Builder.ai: a $1.5B 'AI app builder' that collapsed into insolvency
Builder.ai raised hundreds of millions to build apps with an AI assistant called Natasha. Former staff say humans did most of the work; restated revenue and a cash seizure ended it in May 2025.
- Product
- Builder.ai
- Founder
- Sachin Dev Duggal
- Launched
- n/a
- Model
- Custom software projects sold as an AI-assisted service
Published · 2 min read · 3 sources
- #ai
- #no-code
- #cautionary
- #ai-washing
- #venture-funded
Key takeaways
- Raised more than $450 million and was valued at $1.5 billion, promising apps built mostly by AI.
- Former employees say staff and outsourced developers did the vast majority of the work; projects ran late.
- After revenue restatements and a lender seizing most of its cash, it entered insolvency on 20 May 2025.
What it is
Builder.ai sold app and website development through a chatbot called Natasha, which the company said would build "the first 80%" of a product before human engineers finished it, using a library of pre-coded reusable features (Rest of World).
Timeline
- 2016: founded by Sachin Dev Duggal, based in London (Rest of World).
- 2019: reports that it was exaggerating its use of AI begin (Rest of World). Under its earlier name Engineer.ai, The Wall Street Journal reported the platform "relied heavily on human engineers" (TechCrunch).
- Funding peak: investors including Microsoft and the Qatar Investment Authority put in as much as $445 million, at a $1.5 billion valuation (Rest of World); TechCrunch puts the total at more than $450 million (TechCrunch).
- 2024: the company lowered its estimated revenue for the second half of 2024 by 25% (TechCrunch).
- Mar–Apr 2025: Duggal steps down as CEO; about 270 of roughly 770 staff are cut; 2023 revenue is lowered to $140 million and the 2023 accounts restated (Tech.eu). Lender Viola Credit seizes most of its cash (Rest of World).
- 20 May 2025: Builder.ai confirms it is entering insolvency proceedings (TechCrunch).
Why it worked (for a while)
- A pitch everyone wanted to believe. The goal, per a former employee, was to make building software "as easy as ordering a pizza" (Rest of World).
- Real demand for cheap custom software. Customers did sign up; the trouble came in delivery, not in getting interest (Rest of World).
What to learn
- Your marketing must match what the product actually does. Analysts cite Builder.ai as a possible case of "AI washing" (Rest of World). If humans do the work, say so and price it that way.
- A service business dressed as software keeps service margins and timelines. With humans doing most of the development, delivery took longer than clients expected and some cancelled or refused to pay (Rest of World).
- Revenue you report has to survive an audit. The company hired auditors to review two years of accounts before it failed (Tech.eu).
- Today an indie builder can ship with real AI coding tools; the lesson is not "AI app builders fail" but "don't promise automation you don't have."
Risks and what didn't work
- Debt can end you faster than losses. A lender seizing most of the cash came right before insolvency (Rest of World).
- Inflated numbers. Former employees told Bloomberg the company inflated sales figures by more than 20% on multiple occasions (TechCrunch); Bloomberg also alleged round-tripping with an Indian firm, which that firm called "absolutely baseless and false" (Rest of World). These are allegations, not court findings.
- The other side. Former CEO Manpreet Ratia wrote that Builder.ai "built a real AI-powered platform" and "It wasn't a gimmick" (Rest of World).
FAQ
Short answers drawn from the sections above.
What is Builder.ai?
Builder.ai sold app and website development through a chatbot called Natasha, which the company said would build "the first 80%" of a product before human engineers finished it, using a library of pre-coded reusable features (Rest of World).
Why did Builder.ai work?
- A pitch everyone wanted to believe. The goal, per a former employee, was to make building software "as easy as ordering a pizza" (Rest of World).
- Real demand for cheap custom software. Customers did sign up; the trouble came in delivery, not in getting interest (Rest of World).
What can indie builders learn from Builder.ai?
- Your marketing must match what the product actually does. Analysts cite Builder.ai as a possible case of "AI washing" (Rest of World). If humans do the work, say so and price it that way.
- A service business dressed as software keeps service margins and timelines. With humans doing most of the development, delivery took longer than clients expected and some cancelled or refused to pay (Rest of World).
- Revenue you report has to survive an audit. The company hired auditors to review two years of accounts before it failed (Tech.eu).
- Today an indie builder can ship with real AI coding tools; the lesson is not "AI app builders fail" but "don't promise automation you don't have."
What didn't work for Builder.ai, and what are the risks?
- Debt can end you faster than losses. A lender seizing most of the cash came right before insolvency (Rest of World).
- Inflated numbers. Former employees told Bloomberg the company inflated sales figures by more than 20% on multiple occasions (TechCrunch); Bloomberg also alleged round-tripping with an Indian firm, which that firm called "absolutely baseless and false" (Rest of World). These are allegations, not court findings.
- The other side. Former CEO Manpreet Ratia wrote that Builder.ai "built a real AI-powered platform" and "It wasn't a gimmick" (Rest of World).
Sources
Every link cited above, in order of first use.
- 1.Rest of World (restofworld.org)
- 2.TechCrunch (techcrunch.com)
- 3.Tech.eu (tech.eu)
Related case studies
More in AI tools →Formula Bot: a weekend no-code app, one Reddit post, and a $4–5K API bill
David Bressler built Excel Formula Bot in Bubble over a weekend. A Reddit post brought 100,000 visitors overnight and a surprise OpenAI bill before pricing was in place.
Humane Ai Pin: a $699 AI wearable that was switched off within a year of shipping
Humane launched a screenless AI device with a monthly subscription, got harsh reviews, and sold its platform, staff and patents to HP for $116 million while the Pin itself was shut down.
Lovable: a vibe-coding app builder that says it hit $100M ARR 8 months after its first $1M
A Swedish team rebranded GPT Engineer as Lovable, a browser app builder for non-developers. Company-reported growth was extreme; so was the security lesson for the apps people built with it.