Skip to content
HOME OF LUX AI Join Early Access
PROACTIVE AI ASSISTANT

What if your AI remembered
what happens next?

Most AI waits for another question.

A proactive AI assistant is designed to recognise useful next steps, retain relevant context and help at the moment that help becomes useful.

See how Lux worksJoin Early Access
ILLUSTRATIVE DESIGN / BEING DEVELOPED
YOU / THE INTENTION
“I've got the dentist Thursday at 10:30.”
TIME PASSESRELEVANT CONTEXT CHANGES
LUX / THE USEFUL MOMENT
“Traffic is getting heavier.
Leaving at 09:42 should give you enough time.”

The user didn't have to ask again.

An intended experience, not a live traffic integration or an available autonomous feature.

AFTER THE ANSWER

What is a proactive AI assistant?

A proactive AI assistant is designed to remember a relevant intention, consider the surrounding context and offer a useful next step before you ask again. An AI assistant that reaches out first should do so only when a reminder, preparation or follow-up would help. Lux’s illustrative development examples include preparing two interview questions, revisiting an unfinished commitment and considering when to leave for an appointment.

Reactive assistance responds to the request in front of it. Proactive artificial intelligence considers whether a relevant intention needs attention later. These are modes of assistance; a system can support both.

Reactive AI

  1. You ask
  2. AI answers
  3. Conversation ends

A conceptual request-and-response pattern.

Proactive AI

  1. Understand
  2. Remember
  3. Watch for relevant change
  4. Recognise the useful moment
  5. Help
  6. Follow through

A design model, not a claim about every assistant or competitor.

Sending more notifications doesn't make an assistant more proactive. The useful distinction is whether it retains the intention behind a task, considers permitted signals and decides whether another helpful step is warranted.

An AI that anticipates needs should offer informed help, rather than assume it knows everything you will want. Sometimes the right next step is a question. Sometimes it is no interruption at all.

CONTEXT → OUTCOME → CONTEXT

The Lux loop.

Design direction · Being developed

Deluxevo is building Lux around a continuing loop: carry forward what matters, consider relevant change and return to the task until its next useful outcome. The intended sequence is to remember a commitment, understand its context, anticipate a useful moment, reach out, help resolve it and follow through. This is the product model; every stage is not yet a production-ready capability.

  1. 01

    Remember

    What matters? Retain the intention and the context worth returning to.

  2. 02

    Notice

    Has something relevant changed? Consider supported information the user has authorised.

  3. 03

    Understand

    Does it affect the user's intention? A change matters only in relation to the task.

  4. 04

    Act

    Is there something useful Lux can do or suggest within the permitted scope?

  5. 05

    Follow through

    Did the task actually reach an outcome? A message sent is not the same as a task finished.

  6. 06

    Continue

    Does anything still need attention? Carry useful context into the next step of the loop.

The loop returns to context, not to constant activity. When there is no useful next step, it should rest.

HOW THE MODEL WORKS

From an intention to an outcome.

Three everyday examples show what proactive follow-through could mean. Each is an illustrative development scenario, not a demonstration of publicly available automation.

01 / THE APPOINTMENT

A reminder tells you when.
A proactive assistant understands what surrounds it.

A fixed reminder knows the appointment time. The intended proactive experience connects that commitment with the journey, then suggests a useful departure time when relevant information changes.

The timing in this example is illustrative. It would depend on supported journey information, appropriate permission and enough context to make a useful suggestion.

Proactive journey assistance · In development
  1. 10:30
    THE COMMITMENTDentist appointment
  2. 09:15
    RELEVANT CHANGETraffic changes
  3. LUX
    UNDERSTANDEvaluate the journey context
  4. 09:42
    USEFUL NEXT STEP
    “Traffic is building. Leaving at 09:42 should get you there comfortably.”
  5. YOU
    ACTIONUser leaves
  6. ILLUSTRATIVE OUTCOMERESOLVED
02 / THE UNFINISHED THREAD

Life interrupts you.
Your assistant shouldn't forget the thread.

Conversation history records what was said. Useful continuity also retains what you were trying to achieve, where you stopped and what remains unfinished.

An AI that remembers tasks could bring back the next useful point without asking you to reconstruct yesterday's conversation. Whether and when it follows up should depend on relevance and your controls.

Unfinished-task surfacing · In development
MONDAYAn important start.

You begin the outline.
Life interrupts.

THREAD LEFT OPEN
TUESDAYThe useful point remains.
“You wanted to finish this yesterday.
We stopped at the outline.”
CONTINUE / PREVIEW OUTCOME
03 / THE REASON BEHIND THE DATE

Remember the reason, not just the date.

A calendar notification can tell you that an interview is tomorrow. Context makes the follow-up more useful: you had a specific intention to prepare two questions tonight.

Proactive follow-through reconnects the event with the preparation you wanted to finish. The goal is to help you become ready, rather than merely repeat the date.

