Wall Street is currently panicking over AI’s potential to cannibalize traditional software companies. IBM recently bore the brunt of this anxiety, watching $70 billion evaporate from its market cap after earnings miss.
The knee-jerk reaction from investors is understandable: if advanced AI agents can write code and build applications autonomously, what happens to legacy tech giants?
IBM CEO Arvind Krishna has a remarkably pragmatic answer. While acknowledging that generative AI could automate up to 30% of standard software development across the industry, he insists this shift isn’t a death knell for enterprise tech.
Instead, he views it as a massive tailwind for companies positioned correctly. Here is the mechanical breakdown of how this disruption actually works on the ground.
Breaking Down the 30%: What Exactly Gets Replaced?
When Krishna discusses AI absorbing a third of software development, he isn’t predicting the end of computer science. He’s pointing to the aggressive commoditization of boilerplate coding.
We are rapidly entering an ecosystem where advanced autonomous tools, like Claude Cowork, can handle routine scripting, basic bug fixes, and straightforward SaaS deployments.
For startups whose entire business model relies on selling simple software-as-a-service wrappers around databases, the threat is entirely existential.
However, writing raw code is only a fraction of enterprise software engineering. The 30% that AI replaces is the repetitive grunt work—generating standard UI components, setting up basic API integrations, and translating logic into syntax.
By automating these layers, developers aren’t eliminated; their baseline output just shifts higher up the value chain.
Krishna’s underlying argument is rooted in how large language models actually function. AI is highly efficient at pattern matching and generating syntax based on existing repositories.
But it still requires intense human orchestration to design secure, regulatory-compliant, and highly integrated enterprise systems. The true disruption is happening at the lightweight application layer, not the heavy architectural foundation.
The Infrastructure Tailwind: Selling the AI “Picks and Shovels”
This technical reality explains why Krishna vehemently defends IBM’s market position despite the broader SaaS panic.
While a massive chunk of general coding might get outsourced to AI, Krishna notes that only about 2% of IBM’s specific software portfolio is vulnerable to direct AI replacement.
The reason for this massive gap is architectural. IBM has intentionally anchored itself away from the easily replicated application layer.
The remaining 98% of its software focuses heavily on what companies desperately need before they can deploy AI effectively.
You cannot run advanced machine learning models without unified, real-time data. You cannot deploy secure AI agents across a fragmented corporate network without rigid data governance and hybrid cloud infrastructure.
Krishna is essentially positioning IBM as the plumbing for the AI revolution. Unlocking siloed data streams, reducing the sheer complexity of managing it, and securing hybrid environments are multi-layered problems that autonomous agents cannot currently solve on their own.
Instead of competing directly with generative AI software, IBM is betting that the corporate rush to implement these tools will force enterprises to upgrade their underlying architecture. In this view, AI isn’t the enterprise software killer—it’s the ultimate catalyst for infrastructure sales.
Source: India Today, "IBM Earnings Disappoint; CEO Arvind Krishna Reveals How Much Software AI Can Replace"
