Valliance logo in black
Valliance logo in black

Case study

UK Healthcare Services Provider

Case study

UK Healthcare Services Provider

UK Healthcare Services Provider

Case study

UK Healthcare Services Provider

Case study

UK Healthcare Services Provider

UK Healthcare Services Provider

Building agentic workflows to accelerate growth and accuracy for a leading UK health provider.

Agentic AI in Healthcare showing an MRI scanner
Agentic AI in Healthcare showing an MRI scanner
A leading UK health provider, that delivers outsourced and insourced services has the ambition to scale, but was impeded by a fundamental tension:

Growing the client base of supported clinics required enterprise-scale technology to manage significantly higher data volumes, yet today's business was not large enough to justify that investment.

In the meantime, their insourcing team remained tied up in manual data compilation, transfer, and validation processes, leaving little capacity to support new contracts.

We've built an agentic system spanning the entire clinical outcomes validation and reconciliation workflow, delivering targeted interventions throughout.

  • Automated verification of inbound emails from clinics. Data with issues are now returned to medical secretaries with detailed instructions on how to remedy the issues.

  • Deployed a guard-railed LLM to manage reasoning, exceptions, and ambiguities in the validation rules

  • Automated the compilation, transfer, and validation of data

  • Identified that adding a single data field would eliminate an entirely separate manual process, which was implemented within days

The inbound data was semi-structured.

Clinic emails arrived in varied formats, with edge cases that would have made traditional automation brittle. An agentic approach gave us a way to reason over the messiness, while keeping deterministic checks where they earned their keep.

Evals and test automation kept regression in check.

We ran representative-case evals against the model alongside integration tests for the wider workflow. Thorough QA gave us confidence in what the system was doing at every step.

Observability mattered more than we expected.

Traces across the multi-step orchestration let us see exactly how each step behaved, both functionally and on latency and cost. It surfaced things we'd never have caught from outputs alone.

The inbound data was too varied for fixed rules.

Safety design was built around Simon Willison's 'lethal trifecta'. Access to private data, exposure to untrusted content, and the ability to communicate externally. Inbound emails tick all three. We designed the system so the agent never holds them at once.

Valliance North Star in Orange
Agentic workflows to support UK healthcare - image depicting two caring hands together
Agentic workflows to support UK healthcare - image depicting two caring hands together
50%
50%
50%

Reduction in the time spent by the healthcare administration team on client contracts.

  • Increased data quality and integrity by reducing the human error risk

  • Unlocked the capacity for the client to grow its business by taking on new contract and increased the confidence of their new potential clients in the quality of service

Related insights
Related insights
Related insights
Related insights