The AI Agency Trap

Builder.AI sold itself as SaaS, but ran like an agency.

Hey — It’s Nico.

Welcome to another Failory edition. This issue takes 5 minutes to read.

If you only have one, here are the 5 most important things:

Let’s get into it.

This Week In Startups

🔗 Resources

📰 News

💸 Fundraising

Reproductive health startup, ReproNovo, raises $65M for infertility drugs.

Fail(St)ory

AI, But Make It Manual

This week, Builder.ai, once a $1B AI startup backed by Microsoft,  announced it is going through insolvency proceedings.

It promised to make building software as easy as ordering a pizza. But after burning through $450M, missing revenue targets, and losing trust, the company is now out of cash.

What Was Builder.ai:

Builder.ai sold the dream: don’t hire developers, don’t learn to code — just tell the platform what you want and it’ll spin up an app for you, powered by AI.

When it launched in 2016, the company was called Engineer.ai. Back then, it claimed its AI could build 80% of an app in an hour. But this was before GPT, AI wasn’t anywhere near that level. 

The truth? The “AI” was just a front. Behind the scenes, human engineers were doing almost everything. 

​​Eventually, the lies were exposed. The company rebranded. The tech improved. But that original sin followed them.

Nevertheless, Builder.ai managed to raise $450M, got Microsoft on board, and built out a team of nearly 1,000 people.

In its final form, Builder.ai was a mix of AI tools and human project managers. You’d describe your idea, either by picking a template or chatting with their AI assistant, Natasha, and the platform would scope it out. Then a team (yes, real people) would actually build the app for you, combining automation, templates, and custom dev work.

It wasn’t self-serve. It wasn’t drag-and-drop. It was more like a tech agency with an AI-powered sales layer on top.

The Numbers:

  • 📅 Founded in 2016

  • 💸 Raised $450M+

  • 🔥 At its peak: ~1,000 employees

  • ✂️ Cut 35% of staff in early 2024

  • 📉 Valuation hit $1B

  • 🧾 Owes $85M to AWS and $30M to Microsoft

Reasons for Failure: 

  • Inflated numbers: Former employees said the company inflated sales figures by over 20%, more than once. When Bloomberg started asking questions, Builder.ai admitted it had lowered revenue projections and brought in auditors to review two years of accounts. The damage was already done, investor trust took a hit. You can’t run a $1B startup with spreadsheet fiction.

  • The burn was out of control: Even after laying off 270 employees, the company was still drowning in costs. In the CEO’s own words, he was trying to run the company with “zero dollars” in its UK and US bank accounts. Meanwhile, Builder.ai owed $85M to AWS and $30M to Microsoft. 

  • The AI wasn’t always AI: The 2019 scandal around Engineer.ai hurt their credibility. Telling investors and customers you’ve built a magical AI, when it’s actually people writing code in the background, tends to stick. Even if the tech later improved, that early lie created a trust gap they never fully closed.

  • Weak monetization: They raised like a SaaS unicorn but operated more like an agency. The model depended on high-touch services and custom builds, which don’t scale well without strong margins or repeat business. Turns out, selling custom software with a thin AI layer on top isn’t a compounding business.

Why It Matters: 

  • You can’t fake traction twice. Builder.ai already burned trust with Engineer.ai. When questions came up again, no one gave them the benefit of the doubt.

  • AI wrapping a service business doesn’t scale like SaaS. Selling custom apps with humans behind the curtain looks good at pitch time, but the margins will eat you alive.

  • If you need audits to explain your numbers, it’s already too late. Startups move fast. Trust doesn’t.

Trend

Google I/O 2025

Google I/O happened on Monday, and surprise: it was all about AI.

There was a lot — some impressive, some predictable. But a few things genuinely stood out.

I’ve pulled the ones that feel most relevant if you’re building products or thinking about what’s next.

Let’s get into it.

Why It Matters:

  • Google Search now works more like a chat. This means traditional SEO strategies might start breaking.

  • Gemini is being integrated across devices, screens, and inputs. Users will expect AI help everywhere, not just in a chatbot.

  • Google is laying the groundwork for a real AR platform, and this time, it looks like they might actually ship glasses people want to wear.

Key Announcements

Ask Anything in Search

Google Search now has an “AI mode” that feels more like a conversation. You ask a question and get a direct answer instead of a list of links. You can also follow up naturally, like in a chat. It still shows sources, but they’re secondary.

If you’ve used Perplexity, it’s in that direction. This won’t kill websites, but it will probably change how content gets surfaced. If you rely on SEO, it’s time to start testing what kind of content these models pick up and reference.

Google Beam

Beam is the new name for their 3D video call project. It’s not for consumers yet, as it needs special hardware, but it’s working.

You sit in a booth and talk to someone, and it feels like they’re really there, with eye contact and all. It’s clearly meant for offices or enterprise settings.

The takeaway isn’t “holograms are coming,” it’s that Google’s trying to make remote interaction feel more normal, less flat. Something to keep in mind if your product depends on video calls or remote collaboration.

Veo 3

This is the announcement everyone’s been talking about. Not because it’s another text-to-video model — but because the videos now come with native audio.

Veo 3 can generate voices, background sound, and even ambient noise that matches what’s on screen. The result is videos that feel real in a way that’s kind of unsettling.

People are sharing clips online and calling them creepy, and they’re not wrong. It’s impressive, but also a reminder that we’re entering a phase where AI content won’t just look real — it’ll sound real too.

Gemini Live

Gemini is now available on Android and iOS, and the new “Live” features let you use your camera or screen while chatting with it. You can point your phone at something and ask questions, or share your screen and get help.

This kind of interaction makes the assistant feel more useful in real life — not just for writing or summarizing, but for actual problem solving. Worth trying out.

Android XR

Google is making another push into smart glasses, but this time, they’re focusing on making them wearable and useful.

Partnering with brands like Warby Parker and Gentle Monster, the new Android XR glasses are designed to look like regular eyewear, not tech gadgets.

Powered by Gemini AI, these glasses can provide real-time navigation, translations, and contextual information directly in your field of view. They also feature an optional in-lens display, cameras, microphones, and speakers for hands-free interaction.

While still in the prototype stage, Google plans to release developer tools later this year, signaling a serious commitment to making smart glasses a part of everyday life. 

Help Me Improve Failory

How Was Today's Newsletter?

If this issue was a startup, how would you rate it?

Login or Subscribe to participate in polls.

That's all of this edition.

Cheers,

Nico