The problem was success
Sixteen people doing WordPress and Laravel work for trade bodies and membership organisations. Good reputation, strong pipeline, permanently full.
That last part was the trouble. Their managing director kept a spreadsheet tab labelled "declined" with four entries on it from the previous year. Combined value: just over £310,000. Every one of them was work the agency could do well and simply had nowhere to put.
The alternative was contractors. They'd tried it. Two senior Laravel freelancers at £575 a day, a fortnight of onboarding before either wrote anything useful, and a codebase that needed a rewrite of the queue layer once they left. Margin on that project finished at 11%.
So when a national trade association came back asking for a members' portal (SSO, tiered access, an events engine, invoicing tied to membership renewals) the MD was about to add a fifth row to the spreadsheet.
"Honestly, it sounded like a liability"
Her words. An AI-native studio building client work under her agency's name was not a comfortable proposition for someone whose business runs on reputation.
They tested us on something disposable first: an internal timesheet reconciliation tool, four days, £2,900. Their lead developer read every line of the output. Then they came back with the real job.
How it ran
The spec sprint took nine working days and cost £4,800. The agency's project lead sat in two of the three sessions; the association's operations manager sat in the third. What came out was 140-odd pages of specification, a clickable prototype of every screen, and a build decomposition of eleven milestones with a price against each.
The agency owns that specification outright. It's theirs whether they build with us or not, which is the point.
The build was fixed at £38,400. Agents ran the milestones on an isolated VM; every milestone landed on a staging server the association's team could poke at the same afternoon. The agency reviewed at each gate. Two milestones came back for rework, both on invoicing edge cases nobody had thought about until they saw it running.
Six weeks from kickoff to handover.
What it did to the numbers
| Their usual approach | With Triophase | |
|---|---|---|
| Delivery time | ~5 months | 6 weeks |
| Delivered cost | £73,000 (contractors + internal) | £43,200 |
| Gross margin | 24% | 58% |
| Defects reported in first 60 days | 15 (agency average) | 2 |
| Automated test coverage at handover | not measured | 94% |
They took on three more projects that quarter that would otherwise have been declined. Additional recognised revenue: £212,000.
The defect number matters more to the MD than the margin. "We've had post-launch periods that cost us the profit twice over. Two tickets in two months, both cosmetic, is not a thing that happens to us."