EVE autonomy — how much EVE does on its own
Status: live
EVE autonomy is the single place you decide how far EVE leans in — an overall mode for the whole workspace, plus optional per-area fine-tuning — with a fixed safety ceiling that always routes the sensitive stuff to you no matter what you turn on.
This is the full write-up: read it before you switch anything on, so you know exactly what each setting hands to EVE and what it never will.
Two layers
Autonomy is set in two layers, and the second one is optional.
- The overall mode — one setting for the whole workspace: Suggest, Guided, or Autopilot. Set it once and it governs every kind of work EVE does.
- Per-area fine-tuning — an override for one kind of task at a time (follow-ups, conversations, scheduling, and so on). Each area starts on Inherit, which simply means "follow the overall mode." Change an area only when you want EVE more (or less) hands-on for that one thing than for everything else.
Most people set the overall mode and never touch the areas. The fine-tuning is there for the case where you're comfortable letting EVE send routine follow-ups on its own, but you still want to approve every scheduling move — or the reverse.
Agents pick their own mode. In a Team or Brokerage workspace the broker sets a ceiling — the most autonomous EVE is allowed to be for anyone in the workspace — and each agent chooses their own level up to it. If the broker's ceiling is Guided, Autopilot is locked for the whole workspace and you'll see a note saying so. A solo agent on a Personal workspace has no ceiling above them.
The three modes
The mode names describe how much EVE does before it involves you.
| Mode | What EVE does |
|---|---|
| Suggest | EVE surfaces what needs doing and recommends the next step. You do the work. The most hands-on mode. |
| Guided | EVE prepares the work — drafts the message, builds the task, fills the form — and puts it in front of you for one-tap approval. Nothing goes out until you say yes, but the heavy lifting is done. |
| Autopilot | EVE handles routine, clearly-owed work itself and reports back afterward. It only acts when it can cite exactly why, skips anyone recently contacted so nobody is double-messaged, and records what it did. |
Guided is the right place to start. You see precisely what EVE would do before anything happens, and moving up to Autopilot later is one click.
Fine-tune by area
Below the overall mode is a matrix of areas. Each row is one kind of work, with its own mode selector defaulting to Inherit. Setting a row to Suggest, Guided, or Autopilot overrides the overall mode for that area only; leaving it on Inherit keeps it in lock-step with the overall mode.
Two areas — Marketing and Data entry — deliberately stop at Guided. They never offer Autopilot, for reasons explained in each row below.
Follow-ups
EVE keeps relationships from going cold. It watches for someone who's been waiting on a reply, a relationship that's drifted quiet and is worth reconnecting, a client whose Realm is nearing its close date and deserves a reassuring update, and a promise you made that's coming due.
- Suggest surfaces each as something to handle.
- Guided drafts the follow-up and adds an approval task.
- Autopilot sends the clearly-owed ones itself and records the win.
Promise follow-through is always prepare-only, even on Autopilot: EVE can draft the follow-up and remind you, but a promise is yours to keep, so it never sends on its own.
Range: Suggest / Guided / Autopilot.
Conversations
When a lead replies, EVE can carry the conversation — answer follow-up questions turn after turn, qualify (timeline, financing, area, motivation) naturally, and propose showing times from your calendar. It always introduces itself as your assistant and never pretends to be you.
- Suggest surfaces the reply for you to handle.
- Guided drafts each reply as an approval task.
- Autopilot replies on its own until it hits something that needs you, then hands off.
This is the same engine as EVE Talk; the Conversations area is where its autonomy is set.
Range: Suggest / Guided / Autopilot.
Scheduling
EVE detects when a meeting needs to happen and handles the calendar invite end to end — the message, the appointment, the invite email.
- Suggest flags that a time should be set.
- Guided proposes times and waits for your approval.
- Autopilot offers your real open slots and books.
One rule holds in every mode: a day you've marked confirm-required in your booking guidelines always waits for you. Even on Autopilot, EVE frames those as "let me confirm with you and lock it in" rather than booking them — a consent-required day never auto-confirms.
Range: Suggest / Guided / Autopilot.
Document requests
A transaction is missing a document the client owes — a disclosure, a signed form, a pre-approval. EVE nudges for it.
- Suggest flags the gap.
- Guided drafts the nudge for approval.
- Autopilot sends the reminder and tracks whether it lands.
Range: Suggest / Guided / Autopilot.
Marketing (draft only)
EVE drafts marketing content — social posts, listing copy, a review turned into a post. It never publishes on its own.
- Suggest proposes ideas.
- Guided writes the draft for you to review and post.
There is no Autopilot here on purpose. Publishing to your brand and your audience is a judgment call about voice and timing that stays with you; EVE writes the first draft, you decide what goes out.
Range: Suggest / Guided. No Autopilot.
Data entry (draft only)
EVE reads a conversation and proposes CRM updates — a new phone number mentioned in a text, a corrected email, a stage that clearly moved.
- Suggest points out what looks out of date.
- Guided proposes the exact change for you to approve.
No Autopilot here either. A silent, automatic write to your system of record can quietly overwrite something you'd never have changed, so the human always approves the write.
Range: Suggest / Guided. No Autopilot.
The safety ceiling — this always comes to you
This is the rule that makes autonomy safe to turn on, and it does not bend in any mode. Certain topics are never handled automatically — not in Guided, not in Autopilot, not ever. EVE reads every inbound message before it acts, and the moment it sees one of these, it drops out of automatic handling for that item and routes it to your review queue with a short note on why.
The topics that always come to you:
- Price and negotiation — "can they come down on the asking?", "what's the lowest they'd take?", any back-and-forth on price.
- Offers — "we'd like to put in an offer", "here's our offer".
- Contract terms — a purchase agreement, an addendum, an amendment, a counter-offer, "are we signing today?".
- Legal — a lawyer or attorney, liability, a disclosure issue, anything with legal weight.
- Financing — mortgage, lender, pre-approval, down payment, rate or loan terms.
- Complaints — an unhappy, frustrated, or angry message.
- Opt-outs — "unsubscribe", "take me off your list", "stop texting me".
When one of these trips, EVE never sends. On Guided it was already going to ask; on Autopilot it stops itself and escalates instead of acting. This is the hard ceiling — Autopilot speeds up the routine, and this is the line it will not cross. The escalated item lands in your human review queue.
Where to set it
Everything lives in one place: Settings → EVE. The overall mode is the card at the top; the per-area matrix is directly below it. Change the overall mode to move everything at once, or set an individual area's row to override just that kind of work.
Brokers set the workspace ceiling from the same surface — the ceiling caps both the overall mode and every per-area override for everyone in the workspace.
What's retired
The old Drafting replies area is gone. What it did is now covered by two clearer areas: replying inside an ongoing thread is part of Follow-ups, and answering a lead's inbound reply is part of Conversations. If you had set Drafting replies before, set the mode you want on those two areas instead.
Related
- EVE's autonomy modes — the short version of the modes.
- EVE Talk — the conversation engine behind the Conversations area.
- Compose & review — the human review queue where escalated items land.