The longstanding economic frameworks that have underpinned the tech industry for half a century may soon be dismantled by the rise of agentic AI—autonomous software entities capable of executing complex, multi-step tasks with minimal human intervention. From the subscription-based SaaS model to advertising-driven platforms, the cascading implications threaten to rewrite value chains entirely. At the heart of this shift is a move away from selling access to software tools toward delivering outcomes, a change that could leave incumbents scrambling.

The model most visibly in the crosshairs is software-as-a-service. Since the early 2000s, SaaS has relied on per-user, per-month licensing, aligning costs with seat counts. Agentic AI breaks that logic: when a single AI agent can perform the work of dozens of human operators, per-seat pricing becomes economically absurd. Competitors offering outcome-based pricing—charging per automated resolution, per transaction, or even per business result—will undercut legacy vendors. Analysts cited in the piece project that by 2030, up to 40% of current SaaS revenue could be at risk if providers fail to shift to value-based metrics.

Advertising-based platforms are similarly vulnerable. Today’s internet runs on capturing human attention and monetizing it through targeted ads. But when AI agents browse, research, and transact on behalf of users, the human eyeballs vanish. An agent deciding between insurance policies or travel bookings will optimize for user preferences, not for the highest ad payout. This disintermediation could drain the reservoir of ad impressions, forcing Google, Meta, and others to either erect paywalls or reinvent their models around agent-to-agent commerce. One venture capitalist quoted in the article warns of a “zero-click future” where user interfaces are bypassed entirely.

The platform economy—marketplaces like Airbnb, Uber, or Amazon—faces a parallel unbundling. Agentic AI enables disintermediation at scale: instead of listing on a central platform, providers could deploy their own agents to negotiate directly with consumer agents across a decentralized mesh. Smart contracts and decentralized reputation systems could orchestrate trust without a rent-seeking intermediary. While today’s platforms benefit from network effects and data moats, AI-driven matchmaking may erode those advantages by making discovery and coordination cheaper outside walled gardens.

Fifty years of economic thinking around software—from the PC revolution (1970s–80s) through the internet (1990s) to cloud and mobile (2000s–10s)—has treated software as a tool augmenting human workers. The agentic paradigm inverts that: software becomes the worker. This shift reopens foundational questions about value capture. If an AI agent negotiates a supply contract, handles compliance, and manages logistics, does the value accrue to the company that owns the agent, the platform providing it, or the developer of the underlying model? Early movers like Adept, AutoGPT, and enterprise-focused startups are racing to define the commercial architecture, but no clear winner has emerged.

Several experts interviewed for the piece argue that the transition will be gradual but inexorable—a “scaffolding” being erected around legacy systems before those systems are dismantled. Incumbents are responding with bolt-on agentic features, but co-opting the technology may not save them from the economic logic. Incumbent SaaS vendors, for example, risk merely increasing churn if their agents reduce seat counts without new pricing models. Some, like Salesforce with its Einstein GPT and Microsoft with Copilot, are betting on productivity gains that expand total addressable market rather than cannibalize it, but skepticism remains.

Regulation and labor market disruption add further complexity. The article references a French government white paper exploring how agentic AI could displace white-collar roles en masse, necessitating new social safety nets and tax structures—perhaps taxing AI agents as “virtual employees.” And as agents gain economic agency, questions of liability, accountability, and antitrust will come to the fore. The European Union’s AI Act is only the opening salvo in a long regulatory battle to define the ground rules for an economy where non-human actors make binding decisions.

Ultimately, the piece frames the next half-century not as a simple technological upgrade but as an institutional transformation akin to the Industrial Revolution. The “scaffolding” of the current economic model—licensing, advertising, platform fees—may support the transition for a time, but those who cling to it risk being left hanging when the scaffold finally falls. The winners, the article concludes, will be those who reimagine value creation around autonomous AI outcomes, not those who merely wrap old packaging around new intelligence.

L’article IA Agentique : 50 ans de modèle économique à l’échafaud ? est apparu en premier sur Maddyness - Le média pour comprendre l'économie de demain.