DOC QW-METHOD-01 / FIELD MANUAL

03 / The method

One person · one week · one win

The method is deliberately small. It finds the AI already sitting inside the tools you work in, puts the candidates in order, and stops. What it produces is a list you can act on before lunch, not a program.

Step 01 / Map your stack

Write down the tools you personally open during a normal week. Not the company's software inventory, and not what you wish you used. Email, calendar, meetings, docs, sheets, chat, and whatever system your own function runs on. Six to ten names is typical.

Step 02 / Find what is already there

Look each one up in the directory. Every app carries one of four answers about built-in AI: it ships AI you can switch on yourself, it ships something narrow or gated, it ships nothing a normal user can reach, or nobody has checked it yet. That last answer is reported honestly rather than guessed.

175Ship usable AI
83Checked against docs
34Can act unprompted
19Median actions per app

Only 34 of 589 apps can act on their own. In every other app, a person has to ask. Treat always-on automation as the exception, not the plan.

Step 03 / Run one win

Pick one. Not a plan, not a rollout, one feature in one tool on one real task you were going to do anyway. The right first win is the one with the shortest path between reading about it and using it, which is usually summarizing something long or drafting something routine.

Step 04 / Log it, then repeat

Time the task before and after, roughly. Any estimate you read anywhere, including here, is a model of a typical person at typical volume rather than a measurement of you. Your own number is the only one that decides whether a win is worth keeping. Keep what pays, drop what does not, and add the next one next week.

What this method refuses to doGuardrails
  • No score, grade, or readiness level. You are not being ranked against anyone.
  • No recommendation that requires a new purchase, a developer, or an IT ticket.
  • No claim that a saving has been measured when it has only been estimated.
  • No promise that more automation is better. The right level of delegation is per task.
About the dataSnapshot 2026-08-23
Directory589 apps across 35 categories, holding 26,826 distinct actions.
Withheld411 catalog entries are developer infrastructure and get no page, because a directory of thin pages costs more than it earns.
LimitsThe catalog listing caps at 1000 results with no cursor and returned exactly 1000, so this is the reachable set and very likely truncated. Say 'apps in this catalog', never 'all apps'.
Start at step two. You already know your stack. Tick it off and get the list.
Build your blueprint