For two years, the coaching world has argued about AI in the session. Whether a notetaker should be listening. Whether a chatbot can really coach. That debate matters, but it left out something closer to home. Your client records can now talk to an AI assistant too.
In August and September, HoneyBook connected its CRM to both Claude and ChatGPT. Your pipeline, invoices, contracts, and client messages can now be pulled into a general-purpose AI assistant with a sign-in and a few clicks. That makes the new HoneyBook AI integration for coaches a data decision before it’s a productivity one. The useful question isn’t whether to use it. It’s what you decide before you switch it on.
What HoneyBook actually shipped
On August 19, HoneyBook announced HoneyBook MCP as a Claude connector. MCP stands for Model Context Protocol, the open standard Anthropic released in November 2024 for connecting AI tools to the places your data lives. Three weeks later, on September 9, HoneyBook launched on ChatGPT as part of OpenAI’s new Small Business Collection.
Both versions do more than answer questions. They can read your records and act on them. Here’s how the two compare, based on HoneyBook’s announcements and its setup guide for the Claude connector.
| Claude connector | ChatGPT app | |
|---|---|---|
| What it can read | Pipeline and leads, projects, invoices, contracts, client messages | Clients, projects, schedules, contracts, invoices |
| What it can do | Create and update projects and leads, add notes, create and send invoices and contracts from your templates, issue payment requests | Draft inquiry replies, proposals, invoices, and contracts, add contacts, build lead forms, prepare payment requests |
| Before anything goes out | You approve permissions resource by resource. The announcement doesn’t describe a separate draft step. | Everything arrives as a draft you review and approve before it’s sent |
| How to turn it off | Claude, then Customize, then Connectors, then Disconnect | Revoke access in your ChatGPT settings |
The appeal is easy to see if admin drains you. Helena Bachar, a senior product manager at HoneyBook, described it this way in the Claude announcement.
“Giving our members a way to ask their own business a question – and get the answer from what’s actually in their account, not a guess – changes how the workday starts.”
Asking “which invoices are overdue?” and getting a straight answer is genuinely useful. It also means your client records now travel somewhere they didn’t before.
Your client signed a coaching agreement, not a data-sharing agreement
When a client signs with you, they agree to coaching. They hand over their contact and payment details, and often much more in the messages and notes attached to their project. They didn’t agree to have that information sent to an AI company every time you ask a question about your week.
That routing is real, and HoneyBook is upfront about it. Its help article on how your data is used with Claude says, “Your HoneyBook data is sent to Anthropic, the company that makes Claude, to generate responses to your requests.” The same article says Claude only pulls data relevant to what you ask, and it doesn’t read your account in the background.
The ICF puts a coach’s duty here in writing. Standard 2.5 of the ICF Code of Ethics asks coaches to meet their ethical and legal obligations to clients “directly and through any technology systems I may utilize,” and it names artificial intelligence in that list. Standard 2.4 asks you to store records “in a manner that promotes confidentiality, security, and privacy.”
None of that forbids the connector. It means the call is yours, and you should make it on purpose. If your contract’s confidentiality clause was written before any of this existed, give it a fresh read. Our guide to the legal requirements for a coaching business covers what that clause should do.
Four decisions to make before you connect
Setup takes about a minute, according to HoneyBook. The thinking should take a little longer.
1. Decide which records it can reach
Both versions ask you to approve access by resource instead of handing over the whole account. Read that permissions screen slowly. If you only want help chasing late invoices, the assistant doesn’t need your client messages. Start narrow. You can widen access later, and data that’s already been sent can’t be pulled back.
2. Decide whether it can act or only draft
This is the biggest difference between the two launches. The ChatGPT app hands you drafts, and nothing reaches a client until you approve it. The Claude announcement describes the connector creating and sending invoices and contracts from your templates and issuing payment requests.
An assistant that answers a question badly wastes a minute. An assistant that sends a contract to the wrong client costs you trust you worked hard to earn. If you connect through Claude, approve only what you need for reading at first, and hold off on anything that sends until you’ve watched how it behaves.
3. Decide where your data goes and how long it stays
When you ask a question, the relevant slice of your client records travels to the AI company to produce the answer. What happens to it after that depends on the AI provider and your plan with them, not on HoneyBook. HoneyBook says it doesn’t store or have access to Claude’s responses.
HoneyBook has published that Claude-specific data article. At the time of writing, we couldn’t find an equivalent one for the ChatGPT app, so treat that path with extra care. Before you connect either one, open your AI account’s current data settings and check two things. Can your conversations be used to train future models? How long are they kept? Consumer and business plans can handle these differently, and the policies change, so read them yourself rather than trusting a summary (including this one).
4. Decide what you’ll tell your clients
If an AI tool touches client records, your clients deserve to know. A plain line in your agreement or privacy notice that names the AI tools you use for admin, and what they can see, costs you nothing and builds trust. We covered the wider rules in AI disclosure requirements for coaches, and the session-recording side in AI notetaker consent for coaches. This is the business-records half of the same conversation.
When to leave it switched off
Some coaches should skip this for now. That’s a sound business decision, not a failure to keep up.
- Your project notes hold session content. If you use HoneyBook notes to capture what clients tell you in sessions, that material sits right beside everything else the connector can reach.
- You coach under sponsor contracts with strict confidentiality terms. Executive and corporate engagements may include clauses about how client information is handled. Check yours before a new third party enters the picture.
- You can’t answer decision three yet. If you don’t know your AI plan’s training and retention settings, you aren’t ready to connect.
- The admin isn’t actually a problem. If your invoicing already runs smoothly, the connector adds risk without much relief.
None of this is anti-AI. Our position is that AI belongs in your prep, your reflection, and the business work around your practice, while the human part of coaching is what clients pay for. We made that case in AI coaching platforms are coming for the mid-market. A CRM connector sits squarely in the business half. That’s exactly why it deserves a business owner’s scrutiny.
Make the setup decision on purpose
HoneyBook’s co-founder and CEO, Oz Alon, said in the ChatGPT announcement that AI “changes the ceiling on what one person can do.” For a solo coach juggling sessions and invoices, that’s an appealing promise. It may even be true for you.
If you already run your practice on HoneyBook (and our HoneyBook review explains why plenty of coaches do), give yourself ten minutes before you click Authorize. Know which records it can see. Know whether it drafts or sends. Know where the data goes and what your clients will hear from you. Then connect it, or leave it off, and know exactly why you chose what you did.
