Start here

Each person sets up their own AI-Labs folder, has Claude interview them for their own AGENTS.md, sets the model-training setting, and gets one small piece of real writing checked and ready to send, from their own notes or the practice email thread.

The Brief's surprise: Claude is more capable than you think and less trustworthy than it sounds. In step 8, each person has Claude read their AGENTS.md as an outsider would, then the room hears one question from each. Bring to Lab 2: the mission, two writing samples and an outcomes summary if there is one (or a voice note transcript, or the practice pack), a live RFP if there is one, and the workflow's name with its old-way time from a stopwatch.

Each person builds their org brain as plain files (mission, two writing samples, outcomes and voice notes, plus "How we sound" in AGENTS.md) from their organization's public documents or the practice pack, then writes one real piece in that voice or a compliance matrix for a live RFP.

The Brief tests "it can't sound like us": two thank-you notes from the same ask, one from a bare chat and one from a folder, and the room guesses which came from the folder. In the outside read, a fresh Claude session lists the lines in each draft that don't sound like the voice notes, and each person reads aloud the line that sounds most like them. Bring to Lab 3: one messy real reporting input, plus the name of the report and who reads it, or the practice mess pack.

Each person takes one report section, from their own reporting input or the practice mess pack, through draft, second chair and revision, and saves the second chair as a recipe, Recipes/second-chair.md, with a rule from a real catch.

The Brief tests "if it reads well, it's right" with a vote on a draft made from the mess pack, re-tested on the current default model before the lab. In the outside read, each person runs a prompt in a fresh session that reads their revision as a different skeptic and quotes what their second chair never asks, then reads the most important question aloud in the room round. Bring to Lab 4: the recurring spreadsheet exported as a .csv with red and notes columns deleted, plus one total from the organization's own system, or the practice sheet.

Each person takes a spreadsheet, their own recurring one or the practice sheet, through the round trip: organization names swapped for ORG-01 placeholders by hand, cleaned by Claude 10 rows first with a change log, reconciled by hand, the names put back outside AI-Labs, and the brief saved as a cleanup recipe.

Red and notes columns are deleted by hand, then each person checks their own column headings (the top row only) with Claude before anything is saved. The placeholder key never goes near Claude or into AI-Labs. In the outside read, each person's own cleanup recipe is read in a fresh session and each person reads one rule aloud, never a key. Bring to Lab 5: the meeting the next deck is for and its numbers with placeholders still in, or the practice set.

Each person writes a context file, then checked insights, a narrative checked by the second chair as the meeting's skeptic, and a slide outline for a real upcoming meeting, on their own numbers or the practice set. The .pptx is started with Claude in the session and finished at home if needed.

The Brief tests "start in the slide tool" with a deck made from the raw numbers. Near the end, Claude plays the meeting's skeptic on one of each person's slides, and the room hears each question aloud. Bring to Lab 6: a draft about to ship and the files it came from, and a research brief in the four parts, saved as Working/research-brief.md.

Each person saves their second chair as a skill, Skills/second-chair/SKILL.md, and runs it on a real draft, then turns a Claude Research report on their own question into a one-page summary of claims checked at the source.

The Brief tests "research gives answers" with a pre-run report: a confident claim, its source's date, and whose voices the sources leave out. A five-minute tool map follows the demo. In the outside read, a fresh session reads each person's skill and draft as a funder they've never applied to and names one question the skill never asks, and each person reads it aloud. Bring to Lab 7: the workflow's pieces gathered in Workflows, and the ship log with its old-way and new-way minutes.

Each person writes their workflow down so a colleague could run it without them, with the tool for every step and old-way and new-way minutes, and a 90-day plan with month 1 filled in. No new teaching.

The Brief reads back participants' Lab 1 words. The build and the showcase run in the usual breakout rooms of about eight, each with a note-taking facilitator. In the showcase, each person has 90 seconds on what the workflow does, what it replaced, their old-way and new-way minutes, what surprised them and what it can't do, then one question from someone in another organization. The 90-day plan takes the place of the homework levels. A separate leadership showcase follows after Thanksgiving, in December or January.