The Story

The company that decided "confused user" wasn't acceptable

There's a moment every SaaS founder knows. A user signs up, pokes around the product for ten minutes, hits a wall they can't climb, and never comes back. No support ticket, no feedback - just silence and a churned subscriber. Daryl Budiman and Anirudh Ramprasad watched this problem play out again and again while building at MultiOn, and they decided the answer wasn't better docs. It was a smarter agent.

Zeroframe is the legal entity behind Andoria, an AI-powered customer onboarding agent that doesn't explain how your software works - it shows you. Not through a static tutorial, not through a chatbot that regurgitates help articles, but by actually interacting with the interface in real time and walking users through exactly what to do, action by action.

The founders met at MultiOn, an AI startup backed by General Catalyst and Amazon. In under nine months there, they helped scale the company to a triple-digit million-dollar valuation by building some of the earliest autonomous AI web agents capable of navigating browsers independently. When they left to start Zeroframe, they brought that muscle memory with them.

"When your users get stuck, Andoria steps in and helps them use your software." - Daryl Budiman, CEO, Zeroframe

The pitch is almost unsettlingly simple: paste one script tag into your HTML. That's the entire integration. Andoria then crawls your web application, learns how it's structured, maps the flows, and uses that knowledge to generate personalized walkthroughs for each user who needs help - adapted to that person's background, role, and goals.

Most onboarding tools are essentially fancy slideshows. They walk everyone through the same five screens in the same order, ignoring the fact that a developer onboarding to your analytics platform has entirely different needs than a marketing manager doing the same. Andoria bypasses that with personalization baked into the AI from the first interaction.

The People

Two engineers who built AI that browses the web, now building AI that teaches it

Daryl Budiman studied Symbolic Systems at Stanford - an interdisciplinary program sitting at the crossroads of philosophy, cognitive science, linguistics, and computer science. It's the degree you choose when you want to think about how minds work, artificial or otherwise. He played competitive badminton, has visited 16 countries, and writes about his experiments on a minimalist personal site that doubles as a digital notebook. The kind of person who treats curiosity as a professional tool.

Anirudh Ramprasad studied Computer Science at the University of Michigan Ann-Arbor, then worked at Actively AI and Cue AI before MultiOn pulled them both into the world of agentic systems. As CTO, he's the engineering engine behind a product that has to work reliably enough that a live user interaction doesn't become an embarrassing failure.

Between the two of them: Stanford's human-computer interaction research culture, deep time spent building systems that understand web interfaces at a technical level, and the hard-won operational knowledge of what it takes to scale an AI startup quickly. That's a lot of runway for a two-person founding team.

The Products

Andoria and Teach Mode

🤖

Andoria

The core AI customer agent. Integrates into any web application with a single script tag. Learns the app's interface autonomously, then generates personalized step-by-step walkthroughs adapted to each user's background and goals. Can perform actions directly within the UI to guide users through complex flows.

🎓

Teach Mode

A workflow-recording feature that lets companies show Andoria their proprietary processes. Record your team's most advanced workflows once. Andoria learns them. Users can then be guided through those exact flows on demand - or have them automated entirely. Launched on Product Hunt in December 2024 with 118 upvotes.

The practical difference between Andoria and a traditional onboarding tool is that Andoria is dynamic. It doesn't require a product team to manually write and maintain guided tours every time the UI changes. The agent learns and adapts. For a fast-moving SaaS company shipping updates weekly, that's not a nice-to-have. It's the whole point.

Their target market is B2B SaaS companies - particularly those with complex, multi-step products where confusion is a legitimate retention problem. Think enterprise tools, analytics platforms, developer dashboards, ops software. Places where the gap between "I signed up" and "I understand this thing" is wide enough for users to fall through.

The Edge

Why one script tag changes everything

The most underappreciated thing about Andoria is what it removes from the integration equation. Traditional onboarding platforms require your engineering team to instrument the product, write tours, maintain them as the UI evolves, and update them when features change. Andoria removes most of that burden entirely.

You paste a script tag. Andoria figures out the rest.

For a startup with two engineers and a backlog that never shrinks, this is genuinely compelling. The activation friction is low enough that a founder can test the product during a demo without waiting for an engineering sprint. In the world of B2B SaaS tooling, that's rare enough to be a real competitive advantage.