Teaching the Method

This is Unit Seven of Leading AI at Work, on teaching the method to others. Unit Six closed on governance that needs somewhere concrete to stand. That place only exists if people across a team can build it, so this unit turns to teaching. Carry one question through this lecture: what makes a session on the method stick? The answer arrives with a workshop planned for one agency at the end. By the end you will be able to apply the book's five teaching principles. To run its one-day workshop and its ninety-minute clinic. And to choose exercises, drills and a self-study path, measuring by outcomes.

Seven entries on a session that ended a two-year argument, and the five principles behind it.

Two years of arguing, resolved.

LAI 151 The Clinic That Ended an Argument

The M-CaReM-AI method, by J. R. Carver, opens its teaching chapter with a team that argued about the Customer for two years. A ninety-minute session with a Silver Pack and one card showed that three different things had shared one name. Nobody won; the argument became visible, and the team left with three Packs to write.

The lesson in that session.

LAI 152 Practise on Real Confusion

The book's lesson is that teaching works best when people practise on confusion they already have. A team stuck on what a Claim means learns more from defining the Claim than from any textbook example. An exercise far from their work teaches the steps but not the reason.

Principle one.

LAI 153 Start With Their World

The first principle is to start with the learners' world, not the method. Begin with their pain: failed projects, rework, audit scares, AI misfires. Then show how the method explains it: we never agreed what Case meant. People care about their problems first, frameworks after.

Principle two.

LAI 154 Concrete Examples, Early

The second is concrete examples, early and often, because abstractions do not stick. Draw a messy map on a whiteboard, or show a realistic card and ask whether it is clearer than the last specification they saw. The closer to their world, the faster people connect.

Principle three.

LAI 155 Collaborative, Not a Lecture

The third is to teach collaboratively, in cross-functional groups from business, technology, data, testing and AI. Small-group exercises welcome disagreement, because shared understanding is the method's purpose. With one discipline present, the book says, you miss translation across silos.

Principle four.

LAI 156 Layer the Complexity

The fourth is to layer the complexity rather than deliver everything in the first hour: things and zones, then cards, then Packs, then matrices, and AI patterns only once those four make sense. Each layer builds on the last.

Principle five.

LAI 157 Tie It to AI and Change

The fifth is to keep tying the method to AI and to change. People indifferent to methodology still care about adding AI without risk, and about the next policy change or audit. Lines such as "this makes AI prompts safer" keep it practical.

Six entries on the book's one-day workshop, its rule for AI, and a ninety-minute alternative.

The room for a full day.

LAI 158 The One-Day Workshop: Who and How Many

The book's one-day introductory workshop suits eight to twenty people, in person or online, mixing analysts, product owners, architects, testers, AI engineers, and experts from policy, operations and compliance. The morning covers framing, maps and cards; the afternoon, Packs, matrices and AI.

The morning.

LAI 159 Workshop Morning: Framing, Map, Cards

In the book's suggested agenda, the morning opens with about forty-five minutes of framing: a problem storm about recent projects that went wrong, linked to the method in a few slides. Next, seventy-five to ninety minutes on things and zones: a noun storm, clustering, labelled relationships and a vote on three to seven core things. Then an hour or more on a first card, drafted in pairs and merged. A common discovery is that the group has used Case in several ways. The book adds: let people breathe at the break.

The afternoon.

LAI 160 Workshop Afternoon: Pack, Matrix, AI

The afternoon keeps the morning's thing. Groups build a simple Pack from their card after a look at Bronze, Silver and Gold. Then a mini-matrix, and a hypothetical change, such as a new legal field, to see what must be revisited. Then AI-assisted modeling: run a thing-aware prompt, mark what is correct, wrong and missing, and rewrite. Thirty minutes on next steps close the day. Staying with one thing is the design, because it shows a card feeding a Pack and a Pack feeding a matrix.

A rule for the AI segment.

LAI 161 AI in the Workshop: No Personal Data

For the AI segment, the book sets a condition: use a tool the organization allows, with no personal or confidential data. The input is the thing's definition and notes; the debrief asks where AI helped, where it confabulated, and how the frame exposed mistakes. Training is no exception to data rules.

Ending with commitments.

LAI 162 Turning Energy Into Commitments

The last thirty minutes turn the day's energy into commitments: where to pilot, what a minimum version looks like, and promises such as, within a month, we will build Packs for these two things. A workshop without a named next step usually ends there.

Ninety minutes, one mess.

