The Best AI Assistant Doesn’t Ask What It Can Do — It Tells You, Then Does It

I was sitting with the Lindy product team in San Francisco last December when they asked me a question I’ve been thinking about ever since.

“If you could describe the perfect AI assistant in one word or phrase — the version that would feel truly magical — what would it be?”

I thought about it for a second. And I said: proactive.

Not smart. Not fast. Not accurate. Proactive.

There’s a version of AI that waits for you. And there’s a version that doesn’t.

The Reactive Problem

Most AI tools work the same way: you go to them, you ask, they respond. You forget to ask? Nothing happens. You’re still the one who has to remember. You’re still managing the process. The AI is just the execution layer — and only when you think to activate it.

This isn’t nothing. Reactive AI still saves a lot of time. But there’s a ceiling on how much it can change your life, because you’re still running the mental overhead. The AI gets smarter, but you’re still the one doing the thinking about when and what to request.

That ceiling is real. And most people hit it faster than they expect.

What Proactive Looks Like

The best human assistants don’t work reactively. The A-players — the ones people pay $150K+ a year for — anticipate needs before being asked.

They scan your calendar at the end of the day, notice you have a big pitch meeting Friday morning, and send you a research brief Thursday afternoon without being told. They notice you’re traveling next week and pre-book a restaurant for your dinner meeting without you mentioning it. They spot the follow-up you forgot and send it themselves.

That’s what I mean by proactive. Not reacting to requests. Anticipating and acting.

I was describing this to the Lindy team with a real example from a VC I work with. Before we set up his AI prep system, he was spending roughly 20 hours a week just on meeting preparation. Researching every person he was meeting, digging through old email threads for context, pulling up the latest news on their companies. Every week. 20 hours of prep.

Now he gets a brief every morning. Every call for the day, already researched and summarized. He didn’t ask for it that morning. It just shows up. The AI is running on a schedule, doing the research ahead of time, delivering the package before he thinks to ask.

That’s the difference.

The Linda Example

I’ve seen this work firsthand with my own scheduling setup.

I built a virtual assistant called Linda. When someone wants to meet with me, they CC Linda on an email. She checks my calendar, finds available times, responds within a few minutes with options, confirms when they pick one, and puts it on my calendar. (There’s actually a three-minute delay built in — because when she responded in 60 seconds people got suspicious and thought it was a bot.)

Nobody asked Linda if she wanted to help schedule a meeting. Nobody prompted her with “there’s a meeting to schedule.” She figured it out from the email context and handled the entire workflow without being told.

That’s proactive. She detected a need and acted on it.

The bar for a truly great AI assistant is exactly this: it notices the thing that needs doing and does it, rather than waiting for you to notice and request.

Why This Design Philosophy Matters

When I said this to the Lindy team — “don’t ask me what you can do for me, tell me, then do it” — I was trying to capture something broader than product design.

It’s a different philosophy of what assistance means.

Reactive assistance means you have a powerful tool that you have to learn to use. You have to build the habit of reaching for it. You have to remember it exists and think about what to ask.

Proactive assistance means you have a system that’s paying attention on your behalf. You don’t have to remember to use it because it’s not waiting to be used. It’s running. It notices what you need and handles it.

The practical difference is that proactive AI multiplies you even when you’re not thinking about AI. The reactive version only helps when you’re actively engaged with it. One changes your whole day; the other just makes certain tasks faster.

The Meeting Prep Test

If you want a quick test for whether an AI assistant is genuinely proactive or just a dressed-up chat interface, look at meeting prep.

Does the tool wait for you to ask “can you prep me for my 2pm call?” Or does it just send you a brief at 1:30pm automatically?

That gap — between “ask and receive” and “receives before asking” — is the whole ballgame.

The meeting prep agent is consistently one of the highest-impact things I set up for clients. And the pattern is always the same: they don’t know they want it until they see it the first time. And once they have it, they can’t imagine going back to researching manually.

That’s proactive AI working. Not impressive on a feature list. Transformative in practice.

P.S. The Two Hour Workday program is built around this principle — AI working while you’re not watching. If you want to understand how to set up that kind of system for yourself, that’s where to start.