Proactive assistants: what runs, what triggers it, and what you control

An assistant is useful before you ask only if it does the right work at the right moment. “Proactive” alone does not tell you whether it reads new information, repeats a reminder or changes something on your behalf.

Five behaviors hidden inside one word

Scroll sideways to compare all columns.

BehaviorTrigger and inputOutput to assessWhat you approve / how it stops
Reactive answerYou ask; the assistant uses available materialAn answer nowYou decide what to do with the answer; nothing runs again unless you ask again
Fixed reminderA chosen time; text you suppliedA prompt to do somethingYou approve the text and date when you set it; delete or pause the reminder to stop it
Source-based briefA schedule or meeting time; accessible current materialA summary of what matters nowYou choose the schedule and the material; pause or delete the recurring job. A brief that only informs you has no outside action to approve
Event responseA supported event, such as a new message, plus conditions you setA response to that eventCheck whether follow-on actions wait for your approval, and how to pause the job without disconnecting the account
External actionAn instruction or recurring job, with the necessary permissionA sent message, changed record or other action outside the assistantThe approval setting matters most: approve each action, grant a standing permission, or block the action. Stop by pausing the job or removing the permission

These are a way to evaluate a product, not a ladder where the last row is always better. A client-facing professional may want a brief before each call and prefer to send every reply personally. Someone coordinating a shared inbox may need the action itself completed.

Compare documented behavior, then test the control

Published capabilities below were checked on 21 October 2026. The final column records what each maker documents about approvals and stopping; the screens differ, so test the control in your own account before you rely on it.

Scroll sideways to compare all columns.

ProductDocumented behaviorApprovals and stopping, as documented
ChatGPTScheduled and monitoring tasks; on eligible plans, event-triggered tasks that respond to supported Gmail, Slack or GitHub activityActions that require approval may pause the task. Pause, resume, edit or delete tasks in Scheduled; a monitoring task can stop when an end condition you define is met
ClaudeRecurring work using configured sources; remote and local cases differAn approval mode is set for each task you set up manually. Pause, resume, delete or run a task on demand from Scheduled, where upcoming and past runs are listed
LindyRoutines that run on a schedule (daily, weekly or monthly) or on an event, such as an email arriving, a Slack message, a meeting about to start or a meeting ending; its assistant page also shows meeting preparation and calendar changesGuardrails per connected tool (Always allow, Require approval, Don't offer) cover writes in shared Slack threads; the docs say web app chat, Slack DMs, and iMessage and SMS are not guarded. A routine toggles off without being removed, and a dry run shows what it would do without sending
SanerScheduled proactive updates on its home screen, such as daily summaries, overdue tasks and insight digests, alongside notes, email and connected calendar informationYou set the update schedule under Settings, My AIs, Proactive Settings. The help article does not describe an approval step, so confirm what the updates can change
EquerryRoutines you save, including Morning Brief, Meeting Prep, Watcher and ReminderRoutines bring briefs, alerts and reminders to you through the app and do not send, book or change anything elsewhere, so there is no outside action to approve. Pause a Routine with its Active toggle, or archive or delete it

For step-by-step configuration, use the ChatGPT or Claude task guide. For a broader shortlist by price and job, see the executive-assistant comparison.

A better demonstration: the client changed something

Use this fictional scenario to decide which kind of help you need. You have a Thursday call with a client. Monday's note says the client wants the full rollout. Wednesday's email asks to start with one region and leaves the budget discussion for the call.

Different products can all appear proactive while producing very different results:

Write down the outcome you actually want before the demonstration. If it is preparation, ask whether the brief cites Wednesday rather than recycling Monday. If it is action, inspect the proposed change and who approves it. A convincing description of the change is not evidence that the action happened.

Check what happens when the source is missing

A recurring assistant needs a useful failure behavior. In a test, deliberately withhold Wednesday's update. The assistant should not claim to know it. Then provide a new update that contradicts an earlier assumption and check the next result.

Use three questions: what was read, how recent was it, and what could not be established? Keep your own expected answer beside the result. This makes it possible to distinguish “the information never reached the assistant” from “the assistant had it but produced a weak brief.” Those problems need different fixes.

Also make the stopping rule explicit. A project that ended last month should not keep producing a morning report. Decide when to review recurring jobs, who owns the instruction and which control stops the work. Muting a device alert is not a substitute for checking the task's state.

Equerry's version: arrive at the call already briefed

Save a Meeting Prep Routine and allow calendar access. Equerry uses the meeting details and what you have shared to prepare a brief 15 minutes before each meeting. The resulting brief is available in the app. Morning Brief is a separate Routine you set up for your regular catch-up.

What you sendCC Equerry on the regional-rollout email (or forward it), and save a note describing Monday's agreement.
What you ask“What changed about the rollout, and what needs a decision on Thursday?”
What you get backAn answer from that material, with a source chip when it uses a saved note. The Meeting Prep Routine can also bring the relevant context into your pre-call brief.

Replies that keep Equerry's address on the thread keep arriving. A Watcher checks on the schedule you choose; it is not a promise of an instant reaction to every email. A Reminder sends the text you set on the selected date. Equerry does not book meetings, send messages or read an unshared account.

That makes the choice concrete: choose Equerry for personal preparation from selected material; evaluate a connected-action product when the work requires changing the client's record or completing the booking.

Frequently asked questions

What is a proactive AI assistant?

The phrase can mean a timed reminder, a recurring briefing, a response to a supported event or an action taken in another service. Ask which behavior is included and what starts it. The term is also used for research tools and for customer-support systems that answer a company's customers; this page covers assistants for your own work, not those.

Is a scheduled reminder the same as a briefing?

No. A reminder can repeat fixed text. A briefing must read relevant material and produce a useful current summary. Judge its sources and freshness as well as whether it arrives.

Does proactive mean the assistant can send messages?

No. Preparation and external action are separate capabilities. If you need a reply sent or an appointment booked, check that exact action and its approval controls.

What proactive help does Equerry offer?

You can save Routines such as Morning Brief or Meeting Prep. Meeting Prep uses allowed calendar details and the material you have shared to prepare you before a meeting. It does not send the client's reply or book the appointment.

Last updated 2026-10-21.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.