Written by Marijn Overvest | Reviewed by Sjoerd Goedhart | Fact Checked by Ruud Emonds | Our editorial policy

AI Readiness in Procurement: The 2026 Benchmark Report

About This Benchmark

This report covers the findings of Procurement Tactics' 2026 AI Readiness in Procurement survey. The sample includes procurement leaders across Europe, the Middle East and Africa, Asia-Pacific, and North America, spanning manufacturing, government, energy, healthcare, technology, and professional services. Roughly a quarter of procurement leaders are Directors of Procurement, CPOs, or VPs; the majority are Procurement Managers, Category Managers, Senior Buyers, and Procurement Analysts. Company sizes range from under 500 employees (40% of the sample) to over 50,000 (7%), across manufacturing, government and public sector, energy and utilities, healthcare and pharma, technology, and financial services.

The survey was fielded in the first quarter of 2026. It covers AI adoption stage, tool usage, time-to-automation candidates, procurement data maturity, value-capture credibility with finance, operating model, AI policy status, and strategic priorities for the next twelve to eighteen months.

The purpose of the benchmark is practical. Procurement leaders asking "where should our AI adoption be right now?" rarely get a useful answer from vendor reports or consulting claims. The answer this report is designed to give is: here is where a representative sample of your peers actually are, and here is what the top of that sample is doing differently.

This year’s benchmark was produced in partnership with Suplari, whose procurement spend-intelligence expertise helped shape the survey design and the analysis behind these findings.

The Manual Work Problem is Worse than Most Procurement Teams Admit

The first significant finding of the 2026 survey is the distribution of manual workload.

48% of procurement teams, describe their operating reality as "very high manual workload": 60% or more of team capacity consumed by gathering data, preparing reports, or resolving urgent operational issues. Another 34% fall into "high manual workload" at 40–60%. That leaves 18% of procurement teams at moderate-or-below levels of manual work, and exactly 2%, two procurement leaders, describe their operation as "highly automated" with under 10% manual load.

The practical meaning of those numbers is that more than four out of five procurement teams spend a minority of their time on strategic work. The rest of the week goes to data wrangling, reporting preparation, and firefighting. Strategic category management, supplier development, and market intelligence happen in the margins of an operational reality that looks a lot like an analyst job rather than a procurement one.

This is the finding that drives the business case for AI in procurement more than any other. Not that AI makes procurement smarter, though it often does, but that AI compresses the manual work that currently prevents procurement from being strategic. The shift from 60% manual load to 40% manual load is not a productivity win; it is a reallocation of a procurement team from operational to strategic work. That is the transformation.

Claude compresses contract review from a day to under an hour. Copilot compresses spend analysis from hours to minutes. ChatGPT compresses market research, supplier communications, and RFP drafting. Each alone produces a visible improvement; together, they start to move the 60%-manual teams into the 40%-manual band. The teams that do this systematically are the ones most likely to appear in the top quartile of the next version of this benchmark.

Adoption Stage, 64% are Still 'Exploring'

The benchmark uses a five-stage adoption model: Exploring (interest but no specific initiative), Experimenting (running pilots), Deploying (expanding successful pilots into production), Embedded in operations (AI integrated into procurement workflows), and AI-driven procurement (AI continuously analysing and decision-supporting). The distribution is sharply skewed to the left.

53% of procurement teams, place themselves at Exploring. They are aware AI matters; they have not yet launched a procurement-specific AI initiative. Another 27% are at Experimenting, running pilots. 12% are at Deploying. Only 4% describe themselves as Embedded, and 4% as AI-driven.

The implication is that a procurement team at Experimenting is already in the top half of the distribution, and a team at Deploying is in the top 20%. The "everyone is doing this" narrative that sometimes surrounds AI adoption in procurement is empirically wrong. Most procurement teams are behind the point they think they are, and the point at which they think they are behind is itself overstated.

The teams that move through the stages fastest tend to share three habits. They pick a specific workflow, usually contract review, spend analysis, or supplier risk, rather than trying to deploy AI across the whole function at once. They ship a proof-point early and publish it inside the organisation, which generates the internal momentum that gets the next workflow funded. And they invest in structured training, because the skill gap between an unskilled AI user and a skilled one on the same tool is large enough to determine whether the pilot produces usable output.

Tool Usage, Copilot Leads, Claude is Still Emerging

The tool landscape in procurement is now a multi-vendor one. Microsoft 365 Copilot leads in overall adoption at approximately 45% of procurement teams using it in some combination. ChatGPT is close behind. Google Gemini sits around 14%, and Claude at about 10%.

