The Web’s Middlemen Are Being Repriced for the AI Era

The AI disruption story is often told as a contest between humans and machines. For professionals and investors, a more useful framing is a contest between commodity output and accountable outcome. Generative models have made first drafts radically cheaper. The businesses under the greatest pressure are those that once charged for access to a standard digital result.

Fiverr is the most dramatic example. Its shares closed around $11.06 in late July 2026, compared with a closing peak of $323.10 in February 2021. That is a decline of roughly 96.6%. First-quarter marketplace revenue also fell, and active buyers dropped 17.8% year on year. For a simple logo, short copy or small coding job, a buyer can now try AI before hiring anyone.

Stock imagery faces the same substitution risk. Shutterstock’s second-quarter revenue fell 17% year on year and subscriber numbers declined to 951,000. The company’s market value was around $216 million in late July. Getty Images was worth roughly $171 million, about 92% below its late-2023 level. Generative images clearly put pressure on commodity stock demand, although the companies also face merger, litigation and execution issues that make a single-cause explanation too neat.

The counterexample is Upwork. It is also valued far below 2021 levels, but its operating business remains substantial. First-quarter 2026 revenue rose 1%, gross services volume was nearly flat around $987 million, and AI-related work grew more than 40%. Demand for AI-related skills rose 109%.

That contrast points to the new career hierarchy. The ability to produce a first version of text, code or imagery is becoming less scarce. The ability to define the right problem, select tools, verify facts, integrate systems, understand risk and take responsibility for the final outcome is becoming more important.

The same logic applies to the web. Stack Overflow’s near-99% collapse in new questions from peak levels is not proof that developers no longer value human knowledge. The platform itself says more than 80% still visit regularly and that users often seek another human when AI is untrustworthy. What AI has removed is the need to make every routine question public.

For leaders, the practical response is not to ban the tool or compete with it on speed. It is to redesign roles around judgment. A consultant who merely summarizes information is vulnerable. A consultant who decides what the information means for a high-stakes decision is not equivalent to a chatbot. A designer who produces generic variations is exposed. An art director who owns a brand system is harder to replace.

AI is lowering the price of output while raising the strategic value of accountability. That is why some digital middlemen are collapsing and others are evolving. The winners will not be those who pretend AI cannot do the work, but those who build businesses around the parts of work that still require trust.