Case studies

Automation work, with the numbers attached

Anonymised summaries of recent engagements. Client references are available on request during scoping.

B2B SaaS · 40 staff

Tier-1 support agent handling 62% of tickets

  • 62% of tickets resolved without a human
  • First response time 4h → 40s
  • Deployed in 11 business days

The problem

A support team of six was spending most of the day on repeat questions about billing and onboarding.

What we built

We built a retrieval-based AI agent over their help centre and Stripe billing data, with strict escalation rules to a human on anything account-changing.

Logistics · 120 staff

Invoice intake automated end to end

  • ~30 hours per week returned to the team
  • Data entry errors down 91%
  • Full audit trail per invoice

The problem

Supplier invoices arrived as PDFs by email and were rekeyed into the ERP by two full-time staff.

What we built

A document parsing pipeline extracts line items, validates them against purchase orders, and pushes clean records into the ERP with exceptions routed to a review queue.

E-commerce · 25 staff

Demand forecasting for reorder planning

  • Stockouts reduced 38%
  • Working capital freed in slow SKUs
  • Weekly forecast delivered automatically

The problem

Stock decisions were made from spreadsheets, causing both stockouts and dead inventory.

What we built

We consolidated sales, marketing spend, and supplier lead times into a weekly forecasting pipeline with a simple dashboard for the buying team.