Explainer

Proactive AI vs Reactive AI

Nearly every AI product is reactive: it is excellent once you address it and does nothing until you do. Proactive tools originate — they examine data on their own schedule and raise something you did not request. The distinction decides whether a tool helps with the thing you forgot.

Read summarized version with ChatGPT

Perplexity Gemini

The distinction, and where real tools fall

1Reactive AI produces an output in response to an input you supply. ChatGPT, Claude, Gemini in Gmail, Copilot in Outlook and every "summarise this thread" button are reactive. The quality can be extremely high; the trigger is always you.
2Proactive AI runs without being addressed. It examines data on a schedule, applies a rule for what is worth raising, and raises it. The hard part is not generating text — it is the judgement about what deserves an interruption, because a proactive tool that raises everything is a notification you turn off in week two.
3The test: if you do not open the tool for a week, does it tell you anything? Reactive tools go silent. Proactive ones do not.
4Automation is not the same as proactivity. A Gmail filter runs without you, but only along a path you specified in advance. Proactive means the tool decided something was worth raising, not that it executed a rule you wrote.
5Where real email tools fall: Superhuman's Auto Labels and Gmail's Priority Inbox sort and rank without being asked, which is proactive on incoming mail. Gemini and Copilot are reactive. Google's AI Inbox surfaces to-dos on its own but only from incoming mail, and only on an eligible paid AI plan. Boomerang's follow-up reminders fire on their own but only for threads you flagged at send time — the trigger is still an instruction you gave.
6The gap almost nobody fills: mail you sent that got no reply, where you never set a reminder because you were confident of an answer.

Where it falls short

•You cannot prompt a reactive tool about something you have forgotten — remembering it is the whole problem
•Snoozes and send-time reminders require correct prediction; the threads that go quiet are usually the ones you did not worry about
•Proactive features that raise too much train you to ignore them, which is worse than not having them
•Most inbox intelligence analyses incoming mail, while the commercial risk sits in outgoing mail with no reply
•"Proactive" is used in marketing for anything scheduled, including plain filters

What Clarity does without being asked

1Incoming mail is read and labelled as it arrives — Gmail labels or Outlook categories — so the To Respond queue is built before you open the inbox.
2Sent mail is checked for silence. Threads where you wrote and nobody replied are surfaced, with no flag or reminder set at send time.
3Replies are drafted in your voice and left in the drafts folder, including the chase on a stalled thread.
4A morning brief delivers the two lists — needs a reply, gone quiet — in one message rather than a stream of notifications.
5The proactivity stops before sending. Clarity prepares; you send.

✓Covers the sent folder, which the reactive tools and most of the proactive ones never examine
✓Follow-ups are found retrospectively, so you do not have to have predicted which thread would stall
✓One morning brief instead of per-event alerts, which is what keeps a proactive tool from becoming noise
✓No Google AI plan or Microsoft 365 Copilot licence required — it runs on a standard Gmail or Outlook account
✓Proactive about noticing, never about sending: nothing leaves your account without you

Who this is for

Business development lead

Sends fifteen proposals a month and chases the ones she happens to remember on a Friday.

Engineer turned founder

Uses ChatGPT daily and it has never once told him about an email — because he has to open it first.

Studio director

A client thread stopped after the scoping call and nobody noticed for a month.

Frequently asked questions

Is a Gmail filter proactive AI?

No. A filter is automation: it runs without you, but only along the path you defined in advance, and it makes no judgement. Proactive means the tool decided that something was worth raising — a decision you did not encode.

Isn't ChatGPT proactive if I schedule a task?

A scheduled prompt is still an instruction you wrote, so what comes back is bounded by what you thought to ask for. The distinction that matters is whether the tool can raise something outside the scope of your instruction — an email you never mentioned, from a thread you had forgotten.

What does Clarity actually do on its own?

Three things: it labels incoming mail, it identifies sent mail with no reply, and it drafts replies. It does those whether or not you have opened it. It does not send, and it does not act outside your mailbox.

How does it avoid becoming another source of noise?

Output is concentrated into a morning brief and into labels you see in the inbox you were already opening, rather than per-message alerts. Labels sit in Gmail and Outlook natively, so reading them is not a separate destination.

Does proactive mean it will email people for me?

Not in Clarity. The proactive part is noticing and preparing — the draft is written and waiting. Pressing send is always a human action, which is the line where a wrong judgement stays cheap.

Reference: Google's Priority Inbox documentation

Roughly 9.1 hours a week

Estimate

Writing replies

6.4

Reading & triage

1.1

Follow-up tracking

0.9

Chasing & state

0.7

2.8 hrs of that we could cover

The rest stays yours — it is not repeatable.

106 hrs

a year, if the low end of the range holds

137

threads went quiet and nothing told you

Last 90 days · your own mail · illustrative figures

Clarity Inbox

Get your free inbox report

An interactive report on your last 90 days: where your hours actually went, what is still waiting on you, and how much of your email could run itself. Free, delivered in minutes.

Get my free report

Takes 2 minutes · No credit card · Nothing is sent without you