Builder.ai did not literally prove that it had no artificial intelligence. The London startup, formerly Engineer.ai, marketed AI-assisted app development, automation and a conversational product called Natasha. Yet reporting found that human engineers—many working in India and Ukraine—performed substantial portions of the coding and delivery work. In 2025, the company then suffered a severe liquidity crisis, faced scrutiny over revenue reporting, and filed for bankruptcy.
The defensible conclusion is narrower and more important than the viral headline: Builder.ai appears to have been a labor-heavy software business marketed as a highly scalable AI platform. Whether particular statements amounted to fraud remains a matter for courts, regulators or official investigations—not a conclusion established by the bankruptcy itself.
What Builder.ai actually sold
Founded in 2016 as Engineer.ai by Sachin Dev Duggal, the company promised that customers could describe an app in ordinary language and have it built faster and with less conventional coding. Its public positioning combined no-code and low-code ideas, reusable software components, automation and human implementation.
In its own 2022 Series C announcement, Builder.ai described “knowledge graph-powered code synthesis” and presented Natasha as a conversational AI product. That establishes what the company claimed to offer, not independent proof of how much code was generated automatically. The announcement remains available at Builder.ai’s funding announcement.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
Those categories are not interchangeable:
- No-code or low-code: customers configure prebuilt components with little traditional programming.
- Human-assisted automation: software tools accelerate work while engineers design, integrate, test and revise the result.
- Software outsourcing: people build a bespoke product for a customer, often using ordinary development tools.
- Machine-generated code: a model or automated system produces a significant share of the implementation with limited human intervention.
- End-to-end autonomous generation: a much stronger claim in which the system can reliably turn a specification into production software with minimal ongoing labor.
A company can legitimately use the first two while marketing an experience that sounds like the last one. The key question is the actual ratio of automation to human work.
What the “700 Indian engineers” claim means
Reports in 2025 described more than 700 engineers in India, while earlier coverage also discussed engineers in India and Ukraine doing much of the development work. The figure is an attributed estimate, not an independently audited final headcount. See the Times of India report and Rest of World’s account.
That does not establish that 700 people were literally pretending to be an AI system. Human staff are normal—and often necessary—in legitimate AI products. They may review model output, perform quality assurance, connect systems, handle customer requirements, fix defects or complete work that automation cannot yet do reliably.
The material issue is disclosure. Customers and investors would reasonably want to know:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- What percentage of a project is generated or assembled automatically?
- How much requires engineers to write, adapt or debug code?
- Are those engineers employees, contractors or an outsourced delivery team?
- Does the pricing reflect software usage, labor hours or a hybrid?
Calling ordinary engineers the “fraud” misses the economic question. A labor-intensive delivery model can produce useful applications while having the margins, scalability and risks of a services company rather than a software platform.
Rank #2
Was Builder.ai “fake AI”?
“Fake AI” is shorthand for several different allegations. The strongest supported statement is that Builder.ai was marketed as AI-powered, while reporting found that humans performed much of the valuable work. Commentators describe that gap as AI washing: presenting a product or service as more autonomous, technically advanced or scalable than the evidence supports. Advisor Perspectives discusses that broader issue.
Several stronger claims remain unproven:
- There is no established finding that Builder.ai had no AI software at all.
- It has not been established that every customer project was coded manually.
- The reported 700-person Indian workforce does not prove that every employee was impersonating an AI.
- No public court or regulator determination cited here establishes AI fraud.
- The company’s bankruptcy cannot be attributed solely to the use of human engineers.
Earlier reporting in 2019 had already questioned how automated Engineer.ai’s development process really was. The 2025 collapse therefore was not simply the moment someone discovered that software companies employ people. It was a later combination of operational, financial and liquidity problems.
How a private company reached a reported $1.5 billion valuation
Builder.ai raised more than approximately $445 million, with coverage putting the total in a range of roughly $445 million to $455 million. Investors named in reporting included Microsoft, the Qatar Investment Authority, Insight Partners and other institutional capital. Microsoft’s participation did not mean it owned Builder.ai or guaranteed its results.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →A financing-round valuation of about $1.5 billion was a private-company valuation. It was not a public-market capitalization, a cash balance or the value of assets that could necessarily be sold in an insolvency. Private valuations reflect the price and terms investors accepted in a funding round and can fall sharply when growth assumptions or financing conditions change. Moody’s analysis describes the reported valuation and later insolvency.
The attraction was understandable. The product addressed a real problem—making software development easier for nontechnical customers—and the AI investment market rewarded stories about lower labor costs, rapid growth and potentially enormous software margins. A polished demonstration can also look impressive even when a substantial delivery team works behind the interface.
