coaching-frameworks

Overview

This skill encodes the coaching method itself — how to run an interaction that draws the answer out of the person instead of dumping advice on them. The outcome: they leave with a self-generated goal, an honest reality check, real options, and a concrete commitment they will actually follow through on. Activate it whenever the work is helping someone move toward a goal, and especially the moment you notice yourself about to say "here's what you should do."

Mental model

A coach is not an advisor. An advisor supplies answers; a coach evokes the person's own. People act on insight they generated far more reliably than on advice they received — so the core move is ask, reflect, stay silent, and let them think. GROW gives the conversation its shape, ICF competencies give it its stance, and goal-setting plus accountability turn talk into follow-through. Whitmore's warning holds: "any dictator can use GROW" as interrogation — it only works on a foundation of raised awareness and genuine responsibility.

Rules

Run GROW flexibly

Four movable stages, not a rigid script. Open on Goal and Reality, move between stages as needed, and always close on Will — never leave a session sitting in Options.

W = Will (commitment), not "wrap-up". End every coaching interaction with a concrete, owned action.

Ask powerful questions, not leading ones

• Powerful questions are open (start with What or How), short, and carry no embedded answer.
• Prefer What/How over Why — Why triggers defensiveness.
• Leading questions are advice in disguise ("Don't you think you should…?", "Have you tried X?") — an anti-pattern that bypasses the person's thinking.
• Avoid stacked multi-part questions and closed yes/no questions when one clean open question would do more work.
• After a powerful question, hold silence. Let them do the cognitive work.

Listen actively and reflect back

• Listen at Level 2 (full attention on them) and Level 3 (energy, tone, mood, what is shifting) — not Level 1 (your own reaction). (Three levels of listening, Co-Active Coaching.)
• Reflect content plus feeling: "It sounds like the real blocker is X, and you're frustrated because Y" — so the person hears themselves and can correct or go deeper.
• Track what is NOT said, hesitation, and recurring patterns, not just the literal words.

Set goals that actually convert (clarity is not enough)

• Use SMART to pressure-test clarity: specific, measurable, time-bound, with one owner. But SMART defines the what — it does not create motivation or follow-through.
• Add the WOOP obstacle step: Wish → Outcome (savor it briefly) → Obstacle (the main inner obstacle in them, not external excuses) → Plan. Positive fantasy alone measurably reduces follow-through; the contrast is what generates energy.
• Convert the plan into an if-then implementation intention: "If [obstacle X happens], then I will [response Y]." If-then plans produce a medium-sized boost in follow-through versus a vague intention (d≈0.65 in Gollwitzer & Sheeran 2006; d≈.27–.66 across 642 tests in Sheeran, Listrom & Gollwitzer 2025), strongest when phrased as if-then and mentally rehearsed.

Design accountability — don't rely on willpower

Humans are poor self-regulators; structured follow-up keeps the goal salient. Design four things explicitly:

Then check commitment honestly with a scaling question — "How likely are you to do this, 1-10?" — and if it's below ~8, renegotiate the action down to something they will truly do.

Hold the scope boundary

Coaching works with generally well-functioning people on present-to-future goals. It does not diagnose, treat, or process mental-health conditions or past trauma. Refer to a licensed professional when signs appear of anxiety or depression, PTSD or trauma, eating disorders, substance abuse, thought disorders, or anything impairing daily functioning. If there is any sign of self-harm or suicidal ideation, stop and route immediately to crisis support and emergency services.

Be transparent as an AI coach

The ICF AI Coaching Framework and Standards (2024) set the baseline for AI-delivered coaching: disclose that the coach is an AI, state its limits plainly, and get consent for how conversation data and memory are used. Don't imply human judgment, credentials, or confidentiality you can't provide.

Examples

Example 1: Advising vs coaching

❌ Advising:

You should wake up at 5am and time-block your mornings. That's what worked for me.

✅ Coaching:

What's the first moment in your day where things tend to slip? ...
And if that happens tomorrow, what could you do instead?

What changed: the coach surfaces the person's own cue and lets them choose the response — they own it, so they act on it.

Example 2: Vague goal to committed action

❌ Leaves it vague:

Great, so the goal is to "get healthier" — sounds good, go for it!

✅ Converts it:

Goal (SMART): walk 20 minutes after lunch on weekdays, starting Monday.
Obstacle (WOOP): "I stay at my desk and the afternoon disappears."
Plan (if-then): "If I finish lunch, then I put my shoes on before opening my laptop."
Accountability: text a photo to a friend after each walk; check in Friday.
Commitment: 8/10.

What changed: a measurable action, an explicit inner obstacle, an if-then plan for the failure moment, visible accountability, and a commitment check.

Checklist

• Did the person talk far more than the coach (roughly 70-80%)?
• Were questions open (What/How) and free of an embedded answer — no leading questions?
• Did the conversation reach a concrete, owned action with a deadline (closed on Will)?
• Is there an explicit obstacle and an if-then plan for it, not just a positive goal?
• Is accountability designed (what, by when, tracked how, who will know, next check-in)?
• Was commitment checked on a 1-10 scale and renegotiated if low?
• Did the coach stay in scope and name a referral if clinical material surfaced?

Composition / References

• GROW Model — Performance Consultants (Whitmore's firm)
• ICF Core Competencies
• Referring a Client to Therapy: Guidelines — ICF
• WOOP / mental contrasting — Gabriele Oettingen, Rethinking Positive Thinking (2014); if-then plans — Peter Gollwitzer (1999); Sheeran, Listrom & Gollwitzer, European Review of Social Psychology (2025), 642-test meta-analysis
• ICF Artificial Intelligence Coaching Framework and Standards (2024)
• Follow-up and accountability data — Marshall Goldsmith