What Is Prompting?
Prompting means giving the AI clear instructions.
A strong prompt removes guesswork by explaining what the AI should do, what context it should consider, and what the final output should look like.
A good prompt usually answers five key questions:
Goal — What should the AI do?
Context — What background information should the AI consider?
Format — What should the output look like?
Role — What perspective should the AI take?
Constraints — What should the AI include or avoid?
Weak Prompts vs. Strong Prompts
The difference between a weak prompt and a strong prompt is the difference between guesswork and useful output.
Weak Prompt Example | Strong Prompt Example |
“Write a message to the prospect.” This prompt is too vague because it does not explain:
| “Write a short LinkedIn DM to a VP of Sales at a B2B SaaS company that recently raised a Series B. Reference their growth stage and ask one open question about how they are scaling their outbound team. Keep it under 60 words. Do not mention our product.” |
Because the prompt is unclear, the AI has to guess. | This prompt works better because it includes:
|
How Prompting Works in Alta
Alta AI uses your prompt together with multiple context layers.
These layers help the AI understand what to write, who it is writing to, and how the message should be structured.
Key Prompting Layers in Alta
Message Goal
The message goal explains the purpose of the message.
For example:
Start a conversation
Follow up after no reply
Ask a qualification question
Invite the prospect to book a meeting
Respond to an objection
Message Instructions
Message instructions define how the AI should write the message.
This can include:
Format
Tone
Length
Specific wording
What to include
What to avoid
Variables
Variables allow Alta to pull in structured prospect and company data.
Use variables when you want the AI to personalize messages using information like:
Prospect name
Job title
Company name
Company industry
Recent signal
LinkedIn activity
Important: Use Alta’s variable insertion option to add variables at the exact cursor position.
Global Instructions
Global Instructions apply across messages and campaigns.
They are useful for instructions that should always be followed, such as:
Brand tone
Words to avoid
Formatting preferences
Compliance rules
Messaging style
Global Instructions can also be toggled off per campaign when needed.
Pitch
The Pitch gives the AI background knowledge that can be used across messages.
This can include:
Company positioning
Value proposition
Product information
Target audience
Knowledge documents, if included in the pitch
Control Panel Tips
When working in Alta, use the control panel whenever possible.
Tone and Length
Use Alta’s control panel settings for tone and length instead of adding them only inside the prompt text.
This helps keep prompts cleaner and easier to manage.
Important note: If you write something related to the character count or message length in the message prompt, it may cause message misalignment because this information is already controlled in the control panel.
That is why we recommend using the control panel to define the tone and length of the message, without adding those instructions to the message prompt.
Variables
Always use the system’s variable insertion option to add variables.
This helps make sure the variable is added correctly and placed exactly where needed.
Best practice: When using variables, they become part of the prompt.
For example, you can write in the prompt: Start by saying, “Hi {First Name},” and select the First Name variable.
What Data Can Alta AI Use?
Alta AI can personalize messages using prospect, company, and signal data.
Prospect Data
| Prospect Data
| Company Data
| Signal Data
|
Alta AI can use information such as: | Alta AI can use information such as: | Alta AI can use company information such as: | Alta AI can also use signals such as: |
|
|
|
|
Using these data points helps the AI create more relevant and personalized messages.
Tip: When you want to prioritize signals, make sure to add this as part of the prompt.
For example: When searching for {{socialSignal}}, always prioritize the {{fundingAnnouncement}} signal.
If you cannot find {{fundingAnnouncement}}, check if there is a {{roleChange}} signal.
Why Punctuation Matters in Prompts
Punctuation is not only a style choice. It also helps the AI understand the structure and importance of your instructions.
Use punctuation deliberately to make your prompts easier for the AI to follow.
Useful Prompting Punctuation
Double Quotes: “ ”
Use double quotes when you want the AI to focus on exact text.
Example:
Rewrite this sentence: “We help sales teams book more meetings.”
Triple Quotes: """ """
Use triple quotes when pasting longer content, such as emails, transcripts, or documents.
This helps separate the pasted input from your instructions.
Example:
Review the email below and rewrite it in a warmer tone:
"""
Hi John,
Following up on my last email...
