5 challenges of using AI in procurement

1. Data quality

Procurement data is rarely in a form suitable for algorithms to use, as purchasing records are typically scattered across ERP systems, supplier portals and users' personal spreadsheets, which leads to inconsistent categorization.

Missing fields also lead to problems. Missing data can potentially include supplier contact information or delivery dates, as well as critical information such as compliance certifications and contract terms, which might exist only in signed PDFs.

Organizations that clean, standardize and properly maintain their procurement data will derive more value from using AI in their processes.

2. Integration issues

Connecting AI systems to existing infrastructure can lead to problems. Older legacy ERP platforms often lack the APIs that would enable real-time data exchange, leading to workarounds such as batch uploads and manual exports that can introduce bottlenecks and lead to errors.

Also, supplier platforms often use their own proprietary data formats and standards, which can make it difficult for users to aggregate information from multiple vendors or compare vendor performance.

Companies attempting to use AI in procurement need to create a strategy for integrating platforms, which might include adding middleware layers to connect disparate systems. Integration will require decisions about where data will live, how it will flow between systems and which applications will serve as sources of truth.

3. Stakeholder resistance

Procurement teams might resist AI adoption if they believe that automation will lead to them making fewer decisions or even replace their jobs.

Executive communication about AI in procurement should frame the use of the technology as a business transformation that involves the procurement team rather than an IT project that's being imposed on them. The messaging should convey that AI is a strategic priority that requires cooperation across departments.

Having teams work on the design of AI systems can also help reduce resistance, as employees will learn the strengths and weaknesses of AI.

4. Skills gaps

Procurement workers who were trained in negotiation, supplier relationship management and category strategy might lack the data literacy needed to interpret model outputs. For example, if AI flags a supplier as high-risk or recommends a pricing adjustment, employees could lack the skills to decide whether the recommendation merits action or stems from a model limitation.

Training programs can help procurement employees work with AI tools, including interpreting their outputs, overriding recommendations when needed and explaining decisions to internal stakeholders.

5. Transparency

Regulators and auditors are increasingly asking whether employees made decisions through defensible processes, rather than whether companies made correct decisions. If AI recommends suppliers or flags contracts for review, employees must be able to identify the source of that reasoning using data lineage, the model's logic and the decision thresholds.

Companies must avoid bias in supplier selection or price optimization, which can violate fairness standards and also result in suppliers that could be a good fit being passed over. For example, pricing models that were trained on negotiations with large suppliers might avoid selecting smaller vendors that lack an equivalent bargaining power but could be a good fit otherwise.

Creating governance frameworks can help establish when a human needs to review high-stakes decisions and create an audit trail. For example, if the procurement department selects a supplier based on AI scoring, a decision record can capture which version of the model was used, the data that it analyzed, the weights that were applied to different risk factors and whether a human employee modified the recommendation.

The audit trail will satisfy regulatory requirements, establish internal accountability and create a feedback loop for improving model performance over time.

An analyst's perspective

In a 2025 report, Gartner noted the potential challenges of using generative AI in procurement and urged organizations to plan more thoroughly for the technology's use. While some early adopters are seeing benefits, many organizations are experiencing uneven ROI or falling short of expectations … according to Gartner.

Steps that Gartner suggested chief procurement officers should take to alleviate potential issues included "monitor[ing] evolving regulations to ensure compliant implementation and seek[ing] expert guidance as needed [and] upskill[ing] teams in digital dexterity, human-machine interaction, and prompt engineering to prepare for more AI-enabled processes.

Framing and governance can lead to success

AI in procurement is more likely to scale and perform well when organizations treat it as a governance initiative rather than just a technology purchase.

The organizational prioritization of data quality, cross-functional alignment and transparency will determine whether AI in procurement will deliver value.

Companies should establish data, governance and cultural foundations before deploying models as well as securing executive sponsorship and upskilling employees.

Technical sophistication matters, but organizational readiness matters more.

Donald Farmer is a data strategist with 30-plus years of experience, including as a product team leader at Microsoft and Qlik. He advises global clients on data, analytics, AI and innovation strategy, with expertise spanning from tech giants to startups. He lives in an experimental woodland home near Seattle.