WORK / SELECTED CASES

The work, measured.

Real engagements, anonymised at the client's request. Numbers are pulled directly from production telemetry. Sectors and regions are accurate; names and pricing are withheld by agreement.

27 SYSTEMS DEPLOYED14 AVG DAYS TO LIVE0 RETAINERS SIGNED
CASE #04 · SAAS · ANONYMISED

Onboarding rebuilt in 17 days. Time-to-live: 8 days → same afternoon.

SECTOR · B2B SAAS
REGION · UNITED KINGDOM
TIER · SPRINT
DURATION · 17 DAYS

A mid-market B2B SaaS was losing deals at the finish line. Customers signed, then waited 8 business days to get a working account. Sales blamed Ops. Ops blamed Engineering. Nobody owned the path from contract to live.

RESULT — MEASURED
TIME-TO-LIVE8 business daysSame afternoon
MANUAL TOUCHES PER ONBOARDING232 (exceptions only)
CS HOURS PER NEW CUSTOMER6.4 hrs0.5 hrs
SITUATION — BEFORE
  • ×Manual account creation across 4 internal systems
  • ×CSV migration handled by a single CS lead
  • ×Kickoff scheduling that required 6 emails minimum
  • ×Welcome sequence triggered by a person, not a system
SYSTEM — AFTER
  • Contract-signed webhook into a workflow that provisions all 4 systems in parallel
  • Migration agent ingests the customer's data, validates it, surfaces exceptions
  • Calendar handshake fires the moment provisioning is green
  • Customer lands in their account by end of business the same day
CASE #09 · INSURANCE · ANONYMISED

Underwriting data entry: 4.1% error rate → 0.02% in three weeks.

SECTOR · INSURANCE BROKER
REGION · SOUTH AFRICA
TIER · SPRINT
DURATION · 21 DAYS

An insurance broker had been "going to fix" a manual data-entry problem for two years. Policy details were re-typed from PDF applications into the underwriting system. A transposed digit had just cost them a R200,000 claim shortfall. The committee meetings had stopped helping.

RESULT — MEASURED
ERROR RATE4.1%0.02%
TIME PER POLICY11 mins40 secs (auto) / 3 mins (exceptions)
AUDIT TRAIL COVERAGENone100% (input → output)
SITUATION — BEFORE
  • ×Application PDFs received by email, varying formats
  • ×Re-typed by hand into a legacy underwriting platform
  • ×Errors caught only at claim time — sometimes never
  • ×No audit trail tying policy back to original application
SYSTEM — AFTER
  • PDF ingestion that extracts every field with confidence scoring
  • Validation layer that flags low-confidence values for human review
  • Direct write into the underwriting platform, with audit linkback
  • Exception queue replaces the data-entry queue — humans handle judgment, not typing

MORE CASE STUDIES UNDER NDA · AVAILABLE ON REQUEST

See your problem in one of these? Let's talk.

READY?

Book a 15-minute call and start scoping.

No pre-call form. No pitch deck. Honest answer in 15 minutes; fixed-price proposal in 48.