What it is
Reverse prompting (sometimes called the "flipped interaction" pattern) turns the normal workflow around. Usually you front-load every detail into a single prompt and hope you didn't leave anything out. With reverse prompting, you hand the AI the goal and tell it to ask you questions first — ideally one at a time — until it has everything it needs. Only then does it produce the deliverable. A close cousin is asking the model to write the prompt itself: "Draft the ideal prompt I should give you to get X," which you then review and run.
Why it matters
Most disappointing AI output isn't a model problem — it's a context problem. People don't know what the model needs, so they leave out the details that matter most. Reverse prompting closes that gap by making the AI responsible for gathering its own context, which means fewer vague, generic answers and far less back-and-forth. For Salesforce and Agentforce work specifically, it's a fast way to scope requirements, draft customer-ready content, and design agent instructions without missing the obvious questions.
How it works
- State the goal and hand the AI the wheel — e.g., "I want to write X. Before you write anything, ask me the questions you need to do it well."
- The AI interviews you, one question at a time, surfacing details you'd never have thought to include on your own.
- You answer in plain language — no need to format anything as a prompt or get the wording right.
- The AI confirms its understanding (often as a short summary), then produces the deliverable grounded in your answers.
Examples
“Act as a senior Salesforce solution architect. I want to scope an Agentforce service agent for my support team. Don't propose anything yet — interview me one question at a time about my current setup, my data, and my goals. When you have enough, summarize the requirements back to me.”
“I need a cold outreach email that actually gets replies. Don't write it yet. First, ask me everything you need to know about my offer, my audience, and the outcome I want — then write three versions.”
“I want to use AI to summarize my weekly sales calls. Write the ideal prompt I should give an AI to do this really well, including what context it should ask me for. I'll review it before running.”
“Before we draft the system prompt for my Agentforce agent, ask me about its job, the tone it should use, what it must never do, and when it should hand off to a human. One question at a time.”
How it connects
In Salesforce work, reverse prompting pairs naturally with Prompt Builder: run a quick reverse-prompting session to discover what context a prompt template actually needs, then bake those fields into a reusable template for your Agentforce agents.
Good to know
The trick that makes it work is telling the model to ask one question at a time — otherwise it dumps ten questions at once and you're right back to front-loading. Adding "ask one at a time, and wait for my answer" to almost any prompt turns it into a reverse prompt.
Related terms
Agentforce
Salesforce's AI agent platform that enables businesses to build, customize, and deploy autonomous AI agents across sales, service, marketing, and commerce.
LLM (Large Language Model)
The AI technology behind ChatGPT, Claude, and the intelligence in Agentforce. Trained on massive amounts of text to understand and generate human language.
Prompt Engineering
The practice of crafting precise instructions to guide an AI model's behavior, capabilities, and limitations.
Prompt Builder
Salesforce's tool for creating reusable, templatized prompts that pull in live CRM data. Build once, use across flows, agents, and record pages.