
ICF, EMCC, SF Coach ethics and AI: adding an agent without betraying your framework
Confidentiality, client autonomy, non-substitution, transparency: how to add an AI agent to your practice without clashing with your code of ethics.

Dominique Vives
Strategic Advisor & Co-Founder of Vikl, 20 years at Microsoft
In short. An ICF, EMCC or SF Coach certified coach doesn't have to choose between their code of ethics and an AI tool. The code stays the higher reference: it's the tool that must bend to it, never the other way around. The sensitive points are well known: confidentiality, client autonomy, non-substitution, transparency. Vikl was designed and trained with coaches to respect them by design. And to decide with confidence, an ethics checklist beats a sales promise.
The code comes before the tool, not the other way around
Before we talk about Vikl or any other software, one thing has to be clear. Your ethical framework is not negotiable. The codes of ICF, EMCC and SF Coach existed before AI and will keep applying whatever technology you add to your practice.
A good tool never asks you to loosen that framework. It conforms to it. If it forces you into a compromise on confidentiality, on consent or on your coachee's autonomy, the problem isn't your ethics, it's the tool.
That's the angle of this article. Rather than selling you features, we take the principles at stake one by one and look at what a properly framed agent must do to stay on the right side of the line. Vikl serves as a concrete example, because it was designed and trained with coaches precisely to hold that line. But the grid applies to any tool you might evaluate.
The four principles at stake
Four obligations recur across the three codes. They're the ones an AI agent can reinforce, or betray.
Confidentiality. This is the central obligation, the one the whole relationship rests on. The coachee must be able to share whatever they want without fearing it will reach anyone, including you in certain cases, and above all the funding employer.
Client autonomy. The coachee stays the subject of their own support. They decide their goals, their pace, what they share. A tool must not strip them of that autonomy by making decisions for them or steering the relationship.
Non-substitution. Coaching is a human relationship. No tool replaces the alliance between a coach and their coachee, nor the coach's professional judgment. An agent can extend, never substitute.
Transparency. The coachee has the right to know what they're interacting with, how their data is handled, and the exact role of the tool in their journey. No grey areas.
These four principles structure everything that follows.
Why "the coach never reads the conversations" is an ethical point
This is the most important point, and the most often misunderstood. Many tools offer the coach access to the content of the exchanges, framed as a "plus": summaries, alerts, transcripts "available on request." On paper it looks convenient. In reality it puts you at odds with your own code.
If you can read what your coachee shares at 10 p.m. between two sessions, then confidentiality is no longer guaranteed by the architecture. It depends on your discretion. And a savvy coachee knows it: what can be read can be read, requested, one day demanded by a third party. Trust erodes.
Vikl treats this point as an obligation, not an option. You see your coachee's engagement, their activity, their attendance, the modules they completed, never the content of their conversations. This isn't a setting that could have been switched on, it's a native separation between two worlds, with no bridge between them. It's privacy by design. We detailed it in the confidentiality paradox, and the mechanics of steering by engagement in tracking engagement without reading conversations.
The consequence is ethically clean: no one, not you, not the employer, not the Vikl team day to day, accesses the intimate. You only see what a coach legitimately has reason to see.
Informed consent before assigning a license
A Vikl license is assigned to a coachee. That implies a step that isn't a formality: gathering their informed consent.
Informed means concrete. The coachee must understand what the tool is, what you see and what you don't, where their data lives, and the fact that they keep control. It's not a checkbox, it's a conversation. It fits naturally into the framing work you already do at the start of an engagement.
When the employer funds the licenses, this point is even more sensitive. The coachee must know that the company pays for the licenses without accessing anything of their exchanges. Transparency about that watertight wall is what makes the consent genuinely free. For the technical side of that wall, see where your data lives.
The coaching, therapy, advice boundary
This is a legitimate worry. An agent that converses with a coachee at all hours: won't it drift toward diagnosis, prescription, care? Spill from coaching into therapy or advice?
The answer lies in how the agent is framed. A well-designed agent stays in the coaching posture: it questions, it prompts reflection, it returns the coachee to their autonomy. It does not diagnose and does not prescribe a solution. It doesn't steer the relationship and doesn't substitute for your judgment.
Vikl is customized to your method: your vocabulary, your tone, your posture, your signature tools, whether that's DISC, NVC, Process Com, Gestalt or your own approach. But its foundations stay anchored in organizational psychology and mediation, not in care. When a topic falls outside the scope of coaching, the right reflex stays human: it's up to you, in session, to redirect if needed. The agent extends your framework, it doesn't invent its own.
Are you a coach? VIKL extends your support between sessions, in your method and respecting your coachees' confidentiality. Discover the coach offer.
Hosting, encryption, GDPR, AI Act: concrete guarantees
Ethics doesn't only play out at the level of principles. It takes shape in the infrastructure. A confidentiality promise is only worth as much as the tech that holds it up.
- Hosting in France. The AI runs on Azure France Central. The data doesn't leave the European Union.
- End-to-end encryption. The exchanges are protected across their entire path.
- GDPR compliance. The handling of personal data respects the European framework, which includes the coachee's rights over their data.
- AI Act compliance. The European framework on AI is respected, which matters for a tool meant to converse with people about sensitive topics.
These aren't marketing arguments, they're conditions of possibility. Without them, talking about confidentiality would be hollow. With them, the boundary between visible engagement and invisible content rests on something real.
The ethics checklist before adding an AI tool
Before adding any AI tool to your practice, run through these questions. If even one gets a bad answer, dig in before you sign.
- Confidentiality. Can the coach read the content of the conversations? The right answer is no, never, by design.
- Engagement vs content. Is what I see limited to engagement signals, with no access to the intimate?
- Employer. When the company funds the licenses, is it fully cut off from the content?
- Consent. Can I gather the coachee's informed consent before assigning a license, on the basis of clear information?
- Posture. Does the agent stay within coaching, without diagnosing or prescribing a solution?
- Non-substitution. Does the tool extend my relationship instead of replacing it? Do I keep control of the method and the relationship?
- Data. Where does the data live? Is it hosted in the EU, encrypted, GDPR and AI Act compliant?
- Compatibility. Is the tool compatible with my ICF, EMCC or SF Coach code, without asking me to loosen my obligations?
- Transparency. Can my coachee easily understand what the tool does and doesn't do?
This grid isn't specific to Vikl. It's precisely because it was designed to answer yes to each of these questions that we invite you to put it through the wringer.
In summary
Adding an AI agent to a certified practice requires no ethical compromise, provided you keep the code as the higher reference and choose a tool that bends to it. Confidentiality by design, informed consent, a coaching posture maintained, data hosted in France and compliant: these are verifiable conditions, not promises. Vikl was designed and trained with coaches to hold them, and extends your support without ever substituting for you. The checklist above lets you verify it, for Vikl as for any other tool.
Related articles
Will AI replace coaches? What the profession risks (and doesn't)
Will AI replace coaches? The real answer: it shifts the value of the job toward what it can't do. Relationship, presence, meaning. What changes, and what doesn't.
Tailoring your AI agent to your coaching method (DISC, NVC, Process Com, Gestalt)
Coaches' top fear about AI is standardization. Here is how an agent takes on your vocabulary, posture and signature tools, without ever deciding in your place.
Tracking your coachees' engagement without ever reading their conversations
How to track your coachees' commitment without monitoring what they say. VIKL's privacy by design, aligned with ICF, EMCC and SF Coach ethics.