Contextual preparation follow-up · In development
JOB INTERVIEW / TOMORROW09:00
EARLIER / YOU
“I want to prepare two questions tonight.”
LATER / LUX
“Your interview is tomorrow. You wanted to prepare two questions. Want to finish those now?”
Two questions preparedREADYILLUSTRATIVE OUTCOME
A PLAN ISN'T A FIXED PICTURE

Useful help can depend on what changes.

A plan may have been correct when it was made. Traffic changes. Weather changes. Appointments move. A response arrives. A task remains unfinished.

A context-aware AI assistant should ask whether an authorised, relevant change alters the next useful step. New information is not automatically a reason to interrupt.

For Lux, these signals are design direction. This page does not claim live weather monitoring, inbox monitoring, traffic integrations or every calendar capability. Each would require supported sources, implementation and appropriate permission.

WHEN THE PLAN WAS MADEThe intention is clear.
  • Journey context
  • A moved appointment
  • A new response
  • An unfinished task
WHEN SOMETHING CHANGESIs the next step still useful?
CONCEPTUAL SIGNALS / DESIGN DIRECTION
RELEVANCE BEFORE INTERRUPTION

Helpful when it matters.
Quiet when it doesn't.

A proactive personal AI assistant shouldn't turn every event into a message. The design goal is relevant help, not constant notifications, guessing every need, interrupting every few minutes or acting without appropriate permission.

  1. Not every event
  2. Relevant change
  3. Does this matter?
  4. Is help useful now?
YESSurface it

A useful next step, with context.

NOStay quiet

The thread remains. No interruption.

An AI that follows up earns its place through timing and relevance. Frequency alone is not the measure of helpfulness.

PERMISSION IS PART OF THE LOOP

Proactive doesn't mean uncontrolled.

Contextual assistance may involve a calendar, location, journeys, tasks, connected devices or personal context, where supported and explicitly authorised.

Retaining an intention does not automatically grant permission to access another source or take an action. A suggestion, a reminder and a device action can require different scopes of permission.

The design goal for Lux is appropriate user control over what it can access and how proactive assistance operates. When the context is uncertain or an action is outside that scope, the assistant should ask rather than assume.

These are product design principles, not absolute privacy or security guarantees. The Deluxevo privacy notice explains how this website handles Early Access information.

MEMORY + PROACTIVITY

Without continuity, proactive AI has nothing useful to follow.

MEMORY ALONE

“I remember something.”

PROACTIVITY ALONE

“I detected something.”

“I remember what matters, I noticed something relevant changed, and I understand why it may matter now.”

Persistent memory gives an assistant relevant context to return to between conversations; it should retain useful intentions, not every detail. Memory supplies the reason a change matters. Proactivity supplies the opportunity to help. Combined appropriately, they can turn an isolated interaction into useful continuity.

Explore how Lux is being built around memory and continuity.

THE TASK CAN TRAVEL

A task doesn't always stay on one device.

Lux's cross-device direction is about carrying a useful thread from phone to PC. It isn't permission for unrestricted remote access: computer capabilities depend on a connected, authorised device and supported actions.

Android is in development. Lux PC is a working Windows alpha. Unfinished cross-device tasks, including planned file retrieval, remain development work.

Explore Lux
PHONELuxAndroid — in development
LUX CONTEXT
PCLux PCWorking alpha
A FEW USEFUL ANSWERS

Questions about proactive AI.

What is a proactive AI assistant?

A proactive AI assistant is designed to retain relevant intentions and context, consider authorised information and offer a useful next step when it matters, without requiring the user to repeat every request. It should distinguish a meaningful opportunity to help from an event that needs no action.

How is proactive AI different from a chatbot?

A typical chatbot interaction begins with a prompt and ends with a response. Proactive assistance adds consideration of what happens later: whether the intention is still open, something relevant changed or a follow-up would help. A product may combine reactive conversation and proactive behaviour; the comparison here describes those modes, not every chatbot's capabilities.

Does proactive AI need memory?

Some proactive behaviour can use a single explicit task or reminder. More personal, useful follow-through needs continuity about the intention, relevant preferences or the point where a task stopped. An AI assistant with memory should retain what helps, with appropriate controls, rather than assume all past information is useful.

Can a proactive AI assistant act without permission?

Proactivity is not a blanket permission to act. Access to information and actions should stay within the scope the user has authorised. The design goal for Lux is to distinguish useful suggestions from actions requiring permission and to ask where the scope is unclear.

Is Lux a proactive AI assistant?

Lux is being built as a personal AI assistant centred on context, memory and proactive follow-through. The Lux Loop on this page is its design direction. Proactive reminders, journey assistance and unfinished-task surfacing remain in development; the examples are illustrative rather than claims of public availability.

Does Lux work on Android and PC?

Lux for Android is in development, and Lux PC is a working alpha on Windows. The goal is one continuing assistant across supported, connected and authorised devices. Not every cross-device capability is finished. Meet Lux for the product's current direction and development labels.