Phase Three: Building Object Packs (part two of two)

This is Unit Six of The M-CaReM-AI Method, part two, on drafting, maintaining and choosing Object Packs. Part one built the Incident Report's Pack in ten steps and set out the three levels. This part takes the structure to a second domain, puts AI to work on Packs safely, and builds the library's own. Carry one question through this lecture: is the most important thing in daily work always the one that deserves the deepest Pack? The answer arrives when the library applies the Gold test. By the end you will be able to outline a Pack for a decision thing. To draft a Pack with AI and keep it in sync. To recognize six kinds of thing. And to assemble a Silver Pack from earlier work.

One entry that outlines a second Pack, for a decision in a benefits agency.

A Pack for a decision.

MCM 144 The Determination Pack in Outline

The book outlines a Pack for an Eligibility Determination, to show the structure working in another domain. Its purpose is recording eligibility decisions as the basis for notices and audits. Its lifecycle runs pending, active, superseded and revoked. Its obligations say the policy version must be the one in effect on the effective date, and every determination must trace to its snapshot and policy. Its open questions ask whether a case may hold several active determinations per program, and whether corrections update or supersede. Even in outline, one document serves legal, operational and technical readers.

Two entries on drafting Packs with AI, and keeping them in sync without quiet changes.

AI drafts, you edit.

MCM 145 AI Drafts, You Edit

The method uses AI to assemble a first Pack from pieces the team supplies. Gather the thing's description, cards, fields, lifecycle notes and policy text. Give the tool those pieces and the template, and ask for clear business language. Forbid invented fields, states or relationships, and have it list gaps as open questions. Then edit: remove inventions, clarify wording, fix the lifecycle, and tighten the policy links. The book calls AI fast but occasionally careless; a Pack drafted from a vague prompt encodes the tool's guesses as truth.

Keeping a Pack current.

MCM 146 Keeping a Pack in Sync

The method uses AI to keep Packs current, with a check on the change log. Give the tool the old Pack, the updated cards and the new decisions. Ask it to update the Pack and log every change. Check that each logged change is correct. Then check the other direction: that nothing changed which the log does not mention. Finally, the steward approves. The fourth step is the one teams skip, and the method calls an unlogged edit to a Gold Pack exactly the quiet change it forbids.

One entry on six kinds of thing, each with its own modeling advice.

Six patterns for things.

MCM 147 Patterns for Things

The Toolbox names six kinds of thing. An anchor thing, the center of a domain, gets its Pack early at Silver or Gold. A supporting thing, such as an Action Plan, states how its lifecycle relates to the anchor. A reference list usually needs only Bronze. A snapshot, the state of something at a moment, becomes a thing of its own. An event log focuses on who, what, when and why. A decision, carrying policy weight, is Gold. Recognizing the kind tells you the depth before anyone writes a line.

Three constructed examples that build the library's Packs. The opening question is answered here.

A Silver Pack for the Loan.

MCM 148 A Silver Pack for a Library Loan

A worked example, constructed for this course, assembles a Silver Pack for the library's Loan from earlier units. The header names the head of circulation as steward. The definition, agreed in unit two, says a renewal changes the due date and does not create a new Loan. The properties carry the policy version. The lifecycle includes the pause added in unit three. The card index lists eight cards and shows that no archive card exists. Three open questions remain. Nearly every line came from earlier work: if you did phases one and two, you already hold most of a Pack.

The Gold test.

MCM 149 Does Anything in the Library Earn Gold?

A second constructed example applies the Gold test to the library. The Member's privacy is met by Silver access rules. A mistake with an Item costs a misplaced book. The Loan, the center of daily work, usually costs an inconvenience, so it stays Silver, with a matrix because it changes often. Then the Fine Calculation Record: it decides what a member owes, is audited yearly and disputed months later, so it becomes the only Gold Pack. That answers this lecture's question: the most important thing in daily work was not the Gold one; the test is the cost of a mistake, not how often a thing is used.

A logged change.

MCM 150 A Pack Change, Logged

A third constructed example updates the Gold Pack with AI when fines on children's cards are waived. The tool logs three correct changes: a waiver reason code, a zero amount for junior members, and an effective date. The steward then compares old and new section by section, and finds a fourth, unlogged change: the retention period for fine records cut from seven years to three, copied from a general template. She restores it and records the reversal. Every logged change was right; the dangerous one was unlogged, which is why the two-way check matters for Gold things.

Look at what you can now do. You can outline a Pack for a decision thing, with its states, obligations and open questions. You can draft a Pack with AI as its editor-in-chief, and keep it in sync by checking the change log in both directions. You can recognize six kinds of thing, and let the kind set the depth. And you can assemble a Silver Pack from earlier work, and choose Gold by what a mistake would cost. Unit Seven checks Packs against reality.