The AI Displacement Spiral

By: John F Groom

How AI Erodes Established Businesses While Becoming Their Customers’ Default Way to Work

In 2021, businesses built around digital transactions seemed positioned for years of extraordinary growth. Freelance marketplaces, online homework assistance, image libraries, and software suppliers had benefited from a rapid shift toward remote work and online services. Investors extrapolated the boom. Five years later, AI is challenging the very transactions on which some of those businesses depend.

The central irony is that incumbents are often deploying AI themselves, automating work, reorganizing operations, and reducing their own staffing, while their customers use the same technology to purchase fewer of their services. A second mechanism may make the disruption self-reinforcing. The less frequently customers use an incumbent, the less they think of it. The more frequently they use AI, the more capable and comfortable they become. At the same time, AI itself continues to improve.

Upwork and Fiverr: From Pandemic Favorites to AI Competitors

Upwork went public in October 2018 at an approximate $1.5 billion valuation, while Fiverr followed in June 2019 at about $650 million. These IPO prices provide useful pre-pandemic reference points, although they already reflected expectations of future growth.

Their combined valuation climbed dramatically during the pandemic-era boom. If each company's separate 2021 peak is added together, the figure reaches approximately $19.2 billion. By late September 2026, the combined figure was approximately $1.33 billion. The peak figure is illustrative rather than a simultaneous market value on one trading day.

The operating figures explain part of the repricing. Fiverr's second-quarter 2026 revenue was about $97.8 million, down 10% year over year. Marketplace revenue fell approximately 15.5%, while annual active buyers declined almost 22%. Upwork reported revenue of about $191.7 million, down roughly 2%, while gross services volume declined approximately 4%. These are different businesses with different metrics, so the figures are not directly interchangeable. Upwork also reported growth in AI-related work.

The pandemic boom was already cooling before ChatGPT arrived. AI added a second source of pressure. A customer who once paid $500 for an illustration may now generate and revise it within an existing AI subscription. The freelancer loses the assignment and the platform loses its fee, even if the customer obtains a better result. Complex projects still require human judgment and accountability, but replacing routine transactions can materially change the economics of a marketplace.

The Irony: Using AI to Cut Costs While AI Cuts Demand

The same technology can affect a platform from both sides of its business model. Fiverr announced a roughly 250-person workforce reduction in September 2025 as part of its transition to an AI-first operating model. Upwork announced a substantial restructuring in May 2026 involving approximately 24% of its workforce.

These decisions also reflect profitability and organizational priorities, so AI is not the sole explanation for every eliminated position. Nevertheless, the underlying dynamic is notable. The same technology that reduces demand for freelance services can also allow the platforms themselves to operate with fewer employees. One substitution can therefore affect several layers at once: Customer spending → freelancer income → platform revenue → platform employment The effect is not necessarily limited to the original transaction.

The Opposing Feedback Loops: Lost Habits and Learned Habits

Consider a formerly frequent Upwork buyer who commissioned numerous graphics, software projects, and articles. AI initially replaces only some assignments. But each successful AI project teaches the buyer how to prompt, verify, revise, and combine tools. As models improve, more ambitious work becomes feasible. The buyer visits Upwork less often until AI, rather than a freelance marketplace, becomes the automatic first step for almost every project.

The marketplace then loses more than individual commissions. It loses mental availability. Even assignments that still require a human may no longer begin there. Freelancers experience a corresponding incentive. As earnings decline, they pursue salaried jobs, direct clients, or new businesses. Inactive profiles may remain online while the available and responsive talent pool weakens. Because many contracts are short and switching costs are low, neither side needs to formally terminate a relationship.

Leaving is easy. Winning back habitual use is harder. This creates two opposing feedback loops. On one side, AI replaces some paid assignments, customers visit the marketplace less frequently, sellers earn less, and both groups develop alternatives. On the other, customers use AI more frequently, their skills and confidence improve, new model capabilities expand the range of tasks they can handle, and AI becomes an increasingly natural starting point.

These mechanisms may reinforce one another, but they are not inevitable. An incumbent can still adapt and restore customer value. Freelancers can also adopt AI and increase their productivity. The platform, however, must still demonstrate why paying for a transaction through it creates additional value.

Chegg: When the Educational Subscription Is Displaced

Chegg sold access to textbook solutions and homework assistance and benefited from pandemic-era remote learning. General-purpose AI changed the interaction. Instead of beginning with Chegg, students could ask questions directly, request personalized explanations, and generate practice material through general-purpose AI tools.

Chegg's equity value fell from an early-2021 peak approaching $15 billion to a small fraction of that amount by September 2026. Its valuation was already falling before ChatGPT, so the entire decline cannot be assigned to AI. Subsequent declines in academic-services demand nevertheless demonstrate direct pressure.

Chegg is attempting to build AI-powered career and skills offerings, creating another example of an incumbent using the technology that challenges its existing business model in an effort to reinvent itself.

Shutterstock and Getty Images: When Images Are Generated Rather Than Licensed

Stock-image suppliers face a closely related change. For generic commercial imagery, customers can increasingly describe the image they need rather than licensing and adapting an existing photograph or illustration. Shutterstock's valuation fell sharply from late-2021 levels, while Getty Images' equity value also contracted substantially.

