ICF Checklist for Evaluating AI-Driven Coaching Tools
Last updated: 12/22/2024
We followed the current ICF guidelines when creating and testing our AI tool, Protégé. To find out more about the ICF framework go to this link: ICF AI Guidelines
Selecting an AI-driven coaching tool that meets ICF standards ensures not only ethical compliance but also a high-quality coaching experience that is secure, trustworthy, and effective. By following this checklist, buyers and users can feel confident in their choice, knowing the AI tool is aligned with best practices in coaching. The checklist is also handy for software developers to ensure their AI Coaching System complies with the standards.
Transparency and Ethics
Clearly identifies itself as an AI tool (vs. a human coach).
At the start of each session and from time to time during the session Protégé reminds the user that it is an AI tool. It is learning to recommend a live 1:1 session with Ruth when needed as we develop it.
Discloses any limitations (e.g., not for mental health support).
This is covered in the coaching contracting phase between Pearce Insights and the client. We will be adding some language to the welcome that highlights what coaching is and is not.
Provides accessible information on how it makes decisions.
Protégé lets the user know as it goes along what it is doing. For example, reviewing notes from past sessions, considering the next question or prompt.
Demonstrates commitment to reducing biases in its responses.
Protégé is being trained and is learning about biases. We have an update planned that takes greater account of biases and cultural considerations.
Has a robust consent process and clearly outlines data handling practice.
This is available during contracting, at sign up during the self-service set up process, and is always available on our website. Policies will be regularly updated.
Client Relationship and Trust
Provides options for coaching agreements and includes explicit user consent for AI-based interactions.
Our contracting process addresses this and at the start of each session, Protégé reminds the user that this is AI coaching.
Offers clarity on AI capabilities and limitations.
Protégé frequently reminds the user that of its limitations. It has a safety protocol to handle situations where coaching may not be appropriate or there is concern for the person’s safety and wellbeing.
Includes access to human coaches when needed for additional support like complex or sensitive issues.
Protégé always has access to human coaches as does the coaching client. If a coaching session does not seem to be moving forward the way the client wants, Protégé will recommend an in person session. We will be adding the capability to offer the booking link to make an appointment there and then.
Effective Communication
Adapts responses based on user interactions (active listening).
Protégé is like a human coach in that it learns about you as it has more sessions with you and as long as your close out your sessions each time, it will update its user model to keep up with things it has learned. It will make connections between sessions and highlight themes that you may not have noticed. In this way the coaching experience with Protégé evolves as it gets to “know” you.
Provides feedback, reflection, and self-assessment prompts regularly.
Protégé is designed to use many of the same prompts and questions that I use in our sessions. It regularly reflects and provides feedback. Sometimes it may even feel as though it does too much of that. We are working on this to help it become more observant about how the process is going.
Uses respectful and culturally sensitive language.
Protégé is always learning. If at any time you find something it says inappropriate, please include feedback as you close out your session with Protégé. Protégé has access to many more inputs that a human coach could possibly access, so our expectation is that it will be sensitive.
Fostering Growth
Includes tools for setting and tracking client goals.
Protégé will always encourage you to reflect on goals you have set in prior sessions and will help you stay on track and keep you accountable. However, when you need a change, or you have a new priority you only need to tell Protégé that and it will adjust accordingly. You are in the driving seat of the car of your own life!
Offers reminders and reinforcement for goal progress
Yes, at the start of every session.
Provides insights that adapt to the client’s unique needs and inputs.
Yes, as Protégé learns more about what is helpful to you and what is not, it adapts.
Quality Assurance and Usability
Conducts ongoing testing, validation, and improvement.
Our team will continue to test Protégé, review feedback, select sessions for review and update the model as we learn more about what helps and what does not.
Offers content that is reviewed or developed by language and coaching experts.
Yes, this is part of the process that our partnership with Amigo ensures
Includes mechanisms for user feedback and continuous improvement.
Every session with Protégé ends with an invitation to give feedback and make suggestions. Any user can also email us at admin@allellc.org with feedback.
Security and Privacy –
Please see full information about Security and Privacy on our webpage and at the links below:
https://www.amigo.ai/legal/terms-conditions
Uses encryption for data storage and transmission Yes - https://docs.google.com/document/d/1tX8GYgC5dkq6JaFwRcnJ3KkjY4wX1N-3UNrbCke7768/edit?tab=t.0
Adheres to privacy laws like GDPR.
Yes - https://www.amigo.ai/legal/privacy-policy
Requires secure sign-on like multi-factor authentication.
Yes
Only collects necessary data and provides clear data policies.
Yes - https://www.amigo.ai/legal/privacy-policy
Offers users with control over their data access and sharing preferences.
Yes https://www.amigo.ai/legal/privacy-policy
Created with
