[01]
A century of manual process. Zero AI.
A century-old mutual, deep manual processes, decades of legacy code — and no AI capability to draw on.
Our client is a member-owned UK building society, regulated by the FCA and PRA. Like many established financial institutions, much of its day-to-day work runs on deep manual processes and legacy systems few people fully understand — and it had no in-house AI capability or tooling when we arrived.
Rather than dropping in a black-box product, we embedded with their teams: exploratory sessions and hands-on workshops to find high-value, low-risk places where AI can take the weight out of manual work, then co-building internal tools that run entirely on the client’s own secure cloud. Security, governance and knowledge transfer came first — as they must, for a regulated, member-owned institution.
[02]
Prove value first. Then build.
From conversation to working tools through a repeatable, low-risk cycle — proving value before committing to build.
Exploratory sessions mapped the manual, legacy-bound workflows that cost the most time. Hands-on workshops turned them into candidate use cases, prioritised by business value and delivery risk. Rapid proofs of concept demonstrated value on real material before any large commitment — then we co-built alongside their people and shipped into their environment, feeding a growing backlog of further use cases.
Discover › Workshop › Prototype › Co-build › Deploy & iterate
Everything runs inside the client’s own secure cloud. Sensitive data never leaves their estate. The full cycle is described under Approach.
[03]
Three tools, three honest statuses
Three use cases, each discovered with their teams, prototyped to prove value, then engineered for their infrastructure — with honest statuses.
01 · Code Flow Analyzer
live- Challenge
- Decades of business logic live in legacy SAS programs and SAP data-flow exports that few people fully understand — a risk for audit, change and modernisation.
- Delivered
- A browser-based tool that parses SAS and SAP export files, builds visual data-lineage maps across sources, transforms and targets, and surfaces calculation evidence step by step — down to column-level lineage, with optional AI-assisted checks.
- Value
- Faster, safer understanding of legacy regulatory-calculation code; documentation and audit evidence on demand; a de-risked path to modernisation.
02 · Sanctions & watchlist screening
in design- Challenge
- Members and applicants must be screened against government sanctions and prohibited-persons lists — today a slow, manual, error-prone task.
- Delivered
- An automated name-matching workflow that screens records against government watchlists, scores and flags likely matches for human review, and keeps a full audit trail — designed for the AML and KYC obligations of a regulated lender.
- Value
- Faster, more consistent screening, fewer missed matches, a clean audit trail, and a lighter manual review load.
03 · Contract Analyzer
specified- Challenge
- Contract review is slow and manual; key terms, obligations, dates and risk clauses are buried across long documents.
- Delivered
- A tool that extracts and summarises key clauses, obligations, dates and risk terms to accelerate review. Scoped and prototyped during the engagement; ready to be reactivated when prioritised.
- Value
- Faster contract triage and materially lower review effort.
[04]
The capability stays in-house
Built with them, not just for them.
Knowledge transfer is part of the deliverable. The client keeps a standing pipeline of identified, prioritised use cases — and the in-house understanding to run with them. The capability stays inside the organisation, not locked in a vendor. Discovery and workshops continue to surface new candidates; the backlog of manual workflows ready for AI keeps growing.
[05]
From first tools to a broader footprint
With a first AI capability established and two proofs of concept behind us, the engagement is completing the screening workflow, expanding the code-analysis tooling, and working through a prioritised backlog of further manual workflows — all delivered the same way: on the client’s infrastructure, hand-in-hand with their teams.
Measurement basis
- Basis
- Capability-level: what was stood up, built and scoped. This engagement is measured in capability, not throughput.
- Environment
- The client’s own secure cloud. Sensitive data never left their estate.
- Artifacts
- One live analysis tool, one workflow in design, one specified prototype, workshop outputs, and a prioritised use-case backlog — plus the knowledge transfer to run them.
- Not published
- Client name, timelines and internal data. Baselines and timeframes are shared in conversation, with the client’s approval.