Best Practices for Prompting Intuidy AI
Getting the most out of AI in Vantage starts with how you ask questions and how you configure context. This guide covers practical techniques for writing effective prompts across every AI touchpoint — the assistant, tile AI, context snippets, and workflow nodes.
General Principles
Be Specific
Vague questions produce vague answers. The more specific your prompt, the more actionable the response.
| ❌ Vague | ✅ Specific |
|---|---|
| "Tell me about this data" | "What are the top 3 revenue trends in this quarterly sales data?" |
| "Analyze this" | "Compare this month's conversion rates to last month and highlight any anomalies" |
| "Help me" | "How do I create a bar chart tile showing revenue by region?" |
Provide Context in the Prompt
Even though Intuidy AI is context-aware, adding relevant details to your prompt improves accuracy:
- Mention time periods: "…for Q4 2025"
- Specify metrics: "…focus on MRR, not total revenue"
- Name categories: "…broken down by product line"
Use Natural Language
You don't need to write SQL or code. Speak naturally:
- "Show me which customers had the biggest increase in spending over the last 6 months"
- "What does this waterfall chart tell me about our cash flow?"
AI Assistant Prompts
The Global AI Assistant is your general-purpose AI interface. Here are effective prompt patterns:
Data Questions
What were our top 10 customers by revenue in Q4 2025?
Trend Analysis
Compare this month's sales to the same month last year.
What changed and why might that be?
Action Recommendations
Based on the data in my dashboard, what are the three most
important actions I should take this week?
Configuration Help
How do I connect my Slack workspace to Vantage for workflow
notifications?
Tile AI Prompts
When using Popup AI Chat on a specific tile, the AI already knows the tile's data, type, and configuration. Lean into that:
Good Tile Prompts
- "What's the most notable insight in this data?"
- "Are there any outliers I should investigate?"
- "Explain the trend line in plain language for a stakeholder presentation"
- "If this trend continues, what can I expect next quarter?"
Avoid Redundant Context
Since the tile AI already sees the data, you don't need to describe it:
| ❌ Redundant | ✅ Direct |
|---|---|
| "This bar chart shows monthly revenue from Jan–Dec. What's the trend?" | "What's the trend here and what's driving it?" |
Context Snippet Best Practices
Context snippets are organization-level or personal notes injected into every AI request. They're the most leveraged optimization — write them once, improve every interaction.
Organization Snippets
Write these as if briefing a new analyst on Day 1:
Company Overview:
"Acme Corp is a B2B SaaS company providing supply chain management software to mid-market manufacturers. Founded in 2018, we have 2,400 customers and $35M ARR. Our primary metrics are MRR, net revenue retention, and customer acquisition cost."
Industry & Market:
"We operate in the manufacturing supply chain vertical. Key competitors are SupplyChainPro and LogiTrack. Industry benchmarks suggest 110%+ NRR is best-in-class for our segment."
Custom Instructions
Use these for formatting and behavioral preferences:
- Always format currency as USD with commas (e.g., $1,234,567)
- When discussing growth, default to quarter-over-quarter comparisons
- Keep summaries under 3 paragraphs
- Highlight risks and anomalies before positive trends
- Use bullet points for actionable recommendations
Personal Snippets
Add role-specific context:
"I'm the VP of Sales. I care most about pipeline velocity, win rates, and rep performance. When analyzing data, prioritize metrics relevant to sales leadership."
Workflow AI Node Prompts
Workflow prompts work differently than conversational prompts. They process data rows in bulk, so clarity and structure matter even more.
AI Enrichment Prompts
Be explicit about the output column and format:
Analyze the "customer_feedback" column. Add a new column called
"sentiment" with values: POSITIVE, NEGATIVE, or NEUTRAL.
Add a second column called "key_topic" with the main subject
of the feedback in 2-3 words.
AI Compliance Check Prompts
State the policy clearly with pass/fail criteria:
Evaluate each row against our data retention policy:
- Personal data must have a retention period of ≤ 365 days
- Financial records must have a retention period of ≤ 2555 days
- Any row without a retention period should FAIL
Mark each row as PASS or FAIL with a brief reason.
AI Formatter Prompts
Describe input and output formats:
The "address" column contains unstructured addresses.
Reformat each into: Street, City, State ZIP.
If any component is missing, leave it blank.
AI Conditional Prompts
Define clear routing criteria:
If the "priority" assessment is HIGH or CRITICAL, route to
the "urgent" branch. All others go to the "standard" branch.
Common Mistakes
| Mistake | Fix |
|---|---|
| Too vague — "Analyze this" | Be specific about what to analyze and what you need back |
| Too long — walls of text | Keep prompts focused; use context snippets for persistent details |
| Conflicting instructions — custom instructions say "be concise" but prompts ask for "detailed analysis" | Align your custom instructions with your typical use case |
| No formatting guidance | Tell AI how you want results: bullets, tables, numbers, etc. |
| Forgetting context snippets exist | Review and update snippets quarterly as your business evolves |
Iteration Tips
- Start broad, then narrow — Ask a general question, then follow up with specifics based on the response
- Ask for alternatives — "What other ways could I interpret this data?"
- Request reformatting — "Rewrite that as a table" or "Summarize in 2 sentences"
- Challenge the AI — "Are there any risks or counterarguments to this conclusion?"