Step 2.8: How to Set the AI Agents

Customizing AI Agents for Your Coaching Quests

🔑 Overview: What Are AI Agents?

AI Agents in Rocky.ai White Label empower you to fully customize the behavior and personality of the AI coach per Quest. This means:

How the coach interacts with users
What kind of questions it asks
How it responds to users
What kind of outcomes it drives
How it guides, supports, or challenges your users

You’re not just using AI—you’re designing your own sets of coaches and mentors.

✅ Key Feature: Each custom Quest can include its own tailored AI Agent configuration. Once configured, Rocky will automatically apply these behaviors and styles to the user experience for that Quest.

✨ Why It Matters

This advanced setup is what transforms your brand’s coach from a generic assistant into a highly personalized expert, aligned with your program’s purpose and coaching goals. Your AI Agent becomes a digital embodiment of your methodology, values, and tone.

⚠️ Note: AI Agent customization is only available for your own Quests and Programs. You cannot modify Rocky’s Essentials Programs, but you can draw inspiration from the templates provided there, when adding new Quests with a template.

🛠️ Agent Configuration Guide

To create the perfect coaching experience for your clients, you can fine- tune how the AI coaches and mentors.

To do this, please go to Open Creator Tool.

Please then select AI Agents

Then please select Add/ Manage Questions to the mission you want to work on, as below:

Now you should see a screen like this:

Click on Show Section on the quest you want to work with.

You will now see:

When you click on the red pencil in each of the boxes, you should then see this:

By clicking on the downward arrow, you will be presented with an array of options.

Please find below a description of all the parameters you can set to customise your coaching experience.

Main role of the Quest:

This will help determine the content that appears, the type of language used and the overall tone of the chat.

Define the purpose and tone of the AI within this Quest.

Is the coach here to motivate, inspire, challenge, or guide?
This influences content, language, and emotional framing.

🧠 Example: “This Quest helps users reflect on failure to uncover hidden strengths.”

Solutioning of the AI:

This will determine the specific outcome of the chat and the way in which these outcomes are achieved.

Define how the AI delivers value or outcomes.

Should it guide the user to a solution?
Should it offer direct suggestions?
Or should it focus on user discovery and reflection?

🎯 Example: “Help the user craft 3 action points toward their weekly goal.”

Initial question when the quest is started:

This is the first question your user will be asked when embarking on a quest, so it’s worth choosing carefully.

This is the first question a user sees when they start the Quest.

Should be engaging, clear, and set the tone.
Choose carefully—this sets the direction of the conversation.

💬 Example: “What’s the biggest challenge you're facing this week?”

Questions asked by Agent:

This will determine the type of questions that the user receives. The questions can be more in-depth, more goal-oriented or more set to challenge the user’s specific understanding of a concrete subject. You get to choose the type and tone of question.

Define the types and tone of coaching questions during the chat.

More open-ended, challenging, or supportive?
Can be goal-driven, exploratory, or self-assessing

🧭 Examples:

“Why do you think this keeps happening?”

“What’s one small step you could take today?”

Immediate AI response:

This sets the overall type of response the AI will give the user, as well as the tone the response is given in. You can choose from a variety of responding styles, whether looking for a positive and encouraging tone, an assertive tone to elicit compliance, or an inquisitive response that pushes for greater user autonomy and connects the current chat to past goals.

Define how the AI responds immediately after a user message.

Choose a tone: encouraging, direct, inquisitive, etc.
Tailor for your brand voice and the Quest’s goal.

🎤 Examples:

“That sounds tough—let’s look at it together.”

“Here’s a quick technique that might help…”

Mentoring and Critical Thinking:

This will determine the style of mentoring your user receives. You can choose a more directive style of mentoring or a style that focuses on building the autonomy of the user. There are many other nuances to be explored.

Choose how the AI develops the user’s thinking.

Directive mentoring: gives answers, sets structure.
Socratic mentoring: asks questions, builds user autonomy.

You can mix tones: empathic, challenging, educational.

🌱 Example: “Instead of telling the user what to do, help them explore two paths and decide.”

Guardrailes of AI:

This will allow you to decide what kind of responses you want to avoid. You can choose to avoid finding solutions for the user, being overly directive, or too sympathetic. You are creating the ‘personality’ of the coach and mentor here.

Control what the AI should avoid doing.

Avoid being too solution-focused
Avoid being too directive or too sympathetic
Reinforce the identity and boundaries of your coach

🚫 Example Settings:

“Avoid giving a direct answer to emotional dilemmas.”

“Do not diagnose or offer medical advice.”

Background data for AI:

This will define the context and parameters used for a solution path.

Specify what the AI should aim to collect or analyze after each session.

Useful for creating summary reports or progress tracking
You can define desired data points such as mood, goals, blockers, etc.

🧾 Example: “Capture the user’s top priority this week and save it to history.”

Fallback Questions

Add additional coaching questions to enhance or support the conversation if the flow stalls.

Used to enrich the user experience or provide a backup
Great for maintaining engagement in longer sessions

🔁 Examples:

“What would success look like for you by Friday?”

“Can you recall a moment where you overcame something similar?”