Home/Libra AI

Getting Libra AI ready for launch, and launching it

A proactive AI agent for work. I helped make it ready for the market, then ran the launch. The launch video crossed 1.8M views.

Role
Product & Growth Consultant
Launch
14 May 2026
Scope
Product, website, launch video, creators, events, tracking
The launch video on X

The problem

It worked, but it wasn’t ready for the market.

Libra was building a proactive AI agent for work: it connects to your tools, learns how you work, and starts on things before you ask. A small group of people the team knew were testing it.

New users got a capable chatbot that only got good once it had learned their context, and a sidebar full of features, each impressive, none of them telling you what Libra was for. Everyone we showed the website to said the same thing: “I don’t get what this does.”

The launch was already in motion: a waitlist for our first 1,000 users. The founder brought me in as a force multiplier, to put taste and polish into every part of it.

The product

So I gave the product a point.

  1. By the time you finished, Libra had pulled in the context and showed you how it stitched your work life together. That first aha got people to try it a little. A little goes a long way, because every chat after that makes Libra better.

  2. 02Fewer features, louderI cut back the ones that distracted, so people got to the “aha” fast.
  3. 03A first screen that starts the workSuggestions from your own tools: “You haven’t replied to this email. Want me to start?” Or a starter set, if you hadn’t connected any.
  4. An agent that starts without enough context does bad work. I wrote the PRDs for these changes.

The launch

Then I made sure people heard about it.

  1. It’s since been replaced.

  2. 02Wrote the launch videoScript and storyboard were mine. The agency did the animation, sound and edit, and I ran every review.
  3. 03Warmed up XWe picked AI and tech creators with real engagement, and they built up attention in the days before launch.
  4. We also sent cakes to companies we admired, in SF and Bangalore.

  5. 05Set up the trackingEvents, GA4, Search Console and PostHog, so the team could see what happened after launch.

What happened

We launched on 14 May, and the video travelled.

1.8M+views on the launch video

It went out from the founder’s account, timed for India and SF to be awake together, and ended by opening the waitlist for the first 1,000 users.

Why it spread: launch videos on X were close to their peak, new AI products are fun to engage with, and the founder and the creators we picked carried it far.

What I learnt

Views are attention, not adoption.

If I did it again, I’d spend a few weeks first sitting next to 50–100 early users, watching how they actually use the product, and only then go big.