Your AI Agent's prompt is the set of instructions that shapes how it talks to leads and members: its personality, its goal for the conversation, and the rules it follows along the way. This article covers how to open and edit your AI Agent's prompt, how to write instructions it will actually follow, and how prompt settings differ from other AI Agent settings like training and goals.
Before you begin
Your Gym Lead Machine account needs Premium or AI features enabled to access AI Agents. If you don't see AI Agents in your account, reach out to Support to confirm your plan includes them.
You should already have at least one AI Agent created. This article covers writing and refining its prompt, not creating an agent from scratch.
To create your first Conversation AI Agent, see the article Create a Conversation AI Bot
What you can do with your AI Agent's prompt
Set the agent's role, tone, and communication style
Define the main goal the agent should work toward in a conversation
Add rules for tone, escalation, question order, and what to do when it doesn't know an answer
Pull in dynamic details, like a contact's name or your business information, using custom values
Test your changes against sample messages before they reach a real lead or member
How do I open my AI Agent's prompt settings?
Go to AI Agents.
Open the Agents List to see the AI Agents on your account.
Select the agent you want to update by clicking the three vertical dots, then select Edit, or create a new one.
Click on “Bot Goals”
NOTE: Changes to the prompt aren't saved automatically. If you leave the page before saving, your edits are lost.
How do I write the Personality, Goal, and Additional Information fields?
These three fields work together as one prompt. Each shapes a different part of how the agent responds, so write them as a set rather than in isolation.
Personality. Defines who the agent is and how it should sound. Use this field to set its role and tone.
For example: You are a friendly membership assistant for the gym. Keep responses warm, helpful, and concise.
Goal. Defines the main outcome the agent should work toward in the conversation.
For example: Your goal is to guide the contact toward booking a tour or intro session.
Additional Information. Gives the agent supporting rules, context, and guidance for specific situations, such as escalation rules or how to handle questions it can't answer.
For example: Ask one question at a time. If the answer isn't available in your training, don't guess. Let the contact know a team member can follow up.
Fill in or update the Personality, Goal, and Additional Information fields.
Use Add Custom Value to pull in dynamic details, such as a contact's name or your business information, instead of typing them manually.
Review the full prompt for conflicting instructions before saving. For example, telling the agent to keep replies under one sentence while also asking it to explain every service in detail will produce inconsistent responses.
Click Save.
TIP: Check the available word count in the prompt area before adding long instructions. A shorter, focused prompt is easier for the agent to follow consistently than a long one.
What makes a prompt effective?
A good prompt gives the agent three things: a role, a task, and guidelines.
Role. Who the agent is and how it should carry itself in the conversation. A vague role ("assist customers") produces vaguer responses than a specific one that also says how to acknowledge what the contact says and how to refer to them.
Task. The specific outcome you want and the steps to get there. A vague task invites the agent to guess at what you actually want; a script-like sequence of questions and conditions gives it less room to improvise.
Guidelines. The guardrails the agent should stay inside, such as reply length, tone, or when to stop and hand off to a person.
A few practices make these three elements work harder:
Repeat what matters. If a rule is important, state it in more than one field. For example, referencing your booking goal in both Goal and Additional Information reinforces it.
Show, don't just tell. When you want a specific tone or phrasing style, give examples of what to say and what to avoid, rather than describing the style in the abstract.
Keep static context short. If you add background information about your business or offers directly in the prompt, keep it brief. A long block of context can outweigh your actual instructions, making the agent focus more on the context than on what you told it to do.
Set the scene when it helps. For agents used in a specific situation, like a follow-up after a missed session, telling the agent why the contact is reaching out helps it respond appropriately.
Test and refine. Effective prompts are rarely right on the first save. Send sample messages through the testing tool, see how the agent responds, and adjust.
Prompts and instructions vs. other AI Agent settings
Setting | What it controls | Use it for |
|---|---|---|
Personality, Goal, and Additional Information (the prompt) | How the agent behaves, sounds, and handles situations | Tone, escalation rules, question order, response style |
Training | The factual information the agent draws on to answer questions | FAQs, business details, policies |
Goals | The specific outcomes and actions the agent works toward | Booking appointments, collecting contact details |
Response style settings | Reply length and format | Keeping answers short or long |
Frequently asked questions
Are prompt instructions the same as training my AI Agent?
No. Instructions guide how the agent behaves, while training gives it the factual information it uses to answer questions. If you want the agent to know something about your gym, add it to training. If you want to change how it talks or handles a situation, use the prompt.
Should I put frequently asked questions in the Additional Information field?
No. FAQs and factual business information usually belong in training, not in the prompt. Use Additional Information for behavior rules and response guidance instead.
Where do I control how long the agent's replies are?
Use response style settings if your account has them, or add a direct instruction such as "keep replies under three short sentences" to Additional Information.
Why is my AI Agent ignoring my instructions?
The instruction may be too vague, may conflict with another rule in the prompt, or may be trying to control something that belongs in a different setting, like training, goals, or response style. Review the full prompt for conflicts and make sure the instruction is specific.
Can I use the prompt to control appointment booking?
The prompt can guide how the agent talks about booking, but the actual booking behavior is configured separately in the agent's goals.
Should I use goals or the prompt to collect contact information?
Use goals (or the relevant agent action) to define what information should be collected, and use the prompt to explain when and how the agent should ask for it.
What should I tell the agent to do when it doesn't know an answer?
Add an instruction such as, "If the answer isn't available in your training, don't guess. Let the contact know a team member can help." This keeps the agent from inventing answers it isn't equipped to give.
Do I need to test the agent every time I update the prompt?
Yes. Test with realistic messages after every update to confirm the agent behaves as expected before it talks to a real lead or member.
Helpful tips
Write the Personality, Goal, and Additional Information fields as one set of instructions, not three separate notes.
Use custom values instead of typing repeated details like a contact's name or your business information.
Keep any background context you add to the prompt short so it doesn't outweigh your actual instructions.
Review the whole prompt for conflicting rules before saving, especially after several rounds of edits.
Test after every change with messages that reflect how real leads and members actually write.
When this feature is most helpful
Setting up a new AI Agent and giving it a consistent voice from the start
Fixing an agent that's giving inconsistent or off-brand responses
Adding escalation rules so the agent knows when to hand off to a team member
Adjusting how the agent asks questions during a lead qualification conversation
Refining a prompt after testing reveals unexpected behavior
If you have questions about AI Agents in Gym Lead Machine, our Support team is happy to help. Reach us at hello@usekilo.com.