Every Hey Button AI teammate runs at one of three access levels — read-only, draft-and-approve, or full access — and picking the right one matters more than picking the right first task.
What does read-only actually mean?
At read-only, the teammate watches a channel (texts, an inbox, a calendar) and tells you what it sees and what it would do, but it never sends anything or takes an action on its own. This is the level to start at if you want to see the teammate's judgment before it ever touches a real customer conversation — you get a running summary, not a surprise. The tradeoff is obvious: it saves you almost no time yet, because you still have to act on everything yourself. It's a trial period, not a working setup.
What does draft-and-approve mean?
At draft-and-approve, the teammate writes the reply, confirms the appointment, or drafts the quote follow-up — but it waits for a human to glance at it and approve before anything goes out. This is where most businesses land for the first real stretch of using a teammate. You get most of the time savings (you're reviewing, not writing from scratch) with almost none of the risk (a bad draft gets caught before a customer ever sees it). The honest tradeoff: you still have to be available to approve things, which means it's not fully "off your plate" yet.
What does full access mean?
At full access, the teammate sends the text, books the appointment, or replies to the email without waiting for a human to click approve first — for whatever scope you've defined. This is where the real time savings show up: things get handled at 11 PM on a Saturday without you touching your phone. The honest tradeoff is just as real — a mistake goes out before anyone catches it. Full access only makes sense for tasks you've already watched the teammate handle well at draft-and-approve, not as a starting point.
So which one should I actually start at?
Start at draft-and-approve for anything that touches a customer directly, and reserve full access for narrow, low-stakes, repetitive tasks you've already watched the teammate get right — confirming an appointment time that's already on the calendar, for instance, versus negotiating a price. Most businesses that start at full access everywhere end up dialing it back; almost nobody who starts at draft-and-approve regrets the extra week of reviewing drafts before loosening the leash.
Can I set different levels for different tasks?
Yes — access level isn't one global switch. A lawn care business might run full access on appointment confirmations (low stakes, already scripted) while keeping new-customer pricing questions at draft-and-approve (higher stakes, more judgment required). You set this during the setup questionnaire and can change it later as you build trust with how the teammate actually performs.
Does a higher access level mean a "smarter" AI teammate?
No — access level is a permission setting, not a capability setting. The same underlying system, routing through models like Claude and GPT depending on the task, powers all three levels. What changes is how much it's allowed to do without asking first, not how well it understands the task. Don't confuse "more access" with "more capable" — they're independent decisions. For the broader basics of what an AI teammate is before you even get to access levels, see What Is an AI Teammate?
Frequently Asked Questions
- What's the safest access level to start with?
- Draft-and-approve. It gives most of the time savings of full access with almost none of the risk, since a human reviews everything before it reaches a customer.
- Is read-only worth using at all, or should I skip straight to draft-and-approve?
- Read-only is worth a short trial if you want to see the teammate's judgment before it ever drafts something a customer could see — but it saves little time on its own, so most businesses move past it quickly.
- Can I change the access level later without starting over?
- Yes. Access level is a setting you can revisit once you've seen how the teammate performs, not a one-time irreversible choice.
- Does full access mean the AI teammate can do anything it wants?
- No. Full access still only covers the specific tasks and channels you've defined — it's full access within a defined scope, not unrestricted control of your business.
- Do different tasks have to use the same access level?
- No. You can mix levels — for example, full access on routine appointment confirmations while keeping anything involving pricing or new customers at draft-and-approve.
Conclusion
Access level is the one setup decision worth thinking through carefully — start cautious, prove the judgment, then loosen it. Start the setup questionnaire to pick your levels task by task.