Introduction
Every interaction with AI is a conversation. And like any conversation, the clarity of your request directly determines the quality of the response you get back. A vague prompt leads to vague results. A well-structured prompt leads to exactly what you need - faster, more accurately, and with fewer follow-ups.
These tips teach you how to become a prompt engineer: someone who crafts instructions that unlock the full potential of Sana Agents. Whether you're building sales messages, extracting insights from meetings, or updating your CRM, mastering prompting is a skill that compounds with every message you send.
Key information
The CARE Framework: Your foundation
All effective prompts follow the same structure. We call it CARE:
C - Context
Why are you asking this question? What's the situation, background, or problem that led you here? Sana needs this to understand your world.A - Action
What specifically do you want Sana to do? Analyze? Summarize? Generate? Create? Be direct about the task.R - Result
What does success look like? What outcome are you after? What should Sana optimize for?E - Example
Show Sana what you mean. An example is worth a thousand words—it eliminates guesswork and aligns expectations.
See it in action
Weak prompt (no CARE):
"Tell me about customer feedback"
Strong prompt (with CARE):
Context: "We launched a new product last month and need to understand customer sentiment"
Action: "Analyze all feedback from the past 30 days"
Result: "Identify the three biggest pain points customers mention"
Example: "Pain points might look like: slow onboarding, confusing interface, or lack of mobile support"
The second prompt gets you exactly what you need in one shot. The first one? You'll be clarifying for the next three messages.
The 10 principles of effective prompting
Beyond CARE, these principles guide every prompt you write:
1. Be specific, not vague
Vagueness is the enemy of good prompts. The more details you provide, the better the response.
❌ "Write about solar energy"
✅ "What are the main advantages and disadvantages of residential solar energy systems in northern Europe, specifically considering climate and installation costs?"
2. Provide real context
Sana can't read your mind. Tell it what you're trying to accomplish and why.
❌ "Suggest marketing strategies"
✅ "We're a B2B SaaS company targeting mid-market finance teams. Our biggest challenge is brand awareness. Suggest three marketing strategies we can execute in the next quarter with a team of four people."
3. Choose your tone intentionally
The same information can sound formal, casual, technical, or conversational. Specify which one you need.
✅ "Draft a formal email to a client explaining a contract renewal"
✅ "Write a casual Slack message reminding the team about tomorrow's brainstorming session—make it fun"
4. State your output format upfront
Don't let Sana guess how you want the information structured. Tell it.
❌ "Summarize this report"
✅ "Summarize this report in three bullet points, focusing on revenue impact, timeline, and next steps"
5. Focus on what you want, not what you don't want
"Don't write a long summary" is less effective than "Write a concise summary." Positive framing is clearer.
❌ "Don't include technical jargon"
✅ "Explain this in simple, non-technical language"
6. Break complex tasks into smaller pieces
When you have a multifaceted question, don't ask it all at once. Sequence it.
❌ "Explain climate change, its causes, effects, and solutions"
✅
First: "What are the main drivers of climate change?"
Then: "How does climate change affect agricultural productivity?"
Then: "What are the most effective mitigation strategies?"
7. Use examples as your secret weapon
Showing is better than telling. When you provide an example of what you want, Sana understands instantly.
❌ "Write a product announcement"
✅ "Write a product announcement. Here's an example of the tone and structure I want: 'We're thrilled to introduce [Feature], a game-changer for [User Type]. With [Feature], you can [Benefit] in [Timeframe].'"
8. Guide with keywords and phrasing
Certain words and phrases steer AI responses in specific directions. Use them intentionally.
✅ "Create a slogan that emphasizes sustainability, innovation, and accessibility"
✅ "Write about time management using techniques like prioritization, delegation, and scheduling"
9. Give clear step-by-step instructions
When your task has multiple parts, define each step explicitly. This ensures nothing gets missed.
❌ "Write about product features and benefits"
✅ "Step 1: List the top three features. Step 2: For each feature, explain how it benefits users. Step 3: Add a summary of why these benefits matter together."
10. Build in iteration from the start
You don't have to get it right the first time. Plan to ask follow-up questions to refine, deepen, and enhance Sana's responses.
