Proactive AI That Understands How You Operate
Proactive AI identifies useful work before a person has to formulate a prompt. It uses context, memory, and timing to recognize an upcoming need, an unresolved commitment, a recurring workflow, or a meaningful change—and prepares the next step.
Useful proactive AI is not unrestricted autonomy. It should understand enough about a person's world to prepare relevant work, then let that person review, edit, approve, or dismiss it.
The goal is not more notifications or another inbox. The goal is to reduce cognitive overhead by understanding how a person operates, what they care about, and which moments genuinely deserve attention.
Proactivity is downstream of the personal model
A calendar trigger can say a meeting starts soon. A personal model can say who is attending, what happened last time, which commitments are still open, what you wanted to accomplish, and which questions matter now.
That is the difference between an alert and preparation—and it is why proactivity is the third stage of Sentience's progression, not a standalone feature. Personal AI is intelligence shaped around one person. The digital twin is the continuously evolving model of that person's mind. Proactive AI is that model put to work: using its understanding to prepare what its person would want prepared.
Without the model, a proactive system can only guess from surface signals, and guessing at scale produces noise.
Three levels of proactive
Shallow automation. Fixed triggers and generic alerts: a reminder fires because a timestamp arrived. The system knows that something is happening but nothing about what it means.
Contextual assistance. Responses grounded in the current event or message: a suggested reply built from the thread on screen. Useful, but its understanding ends at the edge of the current item.
Personal proactivity. Work identified from long-term knowledge of a person's relationships, priorities, commitments, and workflows: recognizing that an email quietly changes a project's status, or that a promise made three weeks ago comes due before Thursday's meeting.
Only the third level requires a model of the person—and only the third level can prepare work that feels like it came from someone who knows you.
How Sentience prepares work
Sentience uses the personal model to prepare, not to act on its own.
Before a meeting, it assembles a brief from the history of the people and topics involved. A daily digest surfaces what mattered across connected sources. When new information creates a clear next step, Sentience can propose the action itself: a drafted reply, a Slack message, a calendar change. Recurring routines deliver scheduled work, like a weekly review of open commitments.
Everything surfaces for review. The person edits, approves, or dismisses, and nothing external happens without approval.
A concrete example
Before an investor meeting, Sentience can prepare a brief covering the prior conversations with that investor, the questions left unresolved last time, the follow-ups that were promised, what has changed since the last meeting, and what the user intends to get out of this one.
Afterward, it can propose a follow-up draft that reflects what was discussed and how the user writes to this person. Whether anything is sent remains the user's decision.
The value is not that the AI did something impressive. It is that the user walked in prepared and walked out with the follow-through already started.
Relevance is the hard problem
A proactive system that surfaces everything becomes noise—another inbox to manage, which is a net loss.
The standard is not how many suggestions a system generates. It is whether each suggestion is sufficiently grounded, timely, and useful to deserve attention. That bar is hard, and it is exactly where the personal model matters: knowing what a person cares about is the only reliable filter for what deserves to interrupt them.
Sentience holds proactive work to that standard: grounded in the model, presented as optional, and quiet when there is nothing worth saying.
Work with an AI that already has the context
Sentience prepares briefs, digests, and next steps from a model of your world—so useful work starts before you have to ask, and nothing happens without you.
Apply for early access to build a personal AI that works the way you do.
Frequently asked questions
What is proactive AI?
Proactive AI identifies useful work before a person asks for it. It uses context, memory, and timing to recognize upcoming needs, unresolved commitments, and meaningful changes, then prepares the next step for the person to review.
How is proactive AI different from automation?
Automation runs fixed rules on fixed triggers and treats every firing the same. Proactive AI decides what is worth preparing from an understanding of the person's relationships, priorities, and commitments, and its output is proposed work rather than an executed rule.
Does proactive AI take actions without permission?
Not in Sentience. It prepares and proposes: briefs, drafts, suggested calendar changes. External actions such as sending an email require the person's review and approval, and any suggestion can be edited or dismissed.
Why does proactive AI need long-term memory?
Because relevance comes from history. Knowing a meeting is about to start takes a calendar. Knowing which unresolved question makes that meeting matter takes memory of the relationship, the project, and the commitments made along the way.
How does Sentience decide what may be useful?
It grounds suggestions in the personal model: the people, projects, commitments, and patterns in the person's connected sources. Timing matters too—context is prepared when it can affect what happens next, such as before a meeting or when new information changes a situation.
How can proactive AI avoid becoming another inbox?
By holding a high bar for attention. A proactive system should rank importance, stay grounded in what the person actually cares about, present work as optional, and remain quiet when it has nothing valuable to add.