The issue isn't the technology. It's how we're procuring it.
The core mistake: Treating AI as a product
Most procurement still treats AI as something you buy: a model, a tool, a capability. But in case-driven organizations, regulators, central government functions and arm's-length bodies, AI operates inside decisions, and those decisions need to be consistent, auditable and defensible.
Put simply: You're not procuring AI. You're procuring a decision-making system.
The reality on the ground
Across organizations, the starting points are often fragmented systems, manual case assembly and limited visibility into how a case has actually progressed. This isn't just an internal frustration. 55% of public servants and 56% of citizens say core processes need fixing before new technology is introduced (source: Appian research).
One independent regulator we worked with was managing over 1,000 live cases at any time, with information spread across multiple systems and hours spent each week simply assembling case histories before a decision could even begin.
Adding AI into that kind of environment doesn't solve the problem. It accelerates inefficiency.
What actually works
-
Visibility before AI. AI only works when the process around it is properly understood first. Organizations with real-time visibility across their workflows identify risk earlier, prioritize more effectively and make better-quality decisions, often before AI is introduced at all.
-
A single source of truth. Fragmentation remains one of the biggest barriers to good decision-making. Unified case platforms give organizations centralized records, controlled access and end-to-end traceability. In one program we delivered, over 100,000 records and millions of documents were brought together into a single trusted system, a foundation that made reliable AI-assisted search and summarization possible afterwards.
-
Governance as the differentiator. AI becomes trustworthy when it operates inside a governed process, not alongside one bolted on as an afterthought. That means explainability, auditability and consistency are built into the workflow from day one, not retrofitted once something has gone wrong.
-
Designing for change. Public sector environments evolve constantly. Policies, regulations and citizen expectations all shift. Low-code platforms let organizations adapt workflows and introduce new AI capabilities incrementally, so the system can keep pace with change rather than requiring a lengthy redevelopment program every time something moves.
The Valtech view
The organizations seeing real impact are starting with process, data and governance, then introducing AI once those foundations are in place.
We saw this with the Government Digital Service, where our work to enhance search on GOV.UK was about helping millions of people find the information they need more quickly and confidently, not about introducing AI for its own sake.
We've taken the same approach with a national regulator, combining low-code case management with embedded AI so caseworkers get better context through intelligent search, document summarization and recommended next steps, while staying firmly in control of the decision itself.
By combining low-code case management with embedded AI in this way, teams move beyond pilots into genuine operational decision-making that’s trusted, auditable and built to scale.
Where this leaves procurement
The public sector needs better decision-making, and that starts further back in the process than most AI procurement currently considers: at the workflows, the data and the governance the technology will sit inside.
If you're exploring how to move AI beyond pilots and into real operational impact, Valtech is already working with central government and regulatory organizations to identify high-value AI use cases, design governed workflows around them and deliver platforms that scale. Get in touch to share what you're seeing in your organization, or to talk through what this could look like in practice.