Procurement’s AI opportunity depends on fixing the foundations first

At Everest Group’s 2026 India Procurement and Vendor Management Roundtable, the discussion moved well beyond the long-running question of whether procurement can do more than cut costs. The more consequential issue, participants agreed, is how the function can shape business outcomes in an environment defined by pressure on margins, geopolitical uncertainty, fragmented supplier ecosystems and the accelerating use of artificial intelligence.

That broader ambition is increasing...

ly visible across the market. Fairmarkit has argued that procurement is becoming a strategic differentiator as AI alters how teams assess suppliers, manage spend and respond to volatility, while ISG has described a similar shift from cost reduction towards a more competitive role inside the enterprise. McKinsey, meanwhile, has framed procurement as a source of sustainable advantage when it is linked to digitisation, automation and analytics rather than confined to transactional control. Everest Group’s roundtable reflected that same direction of travel, but with a strong emphasis on the practical choices organisations now have to make.

A recurring point was that procurement’s value cannot be reduced to savings alone. Leaders at the event described value as contextual, changing according to the business problem at hand. In one case, it may mean contract flexibility that allows services to scale quickly; in another, it may involve reducing the cost to serve, limiting geopolitical exposure, speeding up innovation or enabling multiple suppliers to function as a single delivery network without eroding quality.

That creates a familiar problem: the benefits procurement is being asked to deliver are often longer term and less visible than the immediate impact on price. A decision that raises near-term costs may still be the right one if it lowers risk or improves resilience later. Yet procurement teams continue to struggle to define and communicate those wider outcomes in language the business accepts. One example discussed at the roundtable involved a supplier that appeared to be meeting its key performance indicators, but was still regarded internally as unsatisfactory. The issue was not necessarily delivery failure; it was that the measures themselves had not captured what stakeholders actually cared about.

Participants cautioned against the habit of creating large sets of supplier KPIs only to collapse them into a narrow scorecard that oversimplifies performance. The more workable approach, they suggested, is to begin with a small number of measures that matter and refine them over time as the relationship matures. That discipline matters because procurement’s credibility depends on trust, and trust is built when business stakeholders believe the function understands both commercial trade-offs and operational reality.

That trust is not always easy to earn. Some organisations still deal with para-purchasing, where business teams engage suppliers outside established procurement channels. That weakens visibility, reduces control over risk and makes it harder to build a shared evidence base for supplier decisions. Several participants said procurement’s influence improves when it is structurally closer to enterprise strategy, for example through governance models aligned to growth priorities or to enterprise objectives and measures. In those cases, procurement is not merely reporting savings after the fact; it is helping define what value should look like in the first place.

The roundtable also highlighted the increasingly close relationship between procurement and the vendor management office, or VMO. The two functions can no longer be treated as separate stages in a linear process, because value negotiated at contract signature is either preserved or lost during delivery. Procurement and vendor management therefore need connected responsibilities across sourcing, contracting, transition, performance management, risk, relationship oversight and innovation.

Several leaders described models designed to make that connection more deliberate. One organisation begins planning major renewals as much as two years ahead, bringing procurement, vendor management, legal, risk, information technology and data security into a shared governance forum. Another places vendor managers within business units so supplier performance stays tied to stakeholder satisfaction rather than drifting into a purely administrative exercise. The structures differ, but the principle is the same: supplier management must stay close to the business result.

The case for integration becomes even stronger in multi-vendor environments. If several suppliers contribute to the same outcome, then collaboration cannot be left to goodwill. Everest Group argued that shared milestones, interdependencies, acceptance criteria, handoffs, escalation routes and cooperation obligations should be written into the contract. Without that, each supplier may meet its own service levels while the wider programme still fails. Where possible, participants also said, organisations should consider providers with end-to-end capability to reduce the coordination burden and the risk of fragmentation.

Artificial intelligence was another major theme, but the discussion was notably measured rather than breathless. Everest Group outlined how agentic AI can support spend analytics, sourcing, negotiations, contract management, supplier risk assessment, performance management, intake and procure-to-pay. Yet it also warned that AI tends to expose pre-existing weaknesses in process design, technology, skills and data quality. If those foundations are weak, adding another platform can deepen fragmentation instead of improving intelligence.

That caution resonated with participants. Some are exploring in-house AI tools because of data-security concerns and because broad off-the-shelf platforms may not understand the specifics of procurement well enough. But building internally is not a free alternative; it can create heavy demands in specialist skills, integration and maintenance. The choice, as the discussion framed it, is not simply build or buy. It can also involve white-labelled tools, service-provider solutions or a more bespoke approach, depending on how strategically important the capability is and what its long-term cost of ownership will be.

The group was also wary of treating every AI-labelled product as a breakthrough. A proof of concept, several participants argued, should test a real business issue rather than merely demonstrate that a tool can perform an impressive task. Before scaling AI, organisations need to redesign the underlying process, determine data and access requirements, clarify who remains accountable for decisions and define the point at which human judgement is still essential. Reducing manual effort, they noted, does not automatically create a business case if implementation and operating costs outweigh the benefit.

From there, the conversation moved naturally to contracts. Traditional full-time-equivalent pricing models are under strain as automation changes the amount and type of labour required. Output-based, outcome-based, digital-agent and hybrid commercial models are all gaining relevance, but Everest Group’s view was that many buyers should not rush straight to outcome-based pricing. Outcomes are often affected by factors the supplier cannot control, including enterprise adoption, process maturity, data quality, investment appetite and the behaviour of other vendors.

For that reason, output-based models can serve as a useful bridge. They move the commercial discussion away from pure input pricing while allowing both sides to establish baselines, tighten operational discipline and build the trust needed for more ambitious constructs later on. Hybrid arrangements can also help by combining a stable fee with incentives linked to measurable performance or value creation.

Flexibility was a consistent theme. In multi-year technology deals, the delivery model, degree of autonomy, productivity potential and cost structure may change materially over time. Contracts therefore need room for recalibration, benchmarking and amendment without forcing a complete renegotiation every time the market shifts. Transparency matters too. Buyers want to see how AI-related gains will be achieved and shared, while providers need a fair way to recover genuine investment. That means assumptions about productivity, platform charges, model and token costs, human oversight and value sharing should be visible rather than buried in a black-box fee.

The AI talent market adds another layer of complexity. Premium rates may be justified where skills are genuinely scarce and directly improve delivery, but buyers were warned against turning a temporary shortage into a permanent mark-up. Market benchmarking, time-limited premiums, caps, periodic review and knowledge-transfer obligations were all cited as possible controls.

Taken together, the roundtable suggested that procurement’s next phase will be defined less by slogans about transformation than by the discipline with which organisations execute it. The function is gaining influence through a richer value narrative, tighter alignment with vendor management and a deeper role in technology and sourcing decisions. But influence alone is not enough. The real test is whether procurement can use that influence to choose the right problems, structure the right relationships and build the right commercial models in a market that is changing too quickly for rigid thinking.

Source: Noah Wire Services