Micha Kaufman on AI, Work and Fiverr’s Next Chapter

Harry Stebbings interviews Fiverr co-founder and CEO Micha Kaufman about AI, professional adaptation, startups and leadership. Fiverr operates a marketplace for freelance services, so its chief executive has a direct commercial stake in how AI affects work. His employment, technical and policy claims in the interview are opinions and forecasts, not a measured study of the labor market.

Kaufman explains a blunt message warning employees that AI could affect their jobs, including his own. He says it reflected questions already circulating inside the company and was intended to start a discussion about what people could do. He expected a small meeting; by his account around 250 people came to talk with him.

Asked what he told the group, Kaufman says he encouraged everyone to aim to automate as much of their current work as possible, ideally all of it, while acknowledging that complete automation may not be possible. He argues that freed time could go to strategy and work requiring judgment. This is his desired approach, not evidence every task can be automated or every worker will retain a job.

Stebbings raises employees' fear that adopting AI could eliminate their jobs. Kaufman responds that professionals must take responsibility for updating their skills; he says he will support those who want to learn but rejects the idea that an employer alone is responsible for making someone better. Stebbings argues partners and employers should also help people grow, while agreeing personal effort matters.

Asked whether people still want to work, Kaufman says the answer is personal and describes the present as an unusually confusing transition. He sees work as a way to earn a living, express oneself and find meaning. His harsh claim about people who do not work is a personal judgment, not an account of the many reasons someone may be unable to work.

Kaufman compares AI enthusiasm with the dot-com era and says there are too many companies competing for the same needs. He expects some to fail and others to consolidate. His examples and counts are impressions from his feed, not a census or a quantified forecast of market outcomes.

He calls the time needed to clone a product much shorter than when Fiverr launched: he estimates six to nine months then and perhaps ten days for some products now. These are illustrative estimates, not a controlled comparison. He says AI lowers the effort to build an initial product but does not solve the harder problem of reaching and retaining customers.

Stebbings argues Fiverr's defensibility lies in its brand, distribution and trust with suppliers, not merely its website. Kaufman agrees but says a shorter head start gives newcomers less time to build a community. He recalls choosing to focus on Fiverr's own innovation instead of pursuing every clone, and acknowledges the role of luck.

Stebbings suggests defensibility may shift toward the way a company collects and uses customer data. Kaufman agrees that understanding real needs and difficult problems matters more than building a marginal version of an existing product. Both are offering business judgments; Fiverr has an interest in the resilience of its marketplace.

Asked how an investor can choose among many similar AI startups, Kaufman says widely available technology raises the baseline for everyone. He prefers to assess the people: their talent, adaptability, commitment and ability to reason from fundamentals. He says he has invested in many companies, so this is also an investor's perspective.

Kaufman describes an unnamed seed investment he had effectively written off because of a difficult market and founder issues. It later went public, and he recalls a return multiple in the range of 50 times. He does not identify the company or specify the calculation, timing, realized proceeds or full portfolio results, so the anecdote is not a fund-level return.

Stebbings asks whether an AI bubble will burst. Kaufman predicts a severe shakeout and uses an extreme numerical estimate to express how many young companies could disappear, while acknowledging uncertainty about how the market will develop. He expects some durable foundational companies and applications to remain; his figures are not evidence of future failure rates.

He says startups must both attract and retain customers and expects investors eventually to force consolidation or closures by withdrawing support. He also predicts regulation and more expensive technology could change who can compete. The timing, cost and market structure are speculative.

Stebbings describes a bank that says compliance and privacy prevent it from using certain AI tools. Kaufman suggests technical and compliance solutions can emerge, including local hosting, and argues nothing is perfectly safe. This exchange does not assess the bank's obligations or prove any proposed deployment complies with law or protects data.

Kaufman sees a broader problem in distinguishing real from synthetic material and understanding manipulation. He worries that AI might weaken people's central role in creating and sharing knowledge. This is a philosophical concern about a possible future, not proof that systems are already autonomous or indifferent in the way he imagines.

He asks what happens when creators' work is used to train systems without credit or compensation. Stebbings asks why that means copyright is dead; Kaufman invokes the historical role of attribution and commercial rights in encouraging people to share work. His phrase 'copyright is dead' is a critique of enforcement and incentives, not a legal finding that copyright law no longer applies.

Kaufman says uncredited reuse could reduce people's motivation to create and share, while explicitly acknowledging he does not know whether that will happen. Although he says he is speaking as a person concerned about future generations, he is also CEO of a platform whose users sell creative services. The interview does not establish how training data was obtained or how any particular use is treated legally.

Asked whether public-company CEOs feel pressure to have an AI position, Kaufman describes it as a responsibility to navigate technological change. He compares a business to a surfer preparing for successive waves and says leaders must adapt the company as technology shifts. This is a leadership philosophy rather than a demonstration that Fiverr's strategy will succeed.

