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 →
GitHub Copilot proposes definitions and SQL text; you sign business choices and check joins without executing anything.
Collect schema and unresolved churn
Draft query text in Ask
Check joins
Use: You
Have ready: table definitions, key relationships, business question, churn-window policy and SQL dialect; record missing cutoffs.
Use: GitHub Copilot
Open only the schema in VS Code; use GitHub Copilot Ask mode for definitions, comparison SQL, count checks and caveats; copy locally.
What to paste into the tool
Task: Churn schema packet with keys, dialect and unresolved business choices; Proposed churn terms, comparison SQL, count-check SQL and caveats. Use the supplied keys and dialect; identify missing cutoffs as questions; draft churn-window and tenure definitions, a comparison query, and count checks requiring churned plus retained to equal the customer count; state association only and give no computed values; preserve the input packet alongside the draft
Use: You
Trace joins to declared keys; sign churn window and tenure bands, check nulls and duplicates, edit SQL without running it.
The churn analyst releases the response only to local SQL review. The churn analyst releases the query package only after signing its definitions. No computed table or results are produced here. Keep the caveats. Add a note to run the query later on a copy and to hand-count 20 customers outside this workflow. Edit all query text without running it.
You write and sign the cohort definitions; Gemini Code Assist drafts SQL, then you compare every clause without executing it.
Write and sign the definition
Draft SQL from the signed
Check SQL against the signed
Use: You
Have ready: schema documentation and churn policy. Define population, endpoints, retention complement, tenure bands and joins; sign definitions and mark unresolved choices.
The churn analyst holds the worksheet locally until its business definitions are signed.
Use: Gemini Code Assist
Place the worksheet in a scratch file; ask Gemini Code Assist for unexecuted comparison and reconciliation query text and schema questions.
What to paste into the tool
Task: Churn worksheet and supporting schema text; Unexecuted comparison and reconciliation query text with schema questions. Translate only the worksheet definitions into the stated SQL dialect; add count-check queries for disjoint and exhaustive churned and retained groups; flag missing keys instead of filling them; attach schema questions and caveats; do not run a query or claim results; carry the worksheet forward
Use: You
Check each WHERE clause and join; correct SQL, retain caveats and specify a later analyst-owned run on a copy.
Check churned plus retained equals the population. No computed table or results are produced here; keep proposed query text out of an execution console. The churn analyst accepts the text only after the signed worksheet and SQL agree.
You write definitions and SQL; Cursor returns a read-only critique, then you reconcile every concern and sign the text.
Write and sign definitions
List concerns
Check each concern
Use: You
Have ready: schema and business definitions. Write comparison and reconciliation SQL, null cases, duplicate handling and caveats; sign cohort terms without execution.
Use: Cursor
Use Cursor Ask mode in an isolated workspace; copy its line-referenced critique without accepting file changes.
What to paste into the tool
Task: Disclosure-checked churn SQL bundle with schema references; Churn SQL critique with line references and unresolved cohort risks. Inspect the provided SQL against its signed definitions; identify join multiplication, null handling and group-overlap risks with line references; examine whether churned plus retained is checked against the customer count; return critique rather than replacement files; do not execute code
Use: You
Accept or reject each concern with a reason; revise SQL and assemble signed definitions, queries and caveats.
The churn analyst routes the critique to manual reconciliation before accepting changes. The churn analyst signs the reconciled draft before it leaves text-only review. Specify a later run on a copy and a 20-customer manual check outside this workflow.
In short: A proposes cohort logic and SQL, B translates your signed definitions, and C critiques SQL you wrote.
Best when: You want proposed definitions and SQL for analyst sign-off.
What it does: GitHub Copilot proposes definitions and SQL; you sign and check
Strong at: Drafts comparison and count-check SQL from supplied schema context.
Limit: No query execution or computed result; business definitions need your signature.
Free plan: Yes
Best when: You want to define cohort logic before SQL translation.
What it does: Gemini Code Assist translates; you define cohorts and check SQL
Strong at: Translates a supplied signed definition worksheet into comparison SQL.
Limit: You must check clause agreement; the queries remain unexecuted.
Free plan: Yes, daily request caps
Best when: You already have SQL and want a read-only critique.
What it does: Cursor critiques; you write SQL and settle concerns
Strong at: Returns line-referenced SQL concerns without accepting file changes.
Limit: You author all definitions and queries and settle every critique.
Free plan: Yes, limited agent features
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between Cursor and GitHub Copilot for this? See the comparison.
Keep definitions signed and leave query execution and computed results outside this draft workflow.
Check the result against your original inputs. Approve it before sharing or acting.