Cases
Recent work.
Client names are left out on purpose. The systems marked 'our own operation' run this business every day.
Audit · client engagement
Finding a pricing leak
A property services business suspected it was charging below its own rate card. Nobody had checked. We audited 4,213 booking records against the card, kept only the matches we could stand behind, and confirmed recurring undercharging across 58 cases, much of it repeating on the same property for 14 months. The data hinted at a far larger figure, and we reported that one as unverified because most of it depended on low-confidence matching. The client got numbers that hold up, tied to the exact bookings behind them.
4,213 records audited, 58 confirmed cases
Automation · client engagement
Briefs that arrive complete
A global design team was typing the same brief into three separate systems, and incomplete briefs were only caught days into a job. One real card sat open for 164 days and collected 40 back-and-forth comments. IT had locked the toolset, so new software was off the table. We built a five-step intake form inside the project board the company already pays for, with a quality gate that checks each field as it's filled and blocks placeholder answers at the door. One submission now creates the full job card, the board files it automatically, and the tracking spreadsheet writes itself. Tested end to end on the team's real briefs.
Five-step intake with a live quality gate
Productivity · our own operation
AI bills, under control
AI bills grow when expensive models do cheap work. Our own operation runs on a router that sends every mechanical task, summarising, classifying, extracting, drafting, to free AI models on our own machine first, and saves the paid models for judgement calls. Every call is logged: 519 of them across the first nine days, including a 72,000-character job handled entirely on local models. At one measured checkpoint, roughly 320,000 tokens of work had been kept off the paid meter. A budget script reads the log and flags anything that should have been free but ran on a paid model.
519 calls logged across three local model tiers
Productivity · our own operation
The owner's operating system
Prompted.au is a one-consultant practice, and the operation sits on a system we built for ourselves. Anything dropped into a folder, a voice note, a screenshot, a PDF, a video link, comes back as a structured task with dates and dollar amounts attached, waiting for one-click approval. One live file holds every open item and deadline, and the daily briefing is assembled by code, so urgent items can't quietly fall off the list. A local assistant answers questions like 'what did I do last Tuesday?' straight from the file archive, and a phone channel catches ideas away from the desk. It runs on AI models on our own machine, works with the wifi off, and writes to source files only after a human approves.
Capture to approval, running offline
The standard
Every number here survived verification.
Cases are published without client names as a rule, and nothing appears on this page that records can't back. Where a finding was too weak to stand behind, it's described that way.
Your numbers next.
The audit finds what these cases found: the places in your business where the hours and dollars are leaking, with the evidence attached.