The warning signs before the collapse
The failure developed over several stages rather than one revelation.
| Date | Reported development | What it means |
|---|---|---|
| 2016 | Founded as Engineer.ai | The company began with an app-development proposition later branded around AI and automation. |
| 2019 | Reporting questioned the extent of automation | Concerns about human engineering labor predated the bankruptcy by years. |
| 2021–2023 | Substantial venture funding and a reported valuation near $1.5 billion | Investors were valuing the growth story, not buying a public company with continuously disclosed results. |
| 2024 | Approximately $50 million borrowed from Viola Credit | Debt increased the company’s liquidity obligations. |
| March 2025 | Revenue expectations for the second half of 2024 reportedly cut by about 25%; auditors hired | Management and creditors were examining a material gap between expectations and performance. |
| May 2025 | Reports raised questions about sales projections and alleged reciprocal transactions with VerSe | The scrutiny broadened from technology claims to revenue quality. |
| May 20, 2025 | Viola Credit reportedly seized about $37 million, leaving roughly $5 million | The seizure created an immediate cash crisis. |
| June 2, 2025 | Chapter 7 petition filed in Delaware | This was the U.S. bankruptcy case; separate UK insolvency proceedings were also announced or planned. |
| June 5, 2025 | The Delaware filing became public through reporting | News coverage made the bankruptcy chronology widely known. |
The debt, seizure and cash figures come from Bloomberg Law. The March revenue revision and auditor review were reported by Bloomberg Law. The Chapter 7 date was reported by Bloomberg.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →The projected-versus-actual revenue problem
According to people familiar with creditor disclosures, creditors had reportedly been told to expect about $220 million in 2024 sales. Later reporting described actual revenue at approximately $50 million, with some coverage putting it closer to $55 million. Builder.ai reportedly reduced its expectations and engaged auditors to examine two years of accounts. The discrepancy is reported information, not a court finding that the company committed accounting fraud. See Bloomberg’s report.
That distinction matters because startup metrics can refer to different things: bookings, contracted sales, invoices, recognized revenue and cash collected. A forecast is not revenue, and revenue is not cash. Investors and lenders need to see how each number is defined, when it is recognized and whether customers actually paid.
What “round-tripping” would mean
Bloomberg-reported documents, as summarized by the Economic Times, allegedly showed Builder.ai and Indian social-media company VerSe billing each other for similar amounts between 2021 and 2024.
In plain English, round-tripping describes transactions in which companies record sales to one another and the economic activity is then effectively recycled or reversed. If the reported transactions were confirmed and lacked equivalent new demand or cash generation, they could make commercial activity look larger than it really was. The reports describe allegations based on documents; responsibility and legal consequences require official findings.
Recommended Free Tools
Why sophisticated investors still participated
Investor participation does not prove that the product was worthless, nor does it prove that diligence was adequate. Several forces can coexist:
- The generative-AI boom pushed investors toward large software-platform valuations.
- Microsoft’s involvement supplied credibility and potential distribution.
- Builder.ai addressed a genuine customer need and could deliver usable software.
- Human-assisted work can look like automation from the customer’s perspective.
- Private investors depend heavily on management projections, demos, references, technical reviews and financial representations.
A company can have real customers and useful output while still describing its automation, margins or growth too favorably. The central diligence question is not whether humans touched the product, but whether the business economics and disclosures matched the valuation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What customers should ask an AI software vendor
Companies considering an AI development platform should request specific evidence rather than rely on a conversational demo.
- Quantify automation: ask what proportion of output is generated automatically and what proportion requires human intervention.
- Identify the workforce: ask whether reviewers and developers are employees, contractors or outsourced teams, and where responsibility for quality sits.
- Inspect the trail: request access, where practical, to generated code, model calls, logs, prompts, revision history and deployment artifacts.
- Understand pricing: determine whether charges scale with software usage, engineering hours, project scope or a combination.
- Protect portability: confirm that you can export source code, data, workflows, prompts and deployment configurations.
- Plan for shutdown: ask what happens to support, hosting and unfinished projects if the vendor becomes insolvent.
- Check data rights: determine whether customer projects are used to train models and how confidential information is handled.
- Read service commitments: review uptime, response times, ownership, security obligations and remedies for missed commitments.
What investors should have measured
The Builder.ai episode makes unit economics more important than an AI label. A serious diligence process should test revenue per engineer, gross margin, the percentage of work that is reusable, customer concentration, cash collection, debt covenants and cash runway. It should reconcile bookings to invoices and bank receipts, inspect representative customer projects and verify how much of the product works without human intervention.
Best Value
Those checks also distinguish a software company from a managed-services business. Neither model is inherently bad. The danger comes when a labor-heavy operation is valued and financed as though it already has the margins and scalability of autonomous software.
Bankruptcy is not the same as a legal finding of fraud
In the United States, Builder.ai filed Chapter 7 in Delaware. In the United Kingdom, the company announced or planned insolvency proceedings. “Bankrupt” is acceptable shorthand for a general audience, but these are different legal processes.
Likewise, reports of overstated sales, reciprocal billing or misleading AI positioning are not the same as an adjudicated fraud judgment. The documented sequence supports a conclusion of financial distress and serious scrutiny. It does not, by itself, establish criminal or civil liability.
The broader AI-washing lesson
The useful lesson is not that Indian engineers—or human reviewers anywhere—make an AI product illegitimate. Human-in-the-loop systems can be safer and more effective than fully automated ones. The lesson is that companies must explain the boundary between software, automation and labor.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteFor investors, that means asking what is actually scalable and what is being delivered by people. For customers, it means securing source-code access, data portability and continuity protections. For journalists and readers, it means resisting both extremes: “there was no AI at all” and “the AI label proves the business was a software platform.”
The Bottom Line
Bottom line: Builder.ai’s collapse was not proof that humans writing software made the company fake. It was a warning about the consequences of presenting a labor-intensive, financially fragile operation as a more automated and scalable AI company than the available evidence supported.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