"""
Colons: :
Use colons to create clear sections.
Example:
Audience: VP of Sales
Goal: Start a conversation
Tone: Short, curious, and professional
Bullets and Dashes
Use bullets or dashes to separate instructions.
This makes the prompt easier for the AI to follow than a long paragraph.
Example:
Write a message that:
Is under 60 words
Sounds natural
Mentions the prospect’s recent funding
Ends with one open question
Capital Letters
Use capital letters for important, non-negotiable instructions.
Use them sparingly.
Example:
DO NOT mention our product in the first message.
Parentheses: ( )
Use parentheses to add extra context without making the main instruction harder to read.
Example:
Write a follow-up email to the prospect. They already attended a demo but did not reply afterward.
XML Tags:
<tag>
XML tags are useful for long or complex prompts.
They help separate instructions, context, and input.
Example:
<instructions> Write a short follow-up email. Keep it friendly and under 80 words. </instructions> <context> The prospect attended a demo last week but has not replied. </context> <input> Use the prospect’s company name and job title if available. </input>
Common Prompting Mistakes to Avoid
1. Being Too Vague
A vague prompt gives the AI too much room to guess.
Avoid prompts like:
“Write me something about sales.”
Instead, clearly explain what you want the AI to create.
2. Asking for Too Many Tasks at Once
When you ask the AI to do too many things in one prompt, the output quality may drop.
For best results, focus on one task per prompt.
3. Forgetting Tone and Format
Without tone or format guidance, the AI may choose a default style that does not match your brand or channel.
Always explain how the message should sound and what format it should follow.
4. Not Providing Enough Context
The AI needs context to produce relevant output.
Include details such as:
Who the message is for
What happened before
What the goal is
What data or signal should be used
5. Not Iterating
Great prompts are rarely perfect on the first try.
Test your prompt, review the output, and refine it based on what you see.
How to Format Your Prompts
Well-structured prompts are easier for the AI to understand and follow.
1. Use Clear Sections
Separate your prompt into sections.
Recommended sections include:
Role
Context
Task
Format
Constraints
Example:
Role: You are a helpful sales assistant writing on behalf of a B2B sales rep. Context: The prospect is a VP of Sales at a SaaS company that recently raised funding. Task: Write a short LinkedIn message to start a conversation. Format: Keep it under 60 words. End with one open question. Constraints: Do not mention our product. Do not sound salesy.
2.Use Bullets
Bullets make your instructions easier to follow.
Use one idea per bullet.
Example:
Write a message that: - References the prospect’s recent funding - Sounds curious and professional - Stays under 60 words - Ends with one question
3.Use Examples
Examples help the AI understand the style and structure you want.
Instead of only saying “make it casual,” show an example of what casual means.
Example:
Use a style similar to this: Congrats on the recent Series B. Curious how your team is thinking about scaling outbound this year without adding too much manual work.
4.Use Structure for Long Inputs
When pasting long content, such as emails, transcripts, or notes, wrap the content in triple quotes.
Example:
Summarize the following transcript into 3 bullet points: """ Paste transcript here """
This helps the AI understand where the input starts and ends.
Great Prompting Is a Loop
Prompting is an iterative process.
The best prompts improve over time as you test them, review the results, and make adjustments.
The Prompting Loop
Step 1: Write | Step 2: Run | Step 3: Review | Step 4: Refine |
Create a clear first version of the prompt.
Make sure it includes:
| Test the prompt in a real Alta workflow or campaign step. | Review the AI output. Check whether the message is:
| Adjust the prompt based on what you observe. You may need to refine:
|
Repeat this process until the output is consistent and useful.
Best Practices
Be Specific | Use One Task Per Prompt | Add Context | Use Variables | Add Constraints | Test Before Launching |
The more specific your prompt is, the less the AI has to guess. | Focus each prompt on one clear outcome. | Include the details the AI needs to write a relevant message. | Use Alta variables to personalize messages with prospect and company data. | Clearly explain what the AI should not do. For example:
| Always test the prompt before using it in a live campaign. Review the output and make changes if needed. |
Prompting is not a one-time setup. It is a loop: write, run, review, and refine.