✅ First prompt: "Summarize the key findings"
✅ Follow-up: "Can you provide a cost estimate for implementing each recommendation?"
✅ Follow-up: "Which of these would have the highest impact in Q1?"
How-to guidance
Real-world prompting scenarios
Scenario 1: Building a personalized sales message
Your goal: Create a compelling outreach message tailored to a specific prospect.
CARE structure:
Context: "I'm reaching out to a prospect at a Fortune 500 company in the fintech space who recently got promoted to VP of Operations. I found their LinkedIn profile and company website."
Action: "Write four different outreach options for me to choose from."
Result: "Each message should feel personal and targeted—not generic. The goal is to get them interested enough to take a brief 15-minute call."
Example: "The opening line should reference something specific they've done or their company has achieved, not generic praise like 'I was impressed by your work.'"
Additional constraints:
Keep LinkedIn messages under 90 words
Keep emails under 120 words
Avoid clichés like "I hope this message finds you well"
Include a subtle pain point that Sana solves
Use a conversational, C-suite tone (professional but laid-back)
Scenario 2: Extracting insights from a meeting
Your goal: Transform messy meeting notes and a recording into structured insights.
CARE structure:
Context: "I just had a discovery call with a prospect. Here's the recording and my notes. The prospect is from a mid-market healthcare company exploring a new learning platform."
Action: "Create organized bullet points on five specific areas."
Result: "I need to brief my team tomorrow, so it should be clear and actionable."
Example: "For current challenges, don't just list them—explain why they matter to the prospect's business."
What to include:
Their current learning setup and tech stack
Specific pain points and why they're looking for change
What they're hoping to achieve
Realistic timeline and decision process
What I should demonstrate in the next meeting
Scenario 3: Updating your CRM with prospect information
Your goal: Take conversation details and structure them for your sales system.
CARE structure:
Context: "I just discussed a prospect's needs during our call. I need this information structured for Salesforce using the BANCT framework."
Action: "Extract and summarize Budget, Authority, Need, Competitors, and Timeline information."
Result: "Each field should be 1–2 sentences, under 250 characters, with reasoning for what I wrote."
Example: "Budget might read: 'They need 200 licenses. Budget is $50K–75K based on their headcount comment. They mentioned quarterly budget reviews.'"
Key constraints:
Write everything in English (for system compatibility)
Include your reasoning for each statement
If information is missing, write "TBC" (To Be Confirmed)
Focus on accuracy over completeness
Pro tips for advanced prompting
Tip 1: Use variables for reusability
Create template prompts with {{PLACEHOLDER}} variables so you can reuse them across similar tasks.
Tip 2: Combine CARE with your constraints
After your CARE structure, add any specific rules or limitations: "Also, please only mention partners listed on our website" or "Keep technical jargon to a minimum."
Tip 3: Stack examples
Don't just give one example. Give 2–3 examples of the output format or tone you're looking for.
Tip 4: Build prompts iteratively
Your first prompt won't be perfect. Write it, get a response, refine it based on what you see, and iterate. Each round gets you closer to exactly what you need.
Tip 5: Leverage your profile
The more Sana knows about you (your role, industry, goals), the more tailored its responses. Keep your profile settings updated.
Impact and use cases
Prompt engineering isn't magic. It's a skill. And like any skill, it improves with practice. Every prompt you write teaches you something. Every response Sana gives you is feedback.
The more intentional you are with your prompts, the more powerful your results. Whether you're accelerating your sales process, synthesizing complex information, or organizing your workflow, strong prompts save you time and unlock better outcomes. Start with CARE, follow the 10 principles, and iterate. You'll be amazed at what you can accomplish.
FAQ
Q: What if my first prompt doesn't work?
A: That's expected. Iteration is built into the process. Ask follow-up questions, clarify what worked and what didn't, and refine. Each round gets you closer.
Q: Should I always provide an example?
A: Yes, whenever possible. Examples eliminate ambiguity and align expectations instantly. If you can't think of one, describe what success looks like in vivid detail.
Q: Can I use prompt engineering for creative tasks?
A: Absolutely. Creative work benefits tremendously from clear context, specific tone direction, and detailed examples. Be even more explicit about style, mood, and desired output.
For further questions or information about Sana Agents, please contact [email protected] via email.
