How the work runs
Four steps, in that order, every time. The first one is the one most projects skip, and it is the reason the other three go wrong.
-
01
Diagnose where work breaks
Before anything is built, the operation is mapped as it actually runs — not as the org chart says it runs.
Where work enters, who has to remember what, where things wait, what nobody can see, and what breaks when volume doubles. The public website gets a structural read as part of this.
-
02
Design the operational system
The smallest system that removes the worst leaks first.
Which pieces are public-facing, which run in the back office, what fires automatically, and what stays human. Every module is tied to a failure point from the diagnosis.
-
03
Build and integrate it
Built in the tools you already pay for wherever possible, replaced only where they genuinely fail.
Working software in front of the people who will use it every week, not a reveal at the end. Integrations and automations are wired as the modules land.
-
04
Launch, observe, and improve it
A system is only finished once it is the way work gets done.
Launch with the team, watch where it is bypassed, and fix the reasons. Metrics from the diagnosis are re-measured against the new baseline.
Five questions I ask of every operation I am brought into.
The website diagnostic runs a structural version of this on your public site in about a minute. It is a useful first read, not a substitute for the full week.
What this is not
- Not a redesign of your website with a form on the end.
- Not a vendor tool you rent and reshape your team around.
- Not a reveal at the end. You see working software every week from step three.
- Not finished at launch. It is finished when it is how work gets done.