Foundations & security
Audited the whole GitHub org, locked down secrets, and added an automated pre-push gate that blocks credentials before they ever reach a public page. Set the public/private folder discipline the team still works from.
Webprofits
Taking a performance-marketing agency all-in on Claude, safely and at speed.
Webprofits is one of Australia's best-known performance-marketing agencies. Over a three-month weekly retainer, we helped them move from a brave company-wide bet on Claude to a secure, low-cost, partly autonomous way of working that the whole team shares.
A leadership team that decided to change everything at once.
Webprofits runs ecommerce and lead-gen for Australian brands across paid social, paid search, email, SEO and CRO. Co-founder and Director Alex Cleanthous made a hard call: all work moves into Claude, for everyone, no exceptions.
It was the right bet, but a fast one. The setup was growing quicker than it could be hardened, costs were climbing, and the gap between the people who had truly adopted the new way of working and those still catching up was widening. We worked week to week with Alex and with Chief Strategy Officer Ben Fitzpatrick, who drove AI enablement and the Academy across the team, to check the work, close the gaps, and keep them moving without breaking things.
The challenge
Going all-in on Claude is the easy part. Doing it without leaking secrets, burning budget, or leaving half the team behind is the hard part.
- A GitHub org expanding faster than it was being secured.
- Token costs climbing with no clear levers to pull.
- Skills shared by pasting files into chat: no versions, no single source of truth.
- Validated work still run by hand, one skill at a time, with someone watching.
- A real productivity gap between early adopters and the rest of the team.
One hour a week, compounding.
Every week we prepared, reviewed the work, and agreed the next move. The arc went from securing the ground they stood on to handing repeatable work over to routines that run on their own.
Cost & efficiency
Turned runaway token spend into a small set of habits (skill scoping, model selection, session hygiene, output brevity) so the team gets the same work for a fraction of the cost.
Memory that compounds
Standardised one simple Supabase memory layer across the team: every session writes what it learned, every session starts with the relevant context. Knowledge stops living in one person's head.
Sub-agents
Refactored their most expensive skills to fan work out to parallel sub-agents on cheaper models, returning only the summary to the main session: same output, far lower cost per run.
Routines & autonomy
Moved validated skills off the keyboard and into scheduled and event-driven routines (local for cross-repo work, remote for fire-and-forget) chained end to end so a pipeline runs itself.
Distribution
Packaged skills into versioned plugins so one command updates everyone. No more copy-pasting files in Slack, no more drift between team members.
Faster, cheaper, safer, and shared.
From operator to orchestrator
A sales pipeline that used to be four separate manual runs (portal, enrichment, style guide, proposal) now runs as a single chained routine. The work happens without anyone watching it.
Output the whole team can feel
The people fully inside the new way of working went from a handful of creative concepts a week to thirty-plus per run, with well over a hundred shipped across three markets in days.
A safer place to move fast
Secrets out of the codebase, an automated gate that stops them coming back, and a daily security routine watching database access. Speed without the exposure.
Knowledge that stays
A shared memory layer and a plugin system mean what one person learns is available to everyone the next morning, and survives anyone moving on.
“The biggest change we've ever made, and it's such a no-brainer.”
Thinking about going all-in on Claude?
Whether you're just starting or already moving fast, we'll tell you where the gaps are and how to close them. The first call is free.