What the ICF guidance says
The ICF-Aligned Guidelines for the Use of Artificial Intelligence (AI) in Mentor Coaching set out nine principles that align AI use with the ICF Code of Ethics and Core Competencies. AI is framed strictly as a supportive tool: it can help a mentor coach prepare, learn and reflect, but the evaluation of competency, the completion of forms, the safeguarding of confidentiality and the professional judgment behind any credentialing recommendation remain human responsibilities.
Calibrate is a developmental mirror for coaches and the mentor coaches who guide them — never an assessor, never a substitute for the mentoring relationship. That positioning is what makes alignment with this guidance natural rather than retrofitted.
Source: ICF Credentials & Standards — Guidelines for the Use of AI in Mentor Coaching. Published on ICF Engage. ICF member access required. The summaries below are Calibrate's own; the source document is ICF's and is not reproduced here.
Principle by principle
Each ICF principle, in brief, alongside the specific way Calibrate is built to honour it.
| ICF principle | How Calibrate is aligned |
|---|---|
| 1Evaluation of ICF Core Competencies AI supports, but cannot independently evaluate coaching competency. | Calibrate produces a developmental read, never an official assessment. It is positioned as a mirror for the coach and the mentor; results are expressed as a three-state readiness (at standard / approaching / not yet), never as pass/fail. Human discernment stays central. |
| 2Role of AI in mentor coaching AI must not complete mentor coaching forms or be the sole source of evaluation. | The ICF Session Observation and Competency Review forms are generated only as editable drafts. Identity and attestation fields are left blank for the mentor, and nothing is filed automatically. The work must remain the coach's own. |
| 3Professional responsibility AI-generated content must be reviewed, refined and validated by the coach. | A mandatory human-review attestation gates every form export: the mentor must confirm they have reviewed the draft, applied their own judgment, and adopt it as their own before it can leave the app. |
| 4Confidentiality & data protection Identifiable or sensitive data must be anonymised and kept secure. | A compulsory anonymisation step precedes every text upload; original transcripts and audio are not stored beyond the processing call; all processing runs over secured, authenticated channels. |
| 5Informed consent Explicit consent is required from all parties before AI is used. | A consent gate on every upload — audio and text alike — routes by whose session it is and requires written-consent attestations that explicitly cover AI processing, for the coach and, where relevant, the coachee. |
| 6Recordings & transcripts Public or unsecured AI platforms must not be used for such materials. | Calibrate reaches Claude and Gemini only through their commercial / paid APIs with binding no-training terms and data-processing agreements — never the public consumer apps. API keys are held server-side; see our Privacy Policy. |
| 7Ethical use of technology AI outputs are supportive suggestions, not authoritative conclusions. | Every output is framed as developmental guidance, not a verdict, with disclaimers throughout the report and the forms. Calibrate never issues a determination. |
| 8Transparency Be transparent about AI use, especially in credentialing contexts. | Reports and form drafts carry an explicit “AI-assisted draft” disclosure; the Privacy Policy includes a dedicated AI-transparency note aligned with EU AI Act Article 50; and this page states our posture openly. |
| 9Best practice Use AI sparingly and intentionally; keep practice human-centred. | Calibrate is designed as a preparation and reflection tool that feeds the human mentoring conversation — a structured reference before and around mentoring and supervision, not a shortcut that replaces it. |
What we deliberately do not claim
Credibility with experienced mentor coaches comes from precision, not superlatives. So, plainly:
- We do not claim zero data retention. Our providers operate on binding no-training terms; content may be briefly retained under their commercial terms (for example, Anthropic's 30-day abuse-monitoring window) and is never used to train their models. We state exactly this in the Privacy Policy rather than overstating it.
- We do not claim to evaluate competency. Calibrate offers a developmental read; the ICF competency evaluation itself remains the mentor coach's.
- We do not claim ICF endorsement. This page is Calibrate's own alignment statement with ICF's published guidance — not an approval, certification, or partnership with ICF.
Related
This page covers ICF's guidance on AI in mentor coaching. For how Calibrate aligns with the broader frameworks — the EU AI Act, the ICF AI Coaching Framework and Standards, the ICF Coaching Platform Standards, and the EMCC Digital Ethics Guidelines — see AI standards. For the values behind the product, see our Ethical commitment and the Manifesto.