Manifesto
We are trained to hold a mirror for our clients.— The founding premise
Calibrate exists because someone has to hold the mirror for us.
Part I — What Calibrate is
Calibrate is an AI-powered evaluator of coaching practice. It reads a session — transcript or recording — against the ICF competency standards at ACC, PCC, and MCC level, and returns a structured, evidence-based developmental report.
It does not certify. It does not pass or fail. It does not replace human judgement. It holds up a rigorous, honest mirror so that a coach can see their own practice with new eyes — and then return to the work of becoming better at it.
It stands beside the coach and the mentor coach — never in their place. This is not a constraint on what Calibrate could become. It is the condition that lets Calibrate exist at all in a profession built on human relationship. A tool that smells of replacement will be rejected by the coaching world on instinct, and rightly so. Calibrate earns its place by knowing exactly how small its place is.
Everything below is not aspiration. It is a statement of what Calibrate already does and already refuses to do, made explicit so it cannot quietly drift.
Part II — The principles
Principle 0 — the principle that governs the rest
The ICF defines mentor coaching as a collaborative learning process through which coaches receive feedback based on observed or recorded sessions to support them in further developing their unique coaching style and coaching skills in alignment with the ICF Core Competencies. Two words in that definition mark Calibrate's boundary: the process is collaborative — a relationship, not an output — and what the coach receives is feedback — a human, dialogic act, not a verdict.
The ICF Mentor Coaching Competency Model names six competencies across four domains. Read against them, Calibrate's place is deliberately, almost startlingly, small: it touches fragments of one competency — C4, Formative Appraisals — and only its perceptual part: locating competency evidence in a transcript, mapping it to credential level, and tracking it over time. Everything that makes mentor coaching what it is — the collaborative relationship, the partnering, the space for reflection, the recognition of the person, the modelling, the guarding of ethical boundaries between roles — is precisely what Calibrate does not do and must not do.
Principle 1
The report is a resource for reflection, never a verdict. No output has any certificatory or official value for the ICF or any accrediting body. A developmental report is a mirror, not a sentence. Calibrate describes readiness — a mentor's read — never a pass/fail score.
Principle 2
Every session contains fragments of a real person's life, and that person is not in the room when we evaluate. So Calibrate requires full anonymisation of every transcript before upload, and explicit written consent before any audio is used. This is not a formality. It is an act of respect toward someone who trusted their coach.
Principle 3
There are no session databases, no transcript collection, no training of models on user content. Uploaded material is processed in real time and leaves no trace on our systems; only the generated report lives in the user's own account. If this ever changes, we say so plainly and before it takes effect.
Principle 4
An AI cannot perceive silence, sense the quality of presence, or hear what was left unsaid. It recognises linguistic and behavioural patterns in a text — and it does so with rigour. Coaching is more than what appears in a transcript. Using Calibrate well means knowing exactly where it ends. We never pretend otherwise.
Principle 5
Calibrate has no institutional affiliation and is owned by no certifying body. It was built by a coaching professional for coaching professionals — regardless of accreditation path, school of origin, or approach practised. The aim is to widen access to reflective practice, not to build new hierarchies or reinforce old ones.
Principle 6
Every change to the evaluation logic is documented and versioned — a public record, updated at the time of each deployment, not retroactively. An independent panel of five ICF MCC coaches will review Calibrate's evaluations on a semi-annual basis, rating the same anonymised sessions blind — without seeing the tool's output first; the inaugural cycle is scheduled for October 2026. Where systematic divergences appear, the model is refined. This is how the mirror stays true over time.
Living document
This manifesto is versioned alongside the product it describes. Every change to the evaluation model is logged publicly in the Calibrate model changelog, updated at the time of deployment. The version of this document in force is always the one published at this URL.
© 2026 Leading srl STP — Calibrate, professional coaching evaluator.