Welcome
This is where you show the firm where your work actually gets stuck. You know your work better than any tool does. That is why this starts with you.
The promise
- Your entries stay with you. Everything you type is saved in this browser, on this computer. Nothing goes anywhere until you save your file and send it to the pilot coordinator.
- This measures the work, not the people. It is not used to evaluate anyone.
- The firm decides what, if anything, gets built. AI does tasks. People make the decisions.
- The tool is yours. The Aubrey Group keeps it when the pilot is over.
First, who are you?
This puts your name on your file, so the pilot coordinator can fit it together with everyone else's.
How it works
- Fill in the stage that is due. It saves by itself as you type.
- When the stage is due, press Save my file to send at the bottom of the screen.
- Email that file to the pilot coordinator. Send the file itself, so everyone's entries fit together in one record.
The four stages
The due dates are set by the pilot coordinator. Type them in here once you have them.
| Stage | What happens | Who does it | When in the pilot | Due |
|---|---|---|---|---|
| 1 | List your tasks The work you do again and again, step by step, with the steps that break or drag flagged. | Everyone | Days 1–30 | |
| 2 | Time the real thing Each time you do a top task, log how long it really took. | Whoever does the task. Keep logging through the test window. | Days 1–60 | |
| 3 | Decide what to do about each task Five possible answers. Automating is only one of them. | Pilot coordinator and leadership | Days 31–60 | |
| 4 | See what it does to pricing What the changes mean for each service the firm sells. | Leadership | Days 61–90 |
Closing workshop: week of December 14. The task list, the time log and the pricing picture should be in hand a week before it.
Your tasks
List the work you do again and again. Break each task into its steps and flag any step where the work breaks or drags: the slow and tedious parts count as much as the broken ones. You know where; that is the point. The numbers are a quick guess to go with it; Stage 2 times the real thing.
Where your work breaks or drags
Every step you flagged, counted by kind of problem. This is what gets reviewed for duplicate work, waiting, and waste.
Where your time goes
Worked out from your guesses, biggest time cost first. Lost work is counted in: a task where work is lost or started over one time in four costs a quarter more than it looks.
Time log
Now the real thing. Each time you do one of your tasks, add one entry: say what happened, how you did it and how it went, then split the minutes into five parts. Your account of what happened is what explains the numbers.
Your guess against the real time
What you estimated in Stage 1, next to what you actually logged. Further down, the same times are sorted by how the work was done: the usual way, with an AI tool, with a standard process, or with a template. Learning time is left out of the averages and counted separately.
Decide what to do about each task
One decision for each task, biggest time cost first. There are five possible answers: eliminate it, standardize it, template it, automate it, or leave it as it is. Automating is one of the five, and “leave as-is” is a perfectly good answer.
How to decide
The five answers, in this order
- Eliminate. Stop doing it.
- Standardize. Do it one agreed way.
- Template. Start from a fixed form, not a blank page.
- Automate. Let software or AI do part of it.
- Leave as-is. It works, or it is not worth changing.
Try the first three before the fourth, so AI is not put on a step that a simpler fix would solve.
AI is suggested only when all five are true
- The step is drafting, summarizing, reformatting, extracting or comparing, using material the firm already has.
- A named person reviews the result before it leaves the firm.
- The material is allowed under the pilot's rules. Today that means public information only.
- A wrong answer would be caught in the firm's review, not by the client or the funder.
- The time saved is more than the review time it adds.
AI is not suggested for a judgment call, for a step where the source material is missing, or for a step whose real problem is waiting, duplication or an unclear process.
These rules are the firm's to change. The AI review on the Insights page is given the same rules, and it is advice: the decision on every task is made here, by people.
Tools the firm is considering
Before anyone uses a tool, it needs five things written down: who owns the account, the most sensitive material it may touch, what the vendor keeps, who checks its output, and how far it is allowed to go.
What this does to pricing
For each service the firm sells: the hours an engagement takes today, and the hours you expect once the Stage 3 decisions are in place. Prices stay as they are today, so you can see what the change does to revenue.
What happens to revenue
Work billed by the hour loses revenue when it takes fewer hours. Flat-fee work keeps its revenue and gains margin. The mix is what matters.
Insights
Everything in this file on one page: where the work breaks or drags, where the time goes, how it went by method, and what people said, in their own words. The tool only counts and sorts. Nothing here was written by AI.
What came back
Paste the AI's review here. It is saved with this file and travels with it.
Where things stand
What is in this file so far.
Version history
| Version | Date | What changed |
|---|---|---|
| Base 1.0.0 | 2026-09-14 | First release. Four stages, one shared file format, save and combine. |
| Aubrey 1.0.0 | 2026-09-18 | Built for The Aubrey Group on the 90-day pilot clock. Area of work, client label, AI-assisted and learning flags, named reviewer and authority limit, the firm's service tiers, editable due dates, names kept when files are combined, whole-firm view. |
| Aubrey 1.1.0 | 2026-10-04 | Rebuilt for ease of use. Welcome page with the promise and the Dumb on Purpose stamp, each task broken into steps that staff flag where the work breaks (duplicate work, waiting, redoing, extra handoffs, retyping), a Where the work breaks view that counts those flags, an Insights page that gathers everything in one printable place, a Clear button on every section and a Clear everything button (each asks whether to save the record first), a Copy for AI review button that carries the method and the rules with the entries, room to explain each task and each time it is done in their own words, plain wording throughout, an optional one-question-at-a-time guide, only the stages each person fills in, due dates left open for the pilot coordinator. Every 1.0.0 field kept and the same file format, so earlier files still open and combine. |
| Aubrey 1.2.0 | 2026-10-05 | Changes from the firm's review of 1.1.0. The redo question now asks only about work that was lost or had to be started over, not normal revision. Five more kinds of problem a step can be flagged with (repetitive or tedious, information not available, no access to a subject matter expert, client communication breakdown, not enough time to do it to standard), and “breaks” became “breaks or drags.” Research and Written content production are separate kinds of work. Each logged time records how the work was done (the usual way, with an AI tool, with a standard process, with a template) and a satisfaction rating, compared by method on the Time log and Insights pages. The rules for deciding are shown on the Decisions page. The AI review follows the pilot's public-information-only rule and shows which rule each recommendation rests on. Staff are asked to write “a client” in place of a client's name. Every earlier field kept, so earlier files still open and combine. |
The whole firm
Everything the staff sent in, put together. This page only reads; nothing on it can be edited.
Who has sent what
Where the work breaks or drags
Every step the staff flagged, across the firm, counted by kind of problem. Read this for duplicate work, waiting, and waste.