Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
You supply the articles, policy versions, rules and examples. ChatGPT builds the matrix and drafts passages. You check every row.
Collect articles, policies, examples
Build matrix and passages
Check and decide
Use: You
Collect the article variants, the approved simultaneously applicable policy versions, the cohort applicability rules and the approved request examples, keeping source IDs.
Check only the approved articles, policies, rules and examples are included, with no passwords, authentication codes, session tokens or payment credentials.
Use: ChatGPT
In one ChatGPT chat, add the cohort bundle with source IDs; ask for the cohort applicability matrix, then the cohort-labelled passages and unresolved-scope questions.
What to paste into the tool
Task: Draft cohort applicability matrix with conditions, dates, clauses and ambiguous cases. Use only the supplied articles, policy versions, cohort rules and examples. Build a matrix with one row per cohort and example: the rule condition, policy version, effective and transaction dates, the clause that applies, and the cohort. Mark any example the rules do not settle as ambiguous. Do not pick a version by date alone when other conditions apply. Task: Draft cohort-labelled article passages; Cohort applicability matrix with article passage IDs added; Unresolved-scope questions. Write each article passage with a cohort label and its condition, keeping every still-valid legacy passage as it stands. Add the passage ID to each matrix row. Do not describe either policy as obsolete. List scope questions the rules leave open in a separate list. Tie each open question to a worksheet row.
Use: You
Test every matrix row against the approved rules and boundary examples, correct in the same chat, and close the work as preparation only.
Do not let a date alone choose the cohort where other conditions apply. Check each example maps to one supported cohort or an ambiguous state, claims carry a cohort condition, and no valid legacy guidance is overwritten. Settle publication and authority separately; not approval to publish.
Claude drafts the cohort matrix and passages. You test every row against the approved rules before passages are settled.
Collect articles, policies, examples
Draft matrix and passages
Test rows, check passages
Use: You
Collect the article variants, the approved simultaneously applicable policy versions, the cohort applicability rules and the approved request examples, keeping source IDs.
Check only the approved articles, policies, rules and examples are included, with no passwords, authentication codes, session tokens or payment credentials.
Use: Claude
In one Claude chat, add the cohort bundle; ask for the matrix with one row per example, then the cohort-labelled passages and unresolved-scope questions.
What to paste into the tool
Task: Draft cohort applicability matrix with conditions, dates, clauses and ambiguous cases. Use only the supplied articles, policy versions, cohort rules and examples. Build a matrix with one row per example: the rule condition that decides who the rule applies to, the policy version, effective and transaction dates, the clause that applies, and the cohort. Mark any example the rules do not settle as ambiguous. Do not treat either policy as obsolete. Keep the ID beside each row. Task: Draft cohort-labelled article passages; Cohort applicability matrix with article passage IDs added; Unresolved-scope questions. Write each article passage with a cohort label and its condition, keeping every still-valid legacy passage as it stands. Add the passage ID to each matrix row. Do not describe either policy as obsolete. List scope questions the rules leave open in a separate list. Use only the rows you are given.
Use: You
Test each drafted row against the approved rules and boundary examples, correct in the same chat, check the passages, close as preparation only.
Do not let a date alone choose the cohort where other conditions apply. Check each example maps to one supported cohort or an ambiguous state, claims carry a cohort condition, and no valid legacy guidance is overwritten. Settle publication and authority separately; not approval to publish. Check the result against your original inputs. Approve it before sharing or acting.
Gemini Notebook quotes the rules, maps examples to cohorts and drafts. Gemini audits the draft. You check and decide.
Collect articles, policies, examples
Quote rules, map examples
Check each mapping
Draft passages and questions
Audit against the matrix
Decide marks and finish
Use: You
Collect the article variants, the approved simultaneously applicable policy versions, the cohort applicability rules and the approved request examples, keeping source IDs.
Check only the approved articles, policies, rules and examples are included, with no passwords, authentication codes, session tokens or payment credentials.
Use: Gemini Notebook
Add the bundle in Gemini Notebook; ask for a quote table, then a matrix mapping each example to a cohort or an ambiguous state.
What to paste into the tool
Task: Verbatim quote table of cohort rules, policy clauses and request examples with IDs. Use only the supplied material. Quote each cohort rule predicate, each policy clause and each request example exactly, one per row, with its ID. Add no mapping or classification. Task: Cohort applicability matrix with each example mapped to a cohort or an ambiguous state. Apply only the approved cohort rules in the sources. For each request example, set the cohort its conditions satisfy, or mark it ambiguous when no rule settles it, and note which condition decided it. Do not treat either policy as obsolete and do not infer eligibility beyond the supplied rules. Keep the ID beside each row.
Use: You
Compare each row's cohort or ambiguous mark with the approved rules and the quoted example before drafting.
Check the deciding condition is the one the rules give.
Use: Gemini Notebook
In the same notebook, ask for the cohort-labelled article passages, passage IDs added to each matrix row, and unresolved-scope questions.
What to paste into the tool
Task: Draft cohort-labelled article passages; Cohort applicability matrix with article passage IDs added; Unresolved-scope questions. Write each article passage with a cohort label and its condition, keeping every still-valid legacy passage as it stands. Add the passage ID to each matrix row. Do not describe either policy as obsolete. List scope questions the rules leave open in a separate list. Tag each sentence with its worksheet row so the later audit can check it.
Use: Gemini
Add the passages, matrix and questions to a Gemini chat; ask it to mark each claim the corrected matrix does not support.
What to paste into the tool
Task: Draft cohort-labelled article passages with unsupported statements marked; Cohort applicability matrix with article passage IDs added with audit marks; Unresolved-scope questions updated after the audit. Check each labelled passage and each matrix row against the corrected matrix. Mark any claim without a cohort condition, any example that is not covered by exactly one cohort or an ambiguous state, and any legacy passage that was changed. Update the unresolved-scope questions with what you marked.
Use: You
Read each mark against the matrix: add a cohort condition, restore a legacy passage or mark the case ambiguous; close as preparation only.
Do not let a date alone choose the cohort where other conditions apply. Check each example maps to one supported cohort or an ambiguous state, claims carry a cohort condition, and no valid legacy guidance is overwritten. Settle publication and authority separately; not approval to publish. Check the result against your original inputs. Approve it before sharing or acting.
Best when: You want one assistant for the matrix and the passages
What it does: ChatGPT builds and drafts; you check every row
Strong at: Applies a framework from one document to another
Limit: A missed matrix row can reach the passages unchecked
Free plan: Yes
Best when: You want the matrix and passages drafted, checking each stage
What it does: Claude drafts both; you test every row
Strong at: Drafts and edits text in many formats
Limit: A wrong drafted row carries into the passages
Free plan: Yes
Best when: You want quotes, mapping and drafting in one tool, then an audit
What it does: Gemini Notebook quotes, maps and drafts; Gemini audits; you decide
Strong at: Answers grounded only in the sources you add
Limit: Material goes to two vendors, two data-handling terms
Free plan: Gemini Notebook: Yes; Gemini: Yes
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Claude for this? See the comparison.
Never paste credentials or full payment details.
Each example maps to exactly one supported cohort or an explicit ambiguous state, and each claim has a cohort condition.
Every statement in the draft traces to a checked row with a source ID.
Gaps stay visible as gaps; none is filled by inference.
No credentials of any kind are copied.
The named reviewer confirms the evidence and the draft; completion is preparation only.
Check the result against your original inputs. Approve it before sharing or acting.