The Agentic AI invasion: It’s here—but does the software industry face an existential threat?

The trigger was Anthropic’s decision to open-source 11 ‘plugins’ on GitHub for its Claude Cowork AI platform. Compatible with Claude Code, its coding assistant, these plugins let users set goals for its AI model to achieve. Each plugin guides Claude on which tools and what data to use for every task. Together, this toolkit spans productivity, sales, marketing and legal work.

What rattled markets the most was its legal plugin for reviewing contracts, compliance and risk. The panic was understandable, but contracts and revenues have not changed. So, did investors overreact, or do IT firms face a mega-threat?

Until now, humans wrote the script. AI tools such as Replit, Bolt and Claude helped refine it as assistants. Now, an agent can write the script, analyse it, test it, improve it and even deploy it, which explains the market’s extreme reaction.

The first casualties were predictable. Large tech firms with product-heavy business models driven by software-as-a-service (SaaS)—especially those behind banking, financial service and insurance workflows—were hit the hardest. Indian IT majors that sell licensed platforms alongside services suddenly looked exposed.

One may argue that IT firms already use AI platforms like Kiro, Claude and Replit to build their own for clients. But if an agent can reconcile data, generate compliance reports, monitor transactions and flag anomalies, the case for paying recurring licence fees weakens over time.

Globally, the same fear rippled across enterprise software. Layered on top is relentless rhetoric on artificial general intelligence (AGI) that spooks investors with endgame scenarios. Every agent’s launch seems like a step towards that ‘singularity,’ overhyped or not.

That said, markets may have missed key nuances. For instance, agents today excel at preset tasks, not designing complex workflows for organizations with messy legacy systems. That still calls for human architects.

Besides, agents automate work but do not yet own outcomes. So businesses with customized services are at less risk than those counting on licensed stuff—at least for now. Application development and maintenance revenue will not vanish overnight either. Banks, insurers and governments still run old systems. Agents can help, but humans remain essential—particularly in regulated fields.

Markets also need to take the true cost of agents into account. Replacing humans is not a simple swap. A developer’s salary is visible, while an agent’s cost is buried in cloud bills, compute cycles, storage, redundancy, uptime guarantees and energy use. Just to break even, the AI industry will need to charge far more at some point. AI adopters must also price in risks of security and legal liability. Automation won’t be cheap for adopters that must not go wrong. Caution could keep humans in charge.

Yet, market fears are justified in some fields (like SaaS). So, was the tech-stock rout overdone? In the near term, yes. Revenues will not be wiped out. But over time, Agentic AI will surely shrink, reshape and reprice parts of the tech industry. What we don’t know is the profit patience of AI majors and the speed of this agentic shift.