CASE STUDY: How the UK Nuclear Decommissioning Authority use Valyu to power horizon scanning

How a real-time horizon-scanning capability is helping the Nuclear Decommissioning Authority bring relevant developments into view sooner.

The need to see change early

The Nuclear Decommissioning Authority is responsible for cleaning up the United Kingdom’s earliest nuclear sites. Its work spans different locations, organisations and time horizons, so decisions often depend on understanding what is changing beyond the boundary of any one site.

A development elsewhere in the world may not look significant at first. But if it affects the nuclear industry, a region in which work is taking place, or the wider operating environment, it can quickly become relevant. The challenge is to notice those signals early, understand their context and decide whether they matter.

This is the role of risk horizon scanning. It is less about collecting more information and more about helping experienced analysts find the information that deserves attention.

When searching becomes the bottleneck

Before the new capability was introduced, risk analysts relied heavily on manual searching across large volumes of information. Their judgement was not a constraint. The difficulty was the time required to find, filter and organise material before assessment could begin.

That created a familiar tension: every hour spent looking for a relevant development was an hour not spent considering its implications. The risk team needed a way to reduce the noise without narrowing the view so far that an important signal could be missed.

Building RiskLive around the work

RiskLive emerged from a wider programme to modernise how risk is identified, captured and analysed. Rather than asking risk analysts to adapt their work to a generic search tool, the project focused on bringing relevant external developments into a dashboard that could become part of the team’s daily routine.

Working with a university delivery partner, the project team integrated Valyu’s API into RiskLive. The service supports real-time monitoring and allows results to be filtered around the regions, events and developments that are most relevant to the Authority’s work.

The integration was completed in a matter of days. Technical support was used to tune the returned results and ensure that the service could operate under the demands of the environment. RiskLive then moved into daily use by the operations team.

A capability that needs to stay present

For a monitoring service, silence is ambiguous. It may mean that nothing significant has happened, or it may mean that the flow of information has stopped. That makes reliability part of the operational requirement, not simply a technical measure.

RiskLive was therefore designed to run continuously and to give users confidence that the dashboard is current. The underlying search service is intended to support sustained query volumes, regional filtering and consistent retrieval without requiring the team to build and maintain a separate search stack.

What changes for the analysts

The practical change is a shift in where effort is spent. Instead of beginning with a wide manual search, analysts can start with a more focused stream of potentially relevant developments. Human judgement remains central: the system brings information into view, while analysts decide what it means and whether action is needed.

The emerging value lies in earlier awareness, greater consistency and more time for assessment. The scale of that value still needs to be evidenced through usage and operational measures.

“We searched for an easy-to-use API for our RiskLive platform and Valyu met the brief perfectly.”

— Carl Dalby, Group Head of AI & Digital, Nuclear Decommissioning Authority

Public value, told through evidence

The case for RiskLive is not that technology replaces expertise. It is that reliable retrieval can remove some of the repetitive effort that sits in front of expert assessment. If the early operational evidence supports that story, the benefits can be expressed in practical terms: time returned to analysts, faster awareness of relevant developments and less engineering effort devoted to maintaining search infrastructure.

The next chapter

RiskLive is now a working example of how a focused digital capability can support situational awareness in a complex operating environment. The next step is to learn from live use: which sources add the most value, where coverage should expand and how the capability might connect with the wider AI and risk programme.

Those decisions should follow the evidence. The story is still developing, but its direction is clear: help people see what matters sooner, then give them more time to apply the judgement that only they can provide.