Procurement leaders turning to AI insights despite imperfect data landscape

Many procurement leaders still treat imperfect data as a reason to postpone artificial intelligence. That is increasingly the wrong conclusion. The more useful test is not whether spend data is flawless, but whether it contains enough reliable signal to support decisions, spot risks or trigger action.

That distinction matters because procurement data is rarely static. Mergers add new ERP systems, supplier hierarchies break, classification rules drift and migrations wipe out pro...

gress. In other words, the idea of waiting for a permanently clean baseline is unrealistic. According to Hackett Group research cited in Sievo’s analysis, more than 30% of companies still do not plan to invest in spend analytics, which suggests many organisations are leaving value on the table while they wait for conditions that may never arrive.

IBM’s definition of AI-ready data is broadly in line with that view: information must be trusted, accessible and of sufficient quality to support AI use cases. But in procurement, AI-ready does not have to mean perfect. It can mean well-classified spend in certain categories, reliable supplier records in selected areas or enough consistency to answer questions about savings, risk and contract performance.

That is why the best starting point is often not the whole data estate, but the pockets where the signal is already good enough. Large enterprises typically have some categories that are better governed, more actively managed and easier to analyse than others. Those are the places where conversational analytics can begin to deliver value, even if other parts of the business remain messy.

The shift is important because it changes the relationship between procurement teams and data. Traditional reporting often forces people into validation mode, with analysts and category managers spending time arguing over whether numbers are correct. AI, by contrast, allows teams to interrogate the data directly. Users can ask where tail spend is growing, which suppliers look exposed this quarter or where pricing is rising despite softer market benchmarks.

That also helps explain why automation is more than a dashboard upgrade. Once AI can detect a signal, it can also prompt a workflow. A supplier renewal can be flagged, market pricing can be checked and a negotiation review can be triggered without waiting for someone to notice the issue in a weekly report. Over time, each interaction helps expose missing records, duplicate suppliers and unclassified spend that static audits may never uncover.

The broader case for adoption is not mainly about reducing headcount. Sievo argues that the bigger prize is commercial: a modest improvement in savings on addressable spend can dwarf internal efficiency gains. For a business with $7bn of addressable spend, even a 1% improvement in savings could amount to $70m, far outweighing a simple cost-cutting story.

That is why the strongest procurement AI business case usually starts with spend visibility. Deloitte has said that data quality remains one of the biggest obstacles to generative AI in procurement, even as many chief procurement officers prepare to invest. TechTarget similarly notes that procurement data is often fragmented across ERP systems, portals and spreadsheets, making standardisation a prerequisite for reliable AI output.

Governance is another missing piece. A study by ProcureAbility, reported by Supply & Demand Chain Executive, found that while 96% of procurement and IT teams are collaborating to some extent, 54% are not working together on AI governance. That gap matters because AI adoption is not only a data problem; it is also a controls, ownership and operating-model problem.

What good adoption looks like is fairly consistent. Organisations start with the data they can trust, then expand as classification, supplier normalisation and ERP consolidation improve. They use AI to reveal quality issues more quickly, rather than waiting for a perfect remediation programme. And they match the use case to the maturity of the underlying data.

Sievo’s maturity model captures that progression. At the early end, organisations are assembling spend manually and making decisions with limited visibility. Further along, they have a consolidated spend view, but classification is patchy and self-service remains difficult. Higher maturity brings standardised, normalised data, systematic savings identification and, eventually, AI-driven workflows that can move towards more autonomous execution.

The message for procurement leaders is straightforward: data quality is not a reason to stand still. It is the reason to start in the right place, prove value where the signal already exists and let the use of AI improve the data foundation over time.

Source: Noah Wire Services