Their proposed merger, announced in 2025 and terminated in 2026, highlighted the strategic pressure facing the industry. Getty's debt and merger developments also make simple equity-value comparisons incomplete.

Not all images are equally exposed to synthetic generation. Authentic event photography, archival imagery, provenance, and rights-cleared depictions of real people remain distinct from synthetic imagery.

The historical irony is striking. Bill Gates founded Corbis in 1989 to commercialize digital image rights, and its licensing assets later became associated with Getty's distribution network. Digital imagery grew enormously in importance, but generative AI now challenges the need to purchase many pre-existing images.

Teleperformance and Concentrix: Outsourcing Their Own Labor Advantage

Call-center and business-process outsourcing companies built scale by recruiting and managing large workforces. AI voice agents and chatbots give customers an alternative to outsourcing routine interactions. At the same time, the outsourcing providers themselves can use AI to reduce labor costs.

Teleperformance and Concentrix have experienced major valuation declines from earlier levels, although client spending, debt, execution, and other business conditions also matter. Their long-term enterprise contracts may slow switching, but they can also create opportunities for customers to demand lower prices when automation reduces the amount of labor required. The same technology can therefore challenge both the customer's need for outsourced labor and the provider's traditional cost structure.

Software: Adobe, Salesforce, Workday, Atlassian and Intuit

The situation is different for established software companies such as Adobe, Salesforce, Workday, Atlassian, and Intuit. These companies illustrate anticipated disruption rather than a demonstrated collapse of their businesses.

Generative AI can create and edit visual assets, automate customer-service and sales tasks, handle portions of HR and finance, assist software development, and perform accounting or tax-related work. Investors have therefore reassessed whether conventional interfaces and per-seat pricing will retain the same value when customers increasingly direct AI agents to accomplish tasks.

But these companies also possess significant advantages. They have trusted customer relationships, enterprise data, integrated workflows, and the ability to incorporate AI into their own products.

Adobe's creative tools, Salesforce's business systems, Workday's enterprise records, Atlassian's project tools, and Intuit's financial applications could become more valuable with effective AI integration. A falling share price alone does not establish that AI has already destroyed a company's underlying revenue.

Financial Services: Schwab, Raymond James, LPL and Others

AI-related market anxiety also reached financial services in February 2026, when new AI-enabled tax-planning capabilities contributed to sharp selloffs in Charles Schwab, Raymond James, and LPL Financial, alongside declines in Ameriprise, Stifel, and Morgan Stanley. These reactions reflected possible future competition rather than proof of equivalent losses in current business.

AI can automate research, tax scenarios, and routine client interactions, while human advisers may retain advantages in complex judgment, trust, regulation, and accountability. Financial firms can also use AI to increase adviser productivity. The important distinction is therefore between repricing based on anticipated technological change and evidence that existing revenue has already been displaced.

Three Different Forms of Evidence

The examples in this article fall into three broad categories. Observed pressure on established services: Upwork, Fiverr, and Chegg have reported deterioration in parts of their core activity, with AI identified as one contributing factor. Substitution risk amid multiple pressures: Shutterstock, Getty Images, Teleperformance, and Concentrix have business models vulnerable to technological substitution, but company-specific factors make it difficult to attribute their changes to AI alone. Primarily anticipated disruption: Adobe, Salesforce, Workday, Atlassian, Intuit, Schwab, Raymond James, LPL, Ameriprise, Stifel, and Morgan Stanley illustrate markets where investors have repriced potential future risk without evidence of equivalent AI-driven revenue losses.

Keeping these categories separate is important. A decline in market value is not itself proof that AI caused the decline, and a company facing technological risk is not necessarily a company whose existing business has already been displaced.

The Capability Economy and the Disappearing Transaction

The deeper issue extends beyond individual companies. Consider the lost $500 illustration commission. It is recorded as less freelance activity, yet the customer may generate more and better graphics for a fraction of the cost.

Similar substitutions can occur in tutoring, stock imagery, routine support, and other services. Traditional revenue measures capture the disappearing transaction but do not necessarily capture the expanding capability. At the same time, the transition costs are real.

Displaced workers, shareholders, and businesses can experience substantial losses even when customers gain new capabilities. More output and less expenditure can coexist with severe losses for particular suppliers.

This is why technological disruption cannot be understood simply by looking at whether a transaction has disappeared. The economic question is also what replaced it and who captures the resulting value.

Conclusion: The Default Is the Battleground

The critical contest is not simply whether AI can complete an individual assignment. It is where the customer starts the next assignment. Upwork and Fiverr are especially exposed because their short contracts make buyer and freelancer disengagement relatively frictionless. Chegg, image libraries, outsourcing companies, software firms, and financial advisers face different versions of the same challenge: retaining a reason to be chosen when AI becomes the first step in more activities.

The strongest form of disruption combines substitution with habit formation. AI removes some purchases today. It teaches customers to rely on it tomorrow. Then it expands what those customers can accomplish the day after.

Incumbents may adapt by embedding themselves in AI workflows or by offering expertise that automation cannot replace. But the old transaction will not return merely because it once supported a valuable business. The central question is therefore not only what AI can replace. It is whether AI becomes the customer's default starting point for deciding how work gets done.