Buyer's guide
How to evaluate an AI coaching partner. Six questions your legal, HR and procurement teams will ask.
Whether you run coaching programs for client organizations or bring AI coaching into your own platform, these criteria decide whether a partner passes review. Vendor-neutral, with the questions to send and the answers to look for.
The short answer
Ask for documents, not adjectives
Ask every AI coaching partner six things: where its SOC 2 stands, who owns your data, where that data is stored and processed, whether it trains models on your data, whether one studio can run several coaching experiences, and whether you can deliver it under your brand or through an API.
Good partners answer with reports, contracts and named providers. Vague answers such as "enterprise grade" or "GDPR-ready" are where reviews stall.
Last reviewed: November 6, 2026.
The criteria
Six criteria for choosing an AI coaching partner
What to ask, what a good answer includes and what should make you pause.
| Criterion | What to ask | A good answer includes | Red flag |
|---|---|---|---|
| SOC 2 status | Is there a SOC 2 Type II report or an audit in progress? Which systems are in scope? | A report you can read under NDA, or a dated audit plan with the auditor or platform named | "SOC 2 compliant" without a report or a date |
| Data ownership | Who owns the content you upload and the conversations coachees have? Can you export and delete them? | Your methodology and data stay yours, in writing, with export and deletion when you leave | Terms that let the partner reuse your content to improve its service |
| Data location | Where is data stored, where does the model run and which subprocessors are involved? | A named cloud provider and region, a subprocessor list and a data processing agreement | "GDPR-ready" with no region, subprocessor list or agreement |
| No training on your data | Are your content and coachee conversations used to train any model, the partner's or a model provider's? | A clear no in the contract, covering the model providers too | A no on the website and a "may use to improve" in the terms |
| Multi-experience studio | Can you build and run several coaching experiences, for different programs, cohorts or clients, from one place? | Separate experiences with their own method, scope and guardrails, managed centrally | One assistant per account, rebuilt for every client |
| White label and API | Can you deliver it under your brand, and can it be embedded in your product or platform? | Branding where coachees see it, and a documented API | Branding you can't change and no way to integrate |
Beyond the six
Four more questions worth asking
Quality: How is the coach tested against your definition of good coaching before launch, and after every change?
Escalation: What happens when a coachee raises a crisis or a topic outside coaching? Who is told, and how?
Consent: What can the coach, the program lead and the client organization see? Who decides what is shared?
Transparency: Under Article 50 of the EU AI Act, people must be told they are interacting with an AI system unless that is obvious. How does the partner make that clear? Article 50 on the AI Act Service Desk
Checklist
Documents to request from every partner
SOC 2 Type II report, or an audit letter with the expected report date
Status of ISO 27001 or other certifications, with the certificate if there is one
Data processing agreement and the current subprocessor list
Hosting provider and region, including where model inference runs
Data retention and deletion policy, including what happens when the contract ends
Model providers used and their training and retention terms
Summary of the latest penetration test or independent technical audit
How the coach is evaluated before launch, with an example
Branding options and API documentation
Send the same list to every partner on your shortlist. The speed and completeness of the answers tell you as much as the answers themselves.
Our own answers
How CoachBot answers these questions
We build an AI coaching platform, so here are our answers, with links to the details.
SOC 2 status
SOC 2 Type II observation period started October 2026 (report expected Q1 2027); ISO 27001 to follow. Independent technical audit completed July 2026.
Data ownership
Your methodology and data stay yours. A subprocessor list is available on request.
Data location
Hosted on Google Cloud in an EU region, with Gemini running via Vertex AI in the EU. GDPR and EU AI Act transparency compliance reviewed by heyData.
No training on your data
Your data is never used to train external models. Gemini via Vertex AI in the EU is the default; other model providers are used only on Scale, with your own API key.
Several experiences, one place
A separate workspace per client, each with its own CoachBots for different programs and audiences, tested against your definition of good coaching.
White label and API
Your name, logo and avatar as standard; brand colors and fonts on Team; your own domain and emails on Scale. Embed your CoachBot in your website, app or platform via API (Team and Scale).
FAQ
Evaluating AI coaching partners: common questions
- What should I look for in an AI coaching partner?
- Six things: SOC 2 status, data ownership, data location, no training on your data, a studio that runs several coaching experiences and white label or API options. Then ask how the coach is tested, how sensitive topics are escalated and who sees what.
- Is SOC 2 required for AI coaching?
- No law requires it, but many enterprise buyers ask for it. Ask whether there is a Type II report or an audit in progress, which systems are in scope and when the report is expected.
- What is the difference between SOC 2 Type I and Type II?
- Type I checks whether security controls are designed well at one point in time. Type II checks whether they actually worked over a period of several months, which is why buyers usually ask for Type II.
- Does the EU AI Act apply to AI coaching?
- The transparency duties of Article 50 apply since August 2, 2026: people must be told they are interacting with an AI system unless that is obvious. Whether further obligations apply depends on how the system is used, so ask the partner how it classifies its product.
- Should a coaching company build or buy AI coaching?
- Build if AI coaching is the product you sell and you have a team to own it for years. Otherwise buy the governed layer and keep your focus on the methodology. Our build vs buy guide lists the seven cost categories to price in.
- What does white label AI coaching mean?
- AI coaching that your clients experience under your name rather than the platform's. Ask exactly which parts carry your brand (name, logo, domain, emails) and whether that is standard or on request.
You own the method and the brand. We own the guardrails, security and compliance.
Put CoachBot through these six questions in a call with our team.