose by 74%. Those gains were presented as the product of organisational redesign, workload balancing and clearer ownership rather than any breakthrough in automation. (mckinsey.com)

Nick Hudson made a similar point in June 2026 in The World Financial Review, using an MTN Group programme as his example. He wrote that strategic RFP cycle times were cut by 55% not by adding a platform or headcount, but by collapsing hand-offs into structured sessions and bringing requirements, sourcing, legal teams and suppliers together earlier. Hudson’s description of delay was strikingly mundane: a clarification left hanging for three weeks, or a contract waiting a fortnight for a signature that took two minutes to give. His conclusion was that most procurement delay is queue time, not work time, and that the real bottlenecks are decisions and sequencing. (worldfinancialreview.com)

None of that means software is irrelevant. A March 16, 2026 case study from YCP Supply Chain argued the opposite case: technology can accelerate procurement once the operating model has been reset. The consultancy said a Kuwait-based telecoms group with $6.4 billion in annual revenue, 49 million active users and operations across eight markets began by aligning stakeholders around four aims – process efficiency, embedded governance, supplier collaboration and data intelligence – before rolling out a customised source-to-pay platform. According to YCP, the programme introduced automated evaluation workflows, a “Decision Center” for side-by-side bid comparisons and e-auctions for price discovery, while giving leaders real-time visibility across all eight markets and a fully centralised sourcing repository. Those are supplier-side claims, but even on YCP’s own telling the software came after the rules, roles and decision path had been redesigned. (supplychain.ycp.com)

The same sequencing argument is now being made beyond the private sector. Deloitte wrote in March 2026 that governments which begin by buying new digital tools often end up merely reproducing layered, slow processes behind cleaner screens. Its prescription was to strip out low-value reviews, clarify decision rights and standardise common documents first, then automate selectively. Deloitte pointed to examples including the US Internal Revenue Service, where an automated review tool cut review times from hours to minutes after internal rules had been simplified, and India’s Government e-Marketplace, where standardisation and a shared digital marketplace helped reduce purchase cycle times by more than 30%. In that model, dashboards track time to award, supplier access and delivery against results, so procurement is judged by outcomes rather than by how much activity has been pushed through a system. (deloitte.com)

Taken together, the evidence suggests the current procurement-AI debate is being framed too narrowly. Writing in The European Business Review, Hudson argued that AI tends to amplify the process already in place, whether efficient or dysfunctional. BCG’s May 2026 report arrives at much the same destination from another route: it says AI scale-up is hindered above all by people, organisation and process issues, which it groups as 70% of the challenge, compared with 20% for technology and 10% for algorithms. If that diagnosis is right, the procurement leaders most likely to show genuine progress this year will not be the ones with the flashiest copilots. They will be the ones that can say, with dates and numbers, which approvals were removed, which reviews now run in parallel and how much time disappeared from the path to award before AI was asked to do any speeding up. (europeanbusinessreview.com)

Source: Noah Wire Services