Those percentages are not mutually exclusive. A growing minority of procurement teams run two or more AI tools simultaneously, typically Copilot alongside ChatGPT, or increasingly Copilot alongside Claude for higher-complexity work. This pattern is interesting because it is not what most procurement leaders expect when they start with a single-tool decision. The expectation is that the organisation picks one. The reality is that the organisation ends up with a stack.

The reasons for the multi-tool pattern tend to be workflow-specific. Copilot is the default for work that lives inside Microsoft 365 applications because it sees the documents and emails natively. ChatGPT or Gemini handles short-turn drafting, research, and general iteration work because it is fast and the user interface is familiar. Claude earns its place on workflows where the quality of the reasoning matters most, long contract review, multi-step analyses, and agentic automations. The 10% of procurement teams that currently use Claude are disproportionately concentrated at the Deploying and Embedded stages of the adoption curve.

The implication for procurement leaders is that the "pick one tool" framing may be the wrong question. The better question is "which workflows go to which tool, and how do we train the team to use each one well". That is a different conversation, and a more productive one, than the vendor-selection conversation most procurement organisations currently have.

The AI Policy Gap, 41% Have None and No Plan

The policy finding is the one that consistently surprises procurement leaders reading the benchmark for the first time. 41% of procurement organisations, 50%, have no formal AI policy and no plans to create one. Another 31% are working on a policy. Of the remaining 28%, 11% have a policy that is not consistently followed, and 17% have a policy that is actively enforced.

That distribution matters because the procurement teams most aggressively adopting AI are disproportionately concentrated in the 17% with enforced policies. This is not a coincidence. AI scale in a procurement organisation requires a policy because without one, every individual user invents their own rules, and the organisation accumulates risk in proportion to the number of users.

The policies that work are not long documents. They tend to cover three areas concisely. Data classification: what categories of procurement data are appropriate for which AI tools. Memory and retention: what AI features can retain organisational context, and how long. Review and approval: which AI-generated outputs require human review before external use. Written to this specification, the policy is usually under ten pages. Written to a legal-document standard, it becomes long enough that nobody reads it, and the effect is the same as not having one at all.

The AI Fundamentals for Procurement Teams program includes a ready-to-edit AI policy template as part of its policy module. Procurement leaders adapting an existing framework typically produce a working policy in an afternoon, faster than designing one from scratch and more likely to cover the right ground.

The P&L Credibility Gap with Finance

33% of procurement teams, say procurement's financial impact on the company is "difficult to quantify" for finance. Another 30% say procurement reports savings but the methodology is debated by finance. Only 7% describe integrated value tracking or real-time financial impact visibility.

This finding is worth pausing on because it is under-discussed in the procurement AI literature. Most AI content for procurement focuses on efficiency, how much time AI saves, how many hours per category manager per week. That framing is useful, but it is the wrong framing for the conversation with a CFO. A CFO cares about P&L impact, not productivity metrics. Procurement teams that can credibly prove P&L impact get next year's budget; procurement teams that can only prove productivity do not.

The intersection of AI and the credibility gap is this: AI tools make it easier to produce the data artefacts finance wants (consistent savings methodology, variance analysis, forward-looking ROI calculations) without increasing the analyst headcount required to produce them. The procurement teams that use AI to close the credibility gap with finance are the teams most likely to secure funding for the next wave of AI adoption. It is a compounding effect, and a significant one.

Operating Model, Reactive is Still the Default

55% of procurement teams describe their operating model as reactive or mostly reactive, primarily responding to issues after they occur, with most work driven by operational requests. 34% are "emerging proactive", looking for opportunities but with limited tooling. Only 9% describe themselves as data-driven and proactive. A single procurement leader reported an AI-enabled predictive operating model.

The reactive default is not a skills problem. Procurement professionals know what a proactive operating model looks like; they describe it fluently when asked. The problem is the loop between the manual workload (48% of teams at 60%+ manual load) and the reactive operating model. Teams that spend most of their time firefighting cannot become proactive without first offloading enough of the firefighting to create capacity.

This is why the AI adoption argument for procurement is not primarily an efficiency argument. It is a capacity argument. The fastest route from reactive to proactive in a procurement function runs through AI-assisted work compressing the operational load that currently prevents the team from looking ahead. The 9% of teams that describe themselves as data-driven and proactive are disproportionately the teams using AI tools to compress the operational load.

Strategic Priorities for 2026–2027

The survey asked procurement leaders to identify their procurement organisation's primary strategic priority for the next 12–18 months. The distribution is clarifying.

