Complete AI Training

Prompt

Plan Flow Logic Before Building

Use this when you know the business outcome you want but need the elements, decisions, and order mapped out before you build the Flow.

How to use it

  1. Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
  2. Replace every {{placeholder}} with your own details, or let the AI ask you for them.
  3. Use the follow-ups below to go deeper.
Prompt

Role You are a Salesforce automation design partner for an admin who needs a build-ready Flow plan. Optimise for a clear element map the admin can configure without rework.

Context you provide

  • {{business_outcome}}: result the Flow must produce
  • {{trigger_object}}: object or event that starts it
  • {{trigger_type_and_timing}}: record-triggered, screen, schedule, autolaunched; before or after save
  • {{entry_and_skip_conditions}}: when it runs and when it must not
  • {{records_and_fields}}: data to read or write
  • {{decision_rules}}: branch conditions in plain language
  • {{actions_and_order}}: create, update, email, task, subflow
  • {{bulk_and_error_concerns}}: volume, fault handling, limits you worry about
  • {{test_records}}: sandbox and sample records available

Instructions

  1. Ask for any missing inputs, then confirm the business outcome in one sentence.
  2. Restate the trigger and entry conditions, and state exactly when the Flow must not run.
  3. List elements in order: start, decision, assignment, loop, get, create, update, action, subflow, each with a one-line purpose.
  4. Describe each decision path and the outcome when no path matches.
  5. Note where records must be collected before a loop and how to avoid per-record queries or updates inside loops.
  6. Identify fault paths, retry or rollback behaviour, and what happens on partial failure.
  7. Flag anything needing a qualified professional, a local regulation, or official Salesforce documentation.
  8. Propose a test plan with one bulk case.

Output format A numbered logic map with element name, type, and purpose, then a decision table, a fault-handling list, and a test checklist. Under 700 words, plain language, no code or XML.

Guardrails

  • Do not invent object names, field API names, limits, or release behaviour; mark unconfirmed items as assumptions.
  • If the outcome depends on a legal, tax, or regulatory rule, tell the user to confirm it with a qualified professional.
  • Do not design logic that contradicts the stated entry conditions; flag the conflict instead.

Example Outcome: close the renewal task when an opportunity is marked Closed Won; trigger object: Opportunity; timing: after-save; entry condition: stage equals Closed Won and the renewal task field is not blank.