Builder.ai insolvency revives the AI-washing debate
UK-based Builder.ai, backed by Microsoft and Qatar, has announced that it is entering insolvency. Its collapse combines accounts under review, an automation promise questioned since 2019, and the limits of selling AI as a complete substitute for technical work.
Builder.ai, the London-based company that promised to create custom applications with the help of artificial intelligence, told its employees on Tuesday that it would file for insolvency. The collapse affects one of the most visible companies in the applied AI software-development business: it had attracted more than $500 million from investors including Microsoft and Qatar’s sovereign wealth fund.
The case matters for a reason that goes beyond one failed startup. For years, Builder.ai sold the idea that its platform — and particularly its Natasha assistant — could radically simplify app development. But its product relied heavily on human engineering teams, a reality that had already been examined by The Wall Street Journal in 2019.
A financial collapse after reaching unicorn status
Builder.ai came close to a $1 billion valuation and announced a $250 million funding round in 2023. Microsoft not only invested in the company; it also announced an Azure-based partnership and plans to integrate the Natasha assistant into Microsoft Teams, an alliance that lent commercial credibility to its pitch.
The company has now run out of money. Manpreet Ratia, appointed CEO in February, told employees that Builder.ai had been unable to overcome inherited financial problems and past decisions that put significant strain on its finances.
The situation is particularly delicate because the company’s accounts were under review. Financial Times reported that Builder.ai had substantially reduced its revenue figures, from around $220 million initially attributed to 2024 to roughly $55 million. That difference does not, by itself, establish a specific cause of the insolvency, but it intensified doubts about the business’s forecasts.
Insolvency does not necessarily mean immediate liquidation. In the UK, the specific proceedings will determine whether the company enters administration, sells assets or ultimately goes into liquidation. For customers and employees, however, the practical consequences are uncertain: the service, existing contracts and technical support now depend on what happens to the company’s administration.
Natasha and the humans behind the automation
Builder.ai’s most persistent controversy was technological, not financial. When it was still called Engineer.ai, the company was criticized for presenting as automated a process that required the involvement of around 700 human engineers. The company argued at the time that those professionals were part of its development model and that the platform used automation to organize and accelerate the work.
That distinction matters. An AI tool needing human oversight is neither fraud nor an anomaly: today’s code-generation systems are routinely used as assistants. They can draft functions, explain errors, prepare tests and speed up repetitive tasks. But they cannot reliably replace the judgment of someone who understands a business’s requirements, a system’s security and the consequences of modifying existing code.
The problem arises when human involvement is hidden behind a promise of autonomy. The classic precedent is the so-called Mechanical Turk, an alleged 18th-century automaton that played chess but actually concealed a person inside it. In the technology industry, the expression is used to describe apparently automated services that depend on human labor invisible to the customer.
Builder.ai was not simply a generative AI company: it also offered project management, design and application development. That combination can be useful. What was questionable was the gap between the image of an AI-powered software factory and the reality of a service organization dependent on a large engineering workforce.
Coding AI does not eliminate technical responsibility
Builder.ai’s collapse comes as major technology companies push coding assistants aggressively. Microsoft CEO Satya Nadella said last month that AI had already written 30% of the code in some of the company’s repositories. The figure shows that these tools have become part of everyday work, not that they can deliver a complex product without review.
Code generated by language models can contain errors, invent nonexistent functions or suggest insecure solutions. That is why a company adopting these tools needs developers who can review and test the output and take final responsibility for it. Productivity may increase, but the need for technical expertise does not disappear.
For customers and businesses buying these services, the lesson is less spectacular than the original promise: it is worth asking which part of the service is software, which part is automation and which part is performed by people. It is also important to demand verifiable business metrics, especially when a company presents AI as the reason for exceptional margins or growth.
Builder.ai does not show that AI for software development has no value. It shows that a polished sales demo, the backing of well-known investors and the label “AI-powered” cannot replace a sustainable financial model or a precise explanation of how the product is actually built.
