What this is
This article explains how to write effective prompts in Propensity AI. It covers what makes a prompt work well, provides examples across the most common use cases, and describes the mistakes most likely to produce incomplete or inaccurate results.
When to use this
Use this guide when you are:
New to Propensity AI and looking for examples to get started
Familiar with the platform but not getting the results you expect from your prompts
Looking to use specific platform features through natural language
Prompting Best Practices
Be specific about who you are targeting when you already know your audience: Include job roles, seniority, company sizes, and industries when building audiences. The more detail you provide, the more precisely the AI can filter your results. If you're not sure who your target audience should be, start with a general prompt instead. The AI can help you discover and define your audience based on broader criteria like intent signals or campaign goals.
Include time frames when relevant: For intent signals, website visits, and reporting queries, specify a time window. Phrases like "in the last 90 days" or "this quarter" produce more focused results than open-ended requests.
State your goal: Include the intended outcome of the campaign or play. Phrases like "to drive demo requests" or "to accelerate warm accounts to demo stage" help the AI align the proposal with your objective.
Specify channels when you know them: If you know which channels you want to use, include them in your prompt. This prevents the AI from making a default channel selection that may not fit your strategy.
You don't need to know platform terminology to get the right campaign type: Natural, descriptive prompts (e.g. "target people near a conference venue" or "reach companies researching a specific topic") give the AI enough context to select the appropriate type automatically. If you do know specific terms like "contact-based marketing," "account-based marketing," "contextual campaign," or "geofencing," feel free to use them; the AI will recognize the terms.
Prompt Examples by Use Case
Audience Building
Targeting accounts by role, company size, and intent
Build an audience of marketing leaders at mid-size financial services companies actively researching ABM in the last 90 days.
Re-engaging contacts who have visited your site
Create a retargeting audience of contacts who visited the pricing page of my website in the last 30 days.
Campaign Building
Building a campaign from scratch
Create a contact-based marketing campaign targeting my Q1 Account List across LinkedIn and PDN to drive demo requests.
Reaching people based on content consumption
Build a contextual campaign targeting companies researching demand generation strategies to drive webinar registrations
Targeting attendees at a specific event or location
Set up a geofencing campaign around the Dreamforce conference to target attendees with display ads.
Targeting high-value accounts for pipeline acceleration
Create an account-based marketing campaign targeting warm accounts to accelerate them to the demo stage.
Reporting
Pulling leads by warmth and timeframe
Show me all hot leads from my active campaigns in April.
Comparing performance across campaigns
Which of my campaigns drove the most engagement so far this quarter and what channels are they running on?
Checking campaign spend by cohort
How much did my B2B Targeted Marketing campaign spend in Cohort 1, broken down by channel?
Identifying top-performing channels
Which channels are driving the best CTR and CPC across my contact-based marketing campaigns this month?
Reviewing asset performance
Which Facebook ad creatives in my High Value Accounts campaign have the highest CTR and which are underperforming?
Important Notes
Propensity AI generates a proposal before saving anything. Review all proposals carefully before confirming. Nothing is created until you confirm.
You can reference your existing platform data in prompts. Named campaigns, account lists, CRM imports, and cohorts are all accessible through natural language.
Campaign type is chosen automatically based on your prompt, unless specified by the user. Using specific terms like "contact-based marketing" or "account-based marketing" helps the AI select the correct type. Always verify the campaign type in the proposal before confirming.