LAI 163 The Ninety-Minute Clinic

With no time for a day, the book offers a ninety-minute Object and Card Clinic. Spend ten minutes picking one confusing area and the thing at its center. Make a fast inventory and mini-map around it. Spend about half an hour drafting a Silver Pack together, with a card index naming the behaviors people argue about. Write one card for the most painful behavior, then capture next steps. This is the format behind the book's opening scene, and it works because a live disagreement gives every minute a purpose.

Nine entries on practice away from live work, four role drills, self-study, and measuring progress. The opening question is answered here.

Practising somewhere safe.

LAI 164 Practise on Fictional Domains

The book also recommends fictional domains before a mission-critical one, so people learn the steps without fear of getting their own Gold things wrong. Fictional practice builds skill; real confusion builds commitment, so good practice uses both.

Three exercises, rising in difficulty.

LAI 165 Three Fictional Exercises

The book gives three fictional exercises. A public library, the domain of the constructed examples in The M-CaReM-AI Method: a Loan Pack, cards for lending, returns and fines, then an e-book loan rule and a matrix. An online course platform: an enrollment card, a Grade Pack with an appeal lifecycle, and a matrix including a grade-prediction model. A family cash benefit: a Gold Determination Pack and cards, then changed income thresholds traced through the matrix. Groups compare results. The rising order mirrors layering.

A drill for analysts.

LAI 166 Drill: Translate the Spec

The book's role drills take fifteen to twenty minutes a week on live work. For analysts and product owners: take one page of a specification, highlight its nouns, propose an inventory, turn one paragraph into a card, name the Pack it affects, and ask a colleague whether the card says more. A page with no nouns worth keeping is a finding: it describes screens, not things.

A drill for architects.

LAI 167 Drill: Align the Data Model

For architects and data modelers: label each table in a schema with the business thing it belongs to, spot tables mixing several things, which the book calls everything-bucket tables, and propose one primary table per thing. Bring it to the next design review. Labelling first shows hidden domain fracture to people who do not read schemas.

A drill for testers.

LAI 168 Drill: Test Cases From a Card

For testers: take one card, and for each rule write a test that meets it and one that breaks it, tagging each with the thing and card. Send any rule too vague to test back to its author with the question it raised. The book says this surfaces under-specified rules remarkably quickly; a rule no test can break was never precise enough to build.

A drill for AI engineers.

LAI 169 Drill: A Thing-Aware Prompt Library

For AI engineers: pick a Gold thing with a current Pack and write three prompts, one drafting a card from its definition, one generating tests from a card, one summarizing the Pack for non-technical readers. Refine each until it is consistently useful, then add it to the library with an owner and a last-tested date. Consistency is the test, because others will trust it.

Learning alone.

LAI 170 A Self-Study Path

The book's self-study path runs four to six weeks. Week one, think in things: an inventory, a map, one conversation in their terms. Week two, write three to five cards and get a critique. Week three, a Silver Pack used once for real. Week four, a matrix applied to one change. Weeks five and six, two or three prompt patterns and a small pilot area. The first four weeks match The M-CaReM-AI Method; the last two match this course.

Measuring teaching.

LAI 171 Measure Outcomes, Not Templates

The book warns against measuring teaching by filled templates. Measure outcomes: fewer questions about what a field means, fewer surprises after changes, AI drafts needing less rewriting, roles citing the same Packs. The aim, in the book's words, is to feel the ground getting firmer.

A workshop for one agency.

LAI 172 Planning a Workshop for an Agency

A worked example, constructed for this course, plans the workshop for a transport licensing agency. Kofi gathers fourteen people across analysis, product, architecture, testing, AI, policy, compliance and the service desk. The problem storm surfaces a rule change that broke renewal letters and an AI summary quoting the wrong fee. Mapping the Driver License Renewal, the group finds renewal has meant both the application and the issued license. Merging card drafts adds two unwritten rules. The afternoon builds a Pack and matrix, and a new medical-check rule yields eleven items to revisit. In the AI segment, with an approved tool and no personal data, every group catches a draft field for a driver's credit score. Two commitments close the day. That answers this lecture's question: a session sticks when it starts from the group's own confusion and ends with a named commitment.

Look at what you can now do. You can apply the five principles: start with their world, use concrete examples, teach in mixed groups, layer the complexity, and tie it all to AI and change. You can run the one-day workshop on a single thing from morning to close, under the data rule for AI, and the ninety-minute clinic for one live mess. And you can choose fictional exercises, role drills and a self-study path, and measure by firmer ground rather than filled templates. Unit Eight turns to your next ninety days.