There is a difference between a tool and an agent.
A tool sits on a shelf. You take it down when you need it. You use it, you put it back. The tool is faithful — it works the same way every time — but it does nothing while it sits on the shelf. Your hammer doesn't notice that the picture is crooked. Your calendar app doesn't notice that you booked two meetings at the same time. Your notes app doesn't notice that the doctor's office just emailed you a new appointment.
An agent is on the floor with you. It carries some piece of your intention forward while you're busy with something else. The defining test: does it come back to you, or do you have to come back to it?
By that test, most "AI assistants" are still tools. They are very good tools — better than search, better than typing — but they wait. You ask, they answer. Step away, and they stop existing.
MyNaavi is built the other way around. The point is not what happens when you talk to Naavi. The point is what happens while you don't.
"The defining test of an agent: does it come back to you, or do you have to come back to it?"
A day in the life — set once, watched always
What follows is one ordinary day. Nothing in it is dramatic. The interesting part is what Naavi does in the silences — the things you never explicitly ask for in the moment, because you asked once, weeks ago, and Naavi has been quietly following along since.
Naavi finds the appointment letter
A PDF lands in your inbox — a confirmation from your dentist for the cleaning you booked. You haven't seen it. You won't see it for another hour. Naavi sees it now. Reads the file, pulls out the date and the address, files the original in MyNaavi/Documents/medical/, and adds the appointment to your calendar with a 30-minute travel buffer based on where you'll be that morning.
You set this up six weeks ago when you said "watch my inbox for appointments and put them on the calendar." You haven't thought about it since. It just works.
Naavi notices the silence
Your daughter usually emails you every week or two. It's been twenty-three days. Naavi flags it gently in the morning brief — "Sara hasn't been in touch lately, want me to ping her?" You say yes while pouring coffee. By the time you sit down for the 10 a.m., a draft message is waiting for your approval. Short. Warm. In your voice. You tap send.
You asked Naavi to "let me know if I haven't heard from Sara in a while" once, last year, after the family group chat went quiet for a stretch and you felt it. You haven't thought about that instruction since. Naavi has been watching the whole time.
Naavi handles the renewal
Your car insurance renewal notice came in this morning with a 6% increase. Naavi reads it, compares the rate to last year's policy in your MyNaavi/Documents/insurance/ folder, and includes a one-line note in your afternoon brief: "Insurance is up 6% — last year's quote from the broker is in your Drive if you want to call them."
This wasn't an automation rule you set up. Naavi just knows that you care about how money moves in and out of the house, and that you like to be told when something jumps. The bias is the instruction. You don't have to script it.
Naavi adjusts the travel time
Your 3:30 p.m. appointment is at an office on the other side of town. There's an accident on the route Google Maps usually picks for you. Naavi sees this when you walk out of your last meeting and pings: "Leave in eight minutes for the Northcote meeting — traffic added twelve minutes to the usual drive."
You didn't have to ask "how's traffic?" You didn't have to check anything. The decision came to you, with the answer already inside it. Eight minutes. Go.
Naavi brings the list
Two weeks ago, on a Tuesday, you said "add eggs and a roast chicken to my Costco list." Yesterday in the car you said "add laundry detergent." This morning your husband said "don't forget the school t-shirts."
You haven't looked at the list once. You don't need to. As you pull into the parking lot, your phone gently buzzes — "At Costco. 4 items on your list." One tap, and they're in front of you. The list was being held for this moment the whole time.
Naavi sends the thank-you
Three days ago, a colleague did you a real favour — covered a meeting on short notice while your kid was sick. You said to Naavi, between the airport and home, "remind me to thank David next week." Naavi waited. Today is the right day. The draft is ready when you sit down — your tone, his name, a one-sentence reference to what he did. You read it, change two words, send it.
The favour gets repaid. The relationship stays strong. You didn't carry the obligation in your head all week — Naavi did, and gave it back to you at the moment it would land best.
What an agent actually does
Look at the six vignettes again. Notice what they share.
You set the intention once. Sometimes weeks or months ago. Sometimes in a single sentence on a drive home. Sometimes never explicitly — Naavi inferred it from the way you live.
The world changed. An email arrived. Traffic shifted. A silence stretched. A list filled up. A relationship deserved a check-in. None of these moments required a query from you.
Naavi decided whether to act, and how. Not every email becomes a calendar entry. Not every quiet stretch becomes a draft. The agent has to judge. Some things become full actions; some become quiet flags in the morning brief; some get filed silently for later retrieval. The judgment is what makes it an agent rather than a script.
You got the right thing at the right time. Not "here are 47 notifications from today." A single sentence, in the moment it would matter most, with the decision already made.
Why this works
Three reasons, and they compound.
The first is that the bottleneck in most lives is not capacity to do work. It is capacity to notice when work needs to be done. The vast majority of dropped balls are not dropped because you couldn't lift them. They are dropped because no one was watching the corners of your life while you were focused on the centre. An agent watches the corners.
The second is that intention is sticky in a way that tools have never been able to use. When you say "watch my inbox for appointments" on a Wednesday in March, you mean it forever. You shouldn't have to renew the instruction every month. An agent that respects this — that treats your intentions as standing orders — works in a way no app has ever worked for you.
The third is that decisions are easier when they are presented one at a time, with the context already attached. "Leave in eight minutes — traffic added twelve to the drive" is one decision: go or push the meeting. The traffic check, the route comparison, the math — Naavi did it. You just choose. That's what an agent should do for the people it serves.
The shift from tool to companion
An agent that watches and acts on your behalf isn't a better chatbot. It is a different category of thing entirely.
It changes how you talk to MyNaavi. You don't have to remember to ask. You can just tell Naavi — once, the way you'd tell a thoughtful colleague — and trust that the instruction will carry forward.
It changes what you stop carrying. The mental tax of "remember to remember" — the dread that some important thing is slipping while you focus on the urgent thing — that tax drops. Not because you've gotten more disciplined. Because someone else is now holding the corners.
It changes the relationship. A tool is something you use. A companion is something you rely on. The word matters. Naavi is built to earn the second word, every day, in the small ways that compound.
That's the whole bet. Not better answers. Fewer questions you had to ask in the first place.
Set it once. Let Naavi watch the corners.
MyNaavi is in private preview. Tell us one thing you'd want followed on your behalf, and we'll be in touch within a few days.