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 →
ChatGPT drafts the situation table and then the map in one conversation. You check the table first, then the map.
Prepare the excerpts
Draft the table
Correct the table
Draft the map
Check and decide
Use: You
Collect the consented excerpts, objective, offer proof, treatment options and segment definitions; note permitted-use limits, quote IDs and dates.
Check consent and usage permission per excerpt; only permitted excerpts included; no sensitive trait inferred.
Use: ChatGPT
Ask ChatGPT for the situation-by-need table: quote IDs, triggers, treatment, promise, overlap and open items; rows without a quote marked 'still to confirm'.
What to paste into the tool
Using only the attached files, list each buying situation. For each give: supporting quote IDs, buying trigger, proposed campaign treatment from the supplied options, differentiated promise tied to the offer proof, overlap with existing segments, and items still to confirm. Label any situation without a quote as a hypothesis. Do not infer sensitive traits.
Use: You
Correct the table row by row against the originals; remove unsupported rows or relabel them 'still to confirm'.
Check every row has a quote ID or a 'still to confirm' label.
Use: ChatGPT
In the same conversation, ask ChatGPT to turn the checked table into the map with promise drafts, overlaps and inappropriate treatments.
What to paste into the tool
From the checked table only, draft the audience-treatment map. For each situation state the promise hypothesis, the treatment, what offer proof supports it, overlaps with other situations and any treatment that would be inappropriate. Do not add facts not in the files. Do not activate or suggest activating targeting.
Use: You
Decide approval of the map and its open questions; any execution is a separate decision.
Check quote evidence or "still to confirm" labels; promises tie to objective and offer proof; overlaps and inappropriate treatments explicit.
Claude codes the excerpts, groups the situations and drafts treatments and promises. You check the coding, then decide the map.
Prepare the inputs
Code and group
Check the coding
Draft the map
Edit and decide
Use: You
Collect the consented excerpts, objective, offer proof, treatment options and segment definitions; note each item's source ID and date.
Check permitted use and that no protected or sensitive trait is recorded.
Use: Claude
Ask Claude to code each excerpt (quote ID, date, trigger, need, permitted-use status) and group codes into buying situations, unsupported items marked 'still to confirm'.
What to paste into the tool
Using only the uploaded excerpts, objective, offer proof, treatment options and segment definitions, code each excerpt with its quote ID, date, buying trigger, need and permitted-use status, then group the codes into buying situations. Keep quote IDs verbatim. Note overlaps with the existing segments and record missing facts and items still to confirm. Mark anything unsupported as a hypothesis. Do not suggest launching, targeting or tracking.
Use: You
Compare each code with its quoted excerpt; check each situation traces to quote IDs or is marked 'still to confirm'.
Check permitted use and that no protected or sensitive trait is recorded.
Use: Claude
Ask Claude for a treatment from your supplied options and a promise draft tied to the offer proof per situation, with overlaps and inappropriate treatments.
What to paste into the tool
Using only the checked table and files, propose for each situation a campaign treatment from the supplied options and a differentiated promise hypothesis tied to the offer proof. Keep quote IDs verbatim. State overlaps and any treatment that would be inappropriate. Mark anything unsupported as a hypothesis. Do not suggest launching, targeting or tracking.
Use: You
Edit the draft into the final audience-treatment map and decide approval; execution is separate.
Check quote evidence per situation; promises tied to objective and offer proof; overlaps and inappropriate treatments explicit.
Gemini Notebook drafts the situation table and the map with citations from your sources. You check every citation and decide.
Prepare the sources
Draft table and map
Check and approve or hold
Use: You
Add the consented interview excerpts, campaign objective, offer proof, treatment options and segment definitions as notebook sources.
Check consent, suppression and usage permissions before any upload; each excerpt has its quote ID and date.
Use: Gemini Notebook
Ask Gemini Notebook for buying situations with citations, overlaps and gaps, then the map with cited promise and open questions each. Unsourced: state, never infer.
What to paste into the tool
From the sources only, list distinct buying situations. For each give the trigger, the need, which excerpts support it (cite them), overlap with other situations and what is not known. If a situation has no supporting excerpt, say so instead of inferring. Using only the sources, draft the audience-treatment map. For each buying situation give: a treatment from my supplied treatment options, a draft promise tied to my offer proof, overlaps with other situations, and treatments that would be inappropriate. Cite the sources for each. If the sources do not support a treatment or promise, say so instead of inventing one. List open questions at the end.
Use: You
Open each citation against its excerpt; compare each treatment with your options and each promise with the offer proof. Approve or hold.
Check every situation has a checked quote or is marked 'no quote in the excerpts'. Check promises tie to the objective and offer proof; overlaps and inappropriate treatments are explicit. Record the decision separately from any campaign execution.
Best when: One person wants one assistant thread for both stages
What it does: ChatGPT drafts table and map; you check between stages
Strong at: One conversation holds files, table and map; no re-upload
Limit: Free-plan upload caps could interrupt a long interview set
Free plan: Yes, limited (uploads, images, voice, data analysis)
Best when: You want Claude to do the coding, grouping and drafting
What it does: Claude codes, groups and drafts; you check and choose
Strong at: You check the coding before any drafting starts
Limit: Not documented to code interviews or keep quote IDs; read outputs fully
Free plan: Yes
Best when: Each claim must trace to a quoted source passage
What it does: Gemini Notebook drafts table and map; you check citations
Strong at: Answers carry citations you can open and check
Limit: Not confirmed to draft campaign treatments; compare with your options
Free plan: Yes, limited (notebook and source caps)
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.
Named human reviewer (Audience researcher) compares every working record with the original inputs and confirms permitted use before drafting.
Every AI output is read against the supplied inputs; unsupplied facts are removed and open questions kept.
Approval and any later execution (publishing, outreach, targeting, tracking, spend) are separate human decisions outside this workflow.
Check the result against your original inputs. Approve it before sharing or acting.