Kreoh’s Discovery Agent identified qualifying R&D projects from The Code Registry’s internal codebases and documentation for their first UK R&D tax claim, and collected sufficient information to evidence and draft most of the case studies. All feedback is from Catalyst (catalystrd.co.uk), the independent consultancy that ran the claim.

Summary

The Code Registry (TCR), a UK software company, was preparing its first R&D tax claim. Kreoh’s Discovery Agent, an AI securely grounded on a company’s internal codebases and documentation, identified the qualifying R&D projects and collected sufficient information to evidence and draft the technical write-ups, every point cited back to its source. Catalyst reviewed the output with TCR’s lead developer and built the live claim from it. This was a live UK claim, not a demonstration.

The problem it solves

Many R&D claims hit the same bottleneck: working out which projects qualify, then extracting enough technical detail from busy engineers to write each one up defensibly. That means interviews, follow-up calls, and expensive expert time spent reconstructing work that already happened and was already recorded in the company’s own systems.

The Discovery Agent works from that record directly. It surfaces qualifying R&D projects, investigates specific ones, and answers questions about the technical work.

Results

Completeness. The agent found every qualifying R&D project. Of the 15 candidate projects it surfaced, expert review resolved 5 R&D projects (four standalone, plus five related workstreams which formed one R&D project), When Catalyst went through the work with The Code Registry’s competent professional (CP), no projects came up that it had missed. It also pulled enough from the company’s code and documents to enable the writing of each project in full after just one call with the CP. Catalyst called the detailed output “brilliant” for that stage:

“…from what it’s pulled out we have more than enough to actually draft the case studies itself.”

Time. The agent gathers technical detail from the company’s own records, removing the most interview-heavy phase. In the words of the consultant who ran the claim:

“We’ve saved pretty much all the time on the technical information gathering… up to 45 minutes per case study, plus lots of the CP’s time… at least a couple of hours, maybe more.”

Defensibility. Every statement is grounded in the company’s own records, traceable to a specific timestamped source, which matters as HMRC sharpens its scrutiny of how claims are evidenced. Catalyst’s technical specialist rated the write-ups strongly:

“Good descriptions of R&D, good search of similar things to contextualise.”

The result, in Catalyst’s view, is a more defensible claim:

“The level of evidence… means that they are significantly more defensible and… more evidence based than some claims would end up being.”

How the advantage scales

TCR is a small company, where the CEO has visibility over most of the R&D work themselves. Catalyst’s view is that the agent’s value grows with company size. At a larger organisation no one person can see across all the R&D, and the leadership project conversation becomes, in Catalyst’s words, “almost just to come up with a list of potential projects.”