Case studies

The demos prove what I can build fast and cheaply. These prove I ship serious work and have done credible AI builds before. Numbers are concrete where they can be shared, and percentages where they're confidential.

Agentic database enrichment

Vagabod

4,441 gyms58 countries enriched

The situation. Vagabod helps travellers find casual gym access anywhere in the world — but a directory is only as good as its data, and gym pricing and facilities are scattered, inconsistent and constantly changing.

What I built

  • Automated scraping + enrichment pipeline
  • Casual-membership pricing extraction per gym
  • Outreach pipeline so owners can claim listings
  • Continuous re-enrichment as data drifts

Outcome

  • 4,441 gyms across 58 countries enriched
  • Directory kept fresh without manual entry
  • Self-service claim flow for gym owners

Why it matters

  • The same agentic enrichment applies to any large, messy dataset
  • Reactivation & lead-gen lists become assets, not liabilities
Visit Vagabod
Document ingestion · Big-4 (anonymised)

Transaction Management Platform

~80%less manual handling

The situation. A transaction-management platform at a Big-4 firm handled a steady flow of complex deal documents that scarce, expensive people were processing by hand — slow, error-prone, and hard to scale across engagements.

What I built

  • Document ingestion + structured extraction engine
  • Validation rules with anomaly flagging
  • Routing into downstream systems
  • Eval suite + monitoring for accuracy

Outcome

  • ~80% reduction in manual handling
  • Faster, more consistent processing
  • Audit-friendly evidence trail

Why it matters

  • Any firm with contracts, engagement letters or working papers has this exact problem
  • It's the unglamorous workflow that quietly saves a day a week

Client anonymised; metrics shown as a percentage to protect confidential detail.

Proposal / RFP response agent

RFP Responder

1st draftin minutes, not days

The situation. Responding to inbound proposals and RFPs eats senior time: re-reading the brief, hunting for the right boilerplate, and assembling a first draft from scratch every time — when most of the answer already exists in past work.

What I built

  • Agent that reads a brief and a template library
  • Drafts a structured first-cut response
  • Pulls the right past answers per question
  • Hands a clean draft to a human to finish

Outcome

  • First draft in minutes instead of days
  • More proposals answered, none dropped
  • Consistent, on-brand language

Why it matters

  • It's the Document Processing + Follow-up demos applied to winning work
  • Every services firm with an inbound pipeline benefits

Reference build on synthetic-but-realistic data, from a well-subscribed n8n template — a polished example of the pattern.

More case studies land as pilots complete. Want to be the next one — built free during launch?

Your workflow could be the next case study

Book a free Audit and I'll show you a working agent on your data. If it's a fit, your first pilot is built free during launch.

Book a free Automation Audit