We help UK consumers recover money lost through mis-sold financial products. Our work includes managing claims, processing financial documentation and supporting customers throughout the claims process.
We operate at scale, with legal and operations teams working across multiple locations. Technology plays a central role in how we work, with AI already being used to process incoming correspondence, interpret documents and streamline operational workflows. We are looking for someone to take ownership of this capability and help develop it further.
The role:
You will own the business's AI and automation capabilities, overseeing the models, prompts and workflows that process incoming documents, interpret information and determine the appropriate next steps.
These systems are central to day-to-day operations. Their accuracy, reliability and performance directly affect how efficiently cases progress and how quickly customers receive updates. You will be responsible for maintaining and improving these systems, ensuring they are robust, well documented and continually evolving. You will also identify new opportunities to introduce AI and automation across the business.
You will work closely with the CTO and development team, who are responsible for the wider technology platform and its ongoing development. Your focus will be on the AI and automation layer, ensuring its outputs integrate effectively with the wider platform and deliver tangible operational improvements.
Key responsibilities
- Own the end-to-end pipeline that reads, classifies and routes every lender response: acknowledgements, holding letters, final responses, offers, rejections and DSAR data packs.
- Maintain the categories, rules and confidence thresholds that decide what goes straight through and what goes to a person.
- Investigate misclassifications quickly and fix the root cause, not just the case in front of you.
Postal scan categorisation
- Run the OCR and document classification pipeline for scanned post: document type, client, lender and claim matching.
- Improve extraction on poor scans, handwriting and multi-page bundles.
- Keep unmatched and low-confidence items in a clear exception queue, cleared within an agreed turnaround.
Prompts, models and evaluation
- Own the prompt library: version control, documentation, change history and review for every production prompt.
- Build and maintain test sets of anonymised real documents, so every prompt or model change is measured before it ships.
- Monitor accuracy, drift, latency and cost, and choose and tune models against those measures.
- Manage provider dependencies, rate limits and fallbacks, so the pipeline keeps running when a provider has an outage.
Automation and integration
- Build and maintain the automations that move documents and data between the AI layer, the case management system and the client portal.
- Work with the development team so AI outputs (claim status, document type, next step) show correctly in the client portal.
- Spot the next manual processes worth automating and build the case for them with operations.
Governance and compliance
- Make sure AI processing of client data meets UK GDPR, our data protection policies and the standards expected of a regulated legal and claims business.
- Keep a clear audit trail of what each model decided, on what input and with which prompt version.
- Write and maintain runbooks, so no system depends on one person's knowledge.
Working with the business
- Turn needs from the legal and claims teams into system changes, and explain AI behaviour in plain English.
- Report to the CTO on performance, incidents and the roadmap.
Skills and experience
- 3+ years building production software, including at least 2 year shipping LLM or machine learning systems that real users rely on.
- Strong prompt engineering backed by evaluation: you can show how you measured a prompt and improved it, not just that you wrote it.
- Hands-on document AI: OCR, document classification and data extraction (for example Azure Document Intelligence, AWS Textract, Google Document AI or open-source tools).
- Python or TypeScript, REST APIs and SQL.
- Workflow automation and integration (for example Power Automate, Make, n8n, or custom queues and webhooks).
- Experience with major LLM providers' APIs (for example OpenAI, Anthropic or Azure OpenAI), including cost and rate-limit management.
- A track record of picking up an existing, partly documented system and becoming its owner fast.
- Working knowledge of UK GDPR and of handling sensitive personal and financial data.
- Clear communication with non-technical colleagues.