AI in government procurement: practical gains, real risks
The Procurement Act 2023 changed how the public sector buys, with new freedoms and a bigger push for transparency. At the same time, government has published guidance on safe, effective AI use. A recent roundtable of senior officials and industry specialists looked at what AI can deliver in procurement-and what gets in the way.
Here's the signal, without the noise: AI can clear low-value work, improve consistency, and sharpen risk detection. But unclear tech, shifting suppliers, and weak explainability can derail progress if you don't set the right guardrails.
Policy context
The Act introduces simpler routes to market and a new central digital platform. Guidance like the government's AI playbook sets expectations for safe deployment and oversight.
Where AI can add clear value
- Reduce busywork: triage supplier emails, auto-check specs for completeness, and summarise due diligence findings so commercial teams focus on choices, not chase.
- Speed up market insight: faster searches and synthesis help teams understand what they're buying and who can supply it.
- Write better, once: reuse standard specifications (e.g., laptops) across departments instead of starting from scratch each time.
- Risk and fraud detection: flag cartel-like patterns, odd bidding behaviour, and supply-chain fragility before it bites.
- Co-drafting specs: AI agents can help commercial and product leads draft and iterate requirements together.
The sticking points
- Tech shifts fast: terms, features, and vendors change. What looked safe last quarter can age quickly.
- Black-box tools: thin documentation and weak explainability make audit and assurance harder-and errors easier to miss at scale.
- Supplier viability: strong claims, light evidence. Teams can struggle to judge what's good and what's not.
- Change at "supertanker" speed: getting people to adopt new tools is often harder than buying them.
How to buy and implement smarter
Treat the supplier relationship as practical collaboration, not a one-off transaction. You're learning together, so set expectations early and make outcomes the anchor.
- Run small, targeted competitions to find fit, then expand. Keep scope tight and measurable.
- Align on objectives, service levels, and ethics up front. Bake explainability and audit into deliverables, not footnotes.
- Focus on outcomes over a single fixed solution. Expect iteration as the tech and your needs evolve.
Data, platforms, and transparency
The new central digital platform that replaces and extends Find a Tender should make contract data easier to publish and analyse. As more records land, expect better benchmarking, market insight, and accountability-if teams actually use the data.
Work safely: start small, learn fast
- Don't rush. Pick low-risk use cases, prove value, then scale.
- Test in controlled environments with clear guardrails, red-team reviews, and rollback plans.
- Treat failures as shared learning, not blame. Close the loop with short, written post-mortems.
Build capability and culture
You need commercial AI champions-people who can spot opportunities, challenge weak claims, and show what "good" looks like. Pair them with policy, data, and security leads so decisions stick.
If your team needs a structured way to build skills, explore role-based training options for commercial and procurement professionals here: AI courses by job.
Action checklist for the next quarter
- Map two high-volume tasks (e.g., supplier Q&A, spec reuse) and test AI support with real data.
- Define non-negotiables: data handling, audit logs, explainability, and model update controls.
- Set a "living" risk register for AI use in procurement and review it monthly.
- Create a standard supplier questionnaire covering model provenance, training data, evaluation results, and incident response.
- Pilot the new central platform's data features for spend and supplier analysis; publish what you can.
- Name your AI champions and give them time and budget to run proofs of concept.
Bottom line
AI won't fix procurement on its own. But with tight use cases, clear safeguards, honest suppliers, and teams who know what "good" looks like, it can free skilled people to do higher-value work-and make public money go further.
Your membership also unlocks:
