Gartner recently placed generative AI for procurement in the "trough of disillusionment." After months of hype about AI-powered procurement transformation, organisations are discovering that chatbot pilots deliver marginal improvements, then momentum stalls. The problem isn't the technology. It's that most teams are applying AI to the wrong procurement problems.
This is a common scenario: a procurement team pilots a chatbot for answering basic procurement policy questions. Usage is modest, and when leadership asks "What's next?", procurement leaders struggle to identify use cases that justify further investment. The initiative quietly fades.
What follows cuts through the noise to identify which generative AI applications actually work, what prerequisites they require, and which ones remain hype.
The maturity model: Where your procurement organisation actually stands
Before exploring specific applications, assess whether your organisation has the foundation to succeed with gen AI. Three prerequisites determine readiness:
Data accessibility. Do you have digitised historical RFPs, contracts, and supplier communications? Not just stored in SharePoint folders, but searchable and structured enough for AI to process.
Process standardisation. Are your procurement workflows consistent enough to automate? Or does every category manager operate differently, making it impossible to create reusable AI workflows?
Governance framework. Who validates AI outputs before they reach suppliers or legal teams? What's the escalation protocol when AI produces questionable content?
If you answered no to two or more of these questions, address those gaps before piloting generative AI. Otherwise, you're setting up pilots to fail because your procurement organisation isn't ready to use AI effectively.
Low-hanging fruit: Applications that work today
1. Supplier communication drafting
This application requires no system integration, delivers immediate time savings, and carries low risk since outputs are reviewed before being sent.
Three communication types deliver the highest value:
- Regret letters to unsuccessful RFP bidders, maintaining relationships while delivering disappointing news
- Clarification responses during active sourcing processes, ensuring consistent language across supplier queries
- Performance issue documentation, creating defensible records with an appropriate tone
The impact compounds. Human review before sending takes minutes instead of the twenty minutes required to draft from scratch. Implementation happens quickly using existing tools with basic prompt templates.
2. Market research synthesis
Before developing category strategies, procurement teams spend significant time reading analyst reports, supplier whitepapers, news articles, and industry publications. AI condenses these sources into structured summaries, identifies pricing signals across disparate data, and maps supplier competitive positioning from public information.
The workflow is straightforward: AI generates a first-pass synthesis, humans validate the strategic implications, and AI integrates the findings with internal stakeholder input. This accelerates the research phase, freeing category managers for strategy development rather than information gathering.
The critical limitation: AI synthesises existing information but doesn't replace primary research with suppliers or validate assumptions. It's a research assistant, not a research replacement.
Medium complexity: Applications requiring process and technology investment
3. RFx document generation
This isn't quick-win territory despite what vendors claim. Success requires a historical RFx repository, standardised requirement templates, and stakeholder alignment on evaluation criteria—infrastructure that most organisations lack.
What works vs. what doesn't:
- Generating first-draft documents from requirements database
- Ensuring consistency in standard clauses across RFPs
- Creating unique technical specifications without human input — this doesn't work
- Assessing organisational risk tolerance for contractual terms — this doesn't work either
The hidden benefit is that implementing this application forces standardisation of requirements and evaluation frameworks. That delivers value even if AI adoption proceeds slowly.
4. Contract analysis and summarisation
Contract analysis pilots often get blocked by legal teams worried about liability, accuracy, and hallucination risk. These concerns are valid, which is why successful implementations focus on defensible, low-risk applications.
AI adds value by:
- Extracting standard data points across multiple contracts (payment terms, termination clauses, renewal dates)
- Flagging deviations from approved template language
- Summarising amendment histories for contracts with multiple modifications
- Identifying interdependencies between master agreements and statements of work
What AI cannot reliably do: assess the materiality of non-standard terms, interpret jurisdiction-specific regulatory requirements, or provide legal opinions. These tasks require human legal expertise.
The clearest use case is M&A due diligence, where speed matters more than perfection. AI performs initial analysis, flags potential issues, and legal teams review flagged items. Risk mitigation requires both AI and human review for an initial period until accuracy baselines are established.
High complexity: Applications that sound great but have prerequisites most organisations lack
5. Negotiation preparation and scenario planning
It sounds appealing for AI to analyse historical negotiations, supplier financial data, and market conditions and generate negotiation scenarios and counter-arguments. In practice, it rarely works because:
- Most organisations don't have structured data on past negotiation outcomes
- Supplier financial analysis requires integration with third-party data sources
- Negotiation strategy depends on relationship context AI can't access—like whether the supplier is strategically important or whether the relationship is already strained
The narrow use case where this works is repeat negotiations with similar suppliers (freight contracts, MRO agreements, temporary labour) where historical patterns are strong predictors of future outcomes. Even then, AI generates talking points and alternative proposal structures while humans determine strategy and acceptable terms.
What's still hype: Three applications to avoid, for now
Calling these out matters because failed pilots damage AI credibility internally, making it harder to secure support for practical applications that actually work.
Autonomous sourcing agents. AI lacks the judgement to make award decisions or independently assess supplier capability. Procurement decisions require understanding organisational risk appetite, supplier relationship history, and strategic priorities—context AI can't reliably access.
Predictive supplier risk scoring. This requires data integration most organisations don't have and often produces false confidence in risk assessments. Without comprehensive data on supplier financial health, operational capacity, and relationship history, AI risk scores are guesswork dressed up as analytics.
Automated contract generation with dynamic terms. Legal risk is too high and regulatory complexity is too nuanced for current AI capabilities. Contract generation requires understanding jurisdiction-specific requirements, organisational risk tolerance, and negotiation context that AI cannot reliably process.
Measuring success: Beyond "time saved"
Efficiency gains—time saved on routine tasks, faster document creation, accelerated review cycles—prove initial value but don't capture the strategic shift that matters.
Quality improvements matter more: consistent supplier communications, fewer missed contract risks, and more thorough negotiation preparation.
The real measure is whether your procurement team spends more time on strategy and less on administrative work. Track time allocation shifts, category strategy quality (measured by stakeholder feedback and outcomes), supplier relationship health, and faster identification of market opportunities.
Success isn't a dashboard metric. It's your procurement team allocating significantly more time to strategic work instead of document production.
