Ali Ansari’s early inventory included his father’s shoes. At 12, he put a pair on eBay, priced them low and found a buyer almost immediately. He was delighted. His father had a practical question about why his shoes had been sold. His mother encouraged the enterprise. Between them, the family had supplied a business education: demand exists, pricing matters, and it is wise to establish who owns the merchandise.
The next step was to buy things before reselling them. Years later, the founder of micro1 would be dealing in something considerably harder to price: the judgment of people who know how to do difficult work. His company recruits experts whose contributions help train and evaluate AI. The distance between those transactions is large. The habit connecting them is easier to recognize. Ansari pays attention to what somebody else wants, then rearranges the business around it.
A bedroom with an inventory system
Ansari grew up in Tehran before his family moved to Los Angeles. His parents rebuilt their working lives in the United States: his father in kitchen remodeling, his mother through a daycare business. Entrepreneurship was close enough to watch at home. It also involved starting over, with the ordinary obligations that grand accounts of opportunity tend to leave off the balance sheet.
His own textbook venture, CashBooksNow, gave that instinct a system. His bedroom had separate areas for books already listed for sale and those still waiting. By 16, he had sold more than $100,000 worth of books. That figure described sales, but the two piles tell the more intimate story. A teenager had turned part of his room into a small warehouse. Domestic space was becoming working capital.
Then came tutoring. Math-competition lessons with friends expanded into a broader math marketplace. A narrow enthusiasm met a larger customer need. There would be several more versions of that encounter.
Customers wanted the engineers
At UC Berkeley, Ansari ran a software development agency building apps and websites. Every project brought a staffing problem. Before assigning engineers, he had to find them and decide whether they could deliver. The company’s capacity depended partly on how much interviewing its founder could personally fit into the day.
Customers introduced another possibility. They liked the engineers and asked to hire them directly. Ansari experimented with selling access to vetted engineering talent. Within three months, micro1 had outgrown the agency. He closed the older business to concentrate on the new one. A useful side service had acquired the authority to rewrite the plan.
The screening tool addressed the same constraint from a different direction. An AI system could handle an early assessment; Ansari could spend his time on the candidates who progressed. By the time he discussed fundraising in 2024, the problem included persuading investors to see a technology business in work they might otherwise classify as staffing. The product had changed. Explaining the product took work too.
- Build the softwareAgency projects
- Find the builderAI-assisted recruiting
- Teach the modelExpert training data
The student who recruited his professor
School presented another allocation problem. Ansari had studied computer science and mathematics at Berkeley. He dropped the mathematics component and finished in three years. He had considered leaving altogether. Completing the degree early gave him another route into spending more time on the company, without turning his education into unfinished business.
Stanford brought research relationships as well as classes. In 2025 he was studying part time while running micro1. He also approached Andrew Maas, his spoken-language-processing instructor, about becoming the company’s engineering leader. An unanswered message became follow-ups, conversations and, eventually, a hire. Maas joined as employee number 80.
It is an appealing reversal of the usual classroom transaction. Students generally hope a professor will help them find a job. Ansari had a vacancy of his own. The persistence matters more than the reversal: recruiting someone with other options can take a year, even when the person making the offer is accustomed to moving quickly.
Putting the interview to a test
Ansari also put his name on attempts to measure the recruiting machinery. In a 2025 experiment he coauthored with Ada Aka, Emil Palikot and Nima Yazdani, 37,000 junior-developer applicants were randomly assigned to conventional screening or a process that included an AI interview. Candidates who advanced faced a final human interview. Those interviewers did not know which earlier route the candidates had taken.
Among candidates reaching that stage, the pass rate was 54 percent for the AI-assisted pipeline and 34 percent for the traditional one. The comparison concerns a particular hiring process and outcome. It gives the recruiting thesis something more useful than a demonstration video: a result with a defined population and a human assessment at the end.
A separate paper, coauthored with Yazdani and Aruj Mahajan, described Zara as a system for practice interviews, conversational assessment and personalized feedback. That work treats the candidate’s experience as part of the product. An interview is an information-gathering device, but someone still has to sit through it.
When the recruiter changed sides
The next turn arrived through a customer using micro1 to recruit engineers for data work. Large, urgent hiring requests exposed a demand beyond conventional software teams. AI businesses needed people capable of creating and judging the material used to improve models. Ansari could see his recruiting infrastructure becoming the entrance to a different business.
In early 2025, micro1 secured a major frontier AI lab as a customer. The relationship expanded quickly. The company moved toward providing the expert work itself: locating people, assessing their knowledge and managing the production of training and evaluation data. Years spent building the recruiting system suddenly served a new buyer with a new purpose.
The continuity is easy to miss amid the word “pivot.” Ansari was changing what customers paid for while retaining a way to find the people needed to deliver it. Starting again did not require throwing everything away. It required deciding which part of the existing company was worth carrying forward.
The founder supplies the risk
Ansari describes the founder’s role as introducing upside risk that others in the company may have little reason to propose. He names hiring, product and aligning incentives as his central concerns. In a fast-expanding market, he also makes room for immediate rewards: recruitment bonuses tied to large hiring goals, or an equity increase linked to closing a customer.
His account of onboarding is equally direct. New core-team members begin the actual work on their first day. A salesperson joining near the end of a quarter can still receive a proportional quota. He has described a four-week period in which the company and the employee test the fit.
That is a demanding proposition for a workplace. It also explains something about the founder’s pace. The calendar is already running when someone arrives. There is little ceremony between accepting the job and being expected to do it. Whether that arrangement suits someone is a consequential part of the match.
A machine that still needs its teachers
Ansari’s commercial argument runs against the idea that better models eventually eliminate the need for human training input. He argues that synthetic data can multiply the usefulness of expert contributions. People establish the underlying judgment; machine-generated variations extend its reach. In that account, automation changes the value of the human contribution as well as its volume.
He has developed the argument beyond company presentations. In March 2026, he coauthored No Last Mile with Mark Esposito, Ava Fitoussy and Liu Zhang. Their model treats structured human data as a continuing input. Tasks shift, standards evolve and fresh exceptions appear. Capability requires maintenance.
The attraction of that thesis for micro1 is plain. So is the question it leaves open for workers. Training a system can create paid work today while changing the market for the underlying skill tomorrow. A theory about lasting demand does not settle who benefits, how the work feels or how the rewards are divided.
“I hope AI systems will be managed by humans, forever.”Ali Ansari, October 2025
The numbers got bigger. The question stayed human.
In September 2025, micro1 announced a $35 million Series A at a $500 million valuation. A year later, a new financing round was reported at more than $100 million and a $4 billion valuation; the company declined comment. In 2026, micro1 also placed No. 37 on the Inc. 5000.
Those milestones mark a considerable change in scale. They describe different things: money raised, a price attached to the company and a ranking based on growth. None is a substitute for understanding the work Ansari has chosen to organize. That work now reaches beyond recruiting into expert data, realistic environments for AI agents and robotics training.
His 2025 funding announcement opened with a question about where humanity should spend its time. The ambition is to match people with work they find meaningful. It is a much harder assignment than moving inventory from one side of a bedroom to the other. People have preferences. Skills develop. The job that fits can change.
Ansari’s history gives the question a practical texture. He has repeatedly found a new market by watching what people asked him to do next. The next request may change micro1 again. For now, his business depends on finding people who can teach a machine something it has yet to learn.