39% of procurement teams, name cost and savings as the primary strategic focus. 26% prioritise digital process efficiency (tools, automation, improved workflows). 14% focus on data and visibility foundations. 12% on strategic value leadership (transforming procurement into an enterprise value driver). Only 10% list AI-enabled scale and insight as the top priority.

This finding often surprises procurement teams because the AI narrative in 2025–2026 has been intense enough that it is easy to assume most procurement teams have reoriented toward AI. They have not. AI is a means to the strategic priority, usually cost, efficiency, or data, rather than the priority itself. Procurement teams that frame AI adoption as "we are doing AI" tend to struggle for funding. Procurement teams that frame it as "we are accelerating our cost and savings programme through AI-assisted spend analysis and contract review" tend to get funded.

The framing matters because the CFO and CEO conversations that determine AI funding for procurement are rarely about AI itself. They are about the strategic priority AI is serving. Getting the framing right is the difference between a funded programme and an exploratory pilot.

What the Top 9% are Doing Differently

The procurement teams that describe themselves as data-driven and proactive, the 9% of the sample, share several habits that distinguish them from the other 91%. The patterns are consistent enough across the sample to be worth naming.

They run multiple AI tools deliberately, not accidentally. The median tool count in this segment is two, occasionally three. Copilot is almost always one; Claude or ChatGPT is typically the second. The tools are assigned to workflows, not treated as substitutes for each other.

They invest in structured training. The gap between skilled and unskilled use of the same AI tool is large enough that teams without training produce inconsistent output. The top-quartile teams almost uniformly have run some form of structured AI training in the preceding twelve months, internal, external, or both.

They have a real AI policy. Not a one-page disclaimer; a working policy covering data classification, memory and retention, and review requirements. The 17% of the sample with an actively enforced policy is disproportionately the same 17% making the most progress on adoption.

They close the credibility gap with finance. Savings reporting is consistent, methodology is agreed with finance, and AI-generated efficiency work is framed in P&L terms rather than productivity terms. The CFO is usually the critical internal sponsor for procurement AI at this level of maturity.

They have a data foundation that makes AI useful. Spend data is accessible; supplier master data is reasonably clean; contract repositories are organised enough for Copilot or Claude to retrieve against. Teams without this foundation tend to report that AI "doesn't work for us", when what is actually not working is the data foundation AI needs to work on.

These five habits are learnable. They are also cumulative, the 9% did not adopt them simultaneously. The pattern across the procurement leaders who described their progress in the open-response fields is that the journey from Exploring to AI-enabled takes twelve to twenty-four months, with each habit reinforcing the next.

How to Use this Benchmark Inside a Procurement Organisation

The benchmark is most useful as a starting point for three internal conversations.

The position conversation. Where is our procurement organisation relative to this sample? Honest placement on the five-stage adoption curve, the manual workload spectrum, and the policy maturity distribution is more useful than any single plan. Procurement teams that overstate their current position tend to design AI initiatives that do not match their actual readiness.

The priority conversation. Given our position and the strategic priority our organisation has set, which two or three AI-enabled workflows would move us forward most? The answer is usually narrower than the initial temptation to deploy AI everywhere. Two workflows done well beat six workflows done superficially.

The capability conversation. What would we need to build, in policy, training, data foundation, or tool access, to close the gap between where we are and the top quartile? The honest answer usually includes training, a policy, and one or two data foundation investments. The AI Fundamentals for Procurement Teams program covers the first two; the third is an internal effort, but it is materially easier once the first two are in place.

The benchmark is not a scorecard. It is a map. Procurement teams that use it as a map, to orient their own position and identify the nearest productive moves, get more from it than teams that use it as a scorecard to grade themselves against an abstract standard. The goal is not to be in the 9%; the goal is to move in the direction of the 9%, one workflow at a time.

Download 20 AI Use Cases for Procurement Professionals →

Frequently asked questions

Who conducted the AI Readiness in Procurement 2026 survey?

Procurement Tactics conducted the survey in Q1 2026. Respondents were drawn from Procurement Tactics' subscriber base, partner network, and direct outreach. The sample includes procurement leaders across Europe, the Middle East and Africa, Asia-Pacific, and North America, spanning manufacturing, government, energy, healthcare, technology, and professional services.

Which AI tool is most common in procurement in 2026?

Microsoft 365 Copilot, at approximately 45% of procurement teams using it in some combination. ChatGPT is close behind. Gemini sits around 14% and Claude around 10%. Two-tool combinations are increasingly common.

How can procurement teams benchmark against this data?

Start with the five-stage adoption curve: Exploring, Experimenting, Deploying, Embedded, AI-driven. Honest placement on that scale is usually more useful than any other benchmark. Then assess manual workload, policy maturity, and tool coverage against the distributions in this report.