Kaufman compares leading amid rapid change to captaining a ship in a storm, with limited visibility. He is skeptical of precise five-year descriptions. Stebbings challenges that reaction, arguing long-term questions can stretch thinking; Kaufman agrees they are fair while emphasizing that regulation, technology and market structures remain unsettled.

Kaufman speculates that governments might restrict or take control of certain powerful AI capabilities if they become strategically significant. Stebbings asks whether open models make such power more widely available. Kaufman says he does not know the outcome and that many variables remain; neither nationalization nor a particular AGI timeline is established.

Stebbings worries that a small group of AI and infrastructure firms will capture most value. Kaufman compares the pattern to concentration in cloud computing and questions how new it is. Their discussion provides no systematic market-share analysis or conclusion about future distribution of wealth.

Kaufman says he likes being a public-company CEO if leaders understand the long game and do not imagine they can control the market. Stebbings asks whether public investors' demand for predictability discourages long-term bets. Kaufman argues strong execution, cash flow and a solid capital structure can support those investments, a view from an executive with an interest in public markets.

Asked about Fiverr's engineering team in five years, Kaufman predicts it will probably be larger, not smaller. His reasoning is that if competitors can use the same productivity tools, companies may still need more capacity to move ahead. He does not offer a headcount target or guarantee of employment.

Stebbings cites a case where code is produced faster than it can be released and asks when more output becomes harmful. Kaufman distinguishes speed from velocity: work must have direction, and bottlenecks in infrastructure should be addressed before adding features. He also says teams should stop building things that do not matter.

Kaufman wants to reduce the cost of experiments so a failed idea might take one person days instead of ten engineers months. He suggests increasing output and quality per unit of time, but these numbers are illustrative goals, not measured Fiverr results. His speculation about how many tests Amazon runs is also not company data.

He believes AI will dramatically change development and expects strong engineers to benefit especially. When asked if average developers will become ten times as productive, he is skeptical; he is more confident that exceptional developers will gain. Those judgments and multipliers are not established by comparative productivity evidence in the interview.

Asked which Fiverr function is most affected, Kaufman identifies marketing rather than customer support. He says coding tools can automate some repetitive research and implementation work, while marketers still have to understand human audiences. He cautions against giving generative tools too much freedom in customer-facing material.

Kaufman says the industry is emerging from an illusion that marketing can be fully automated. Stebbings contrasts routine corporate social posts with nuanced, timely creative work. Kaufman believes junior marketing roles have been replaced faster than junior development roles, but the interview supplies no hiring or displacement data to substantiate the comparison.

On leadership, Kaufman says he does not use AI to write his own communications. He jokes about automating investor-relations work for machine readers, then says other CEOs often plan on a nearer rather than later AI transition as a precaution. These impressions are not a survey of public-company leaders.

Kaufman says he has not seen widespread mass layoffs solely as a result of AI, though some companies have begun cutting costs. At Fiverr he wants to use the existing team's capacity before hiring more, and expects new hires across functions to be fluent in AI. His statements do not quantify Fiverr's job changes or prove what will happen to employment.

Asked what cost he would most like to cut, Kaufman names marketing, but explains he would do so if organic growth were strong enough to replace paid acquisition. He says Fiverr was not in that situation at the time. This is a conditional business preference, not an announced cost-cutting plan.

In quick fire, Kaufman says he once hoped modern AI would distribute power widely and now worries about concentration. He would prefer foundational technology to be open rather than closed. The benefits and risks of openness, and the governance of specific organizations, are not settled by the interview.

Asked what advice he would give a child entering the workforce, Kaufman says he would often advise learning independently rather than attending university. He claims most modern professions can be learned faster that way. This is his personal view, not evidence that degrees are unnecessary for regulated professions or all careers.

He says an early Fiverr version showed demand for an adult creator service resembling a later OnlyFans model, but the founders and board chose not to pursue it because of their values. This is his retrospective account of a missed strategic opportunity, not evidence Fiverr would have achieved a comparable outcome.

Kaufman says he reviews annually whether he remains the right leader, considering his energy, contribution, curiosity and interest. His personal warning sign would be waking up five days in a row dreading the job. He thinks a founder should be willing to find a successor if they can no longer serve the company well.

Asked about happiness, he says meaning matters more to him and can coexist with both difficult and joyful periods. He describes family values inherited from his parents and grandparents: giving rather than only taking, and making home a safe place for children to seek help. These are personal reflections, not a general formula.

Kaufman says those values also inform his management and describes himself as a builder who expresses creativity through businesses, products and communities. The conversation closes with his view of leadership as an ongoing practice of creating and adapting, rather than a settled prediction about AI's effects.