Software’s Death by AI Is Exaggerated

Artificial intelligence can do a lot these days—and will be able to do a lot more in the future. But killing a $1.2 trillion industry will be a stretch. That is how much the world’s businesses are expected to spend on enterprise software this year, according to projections from market research firm Gartner. It is a big number, and nearly 11% higher than the $1.1 trillion that was spent last year. Businesses spend more on software than just about any other technology-related category. 

But the age of AI has made that budget line look increasingly under threat. Among the early promises of tools such as ChatGPT was that it could allow novices to create software by simply telling the LLM what they want in natural language. Such “vibe coding” could theoretically render premade software obsolete. In a live demo of its new GPT-5 model earlier this month, OpenAI employees created an app for teaching French to English speakers in a few minutes, all on stage. “We think this idea of software on demand is going to be one of the defining characteristics of the GPT-5 era,” OpenAI Chief Executive Sam Altman said at the event. He is actually a bit late to the party; Nvidia CEO Jensen Huang proclaimed in 2017 that “AI is going to eat software.” The prospect of an existential threat has cast a darker shadow on an industry that was already under a cloud. 

Global economic uncertainty caused by trade wars, real wars and the prospect of inflation has created a “business pause on net-new spending” on corporate IT needs, Gartner’s analysts wrote in a report last month. Booming investments in AI infrastructure are also pressuring other areas of the corporate tech budget line. Wall Street is getting worried. Analysts are asking executives about topics such as “AI disruption” with increasing frequency, according to an analysis of earnings calls and other events by AlphaSense. Software stocks have also been among the weakest categories in tech of late. The BVP Nasdaq Emerging Cloud Index has slumped nearly 6% over the past month and is the only major tech subcategory to be in the red for the year to date, according to Fact- Set. 

AI tools are certainly changing the game for software development. JPMorgan Chase Chief Financial Officer Jeremy Barnum told analysts at a meeting in May that he had been indulging in vibe coding, describing it as “pretty amazing.” But highly complex software applications running missioncritical tasks won’t be simple to replace. Especially those running on data related to sensitive—and often regulated—areas such as human resources and finance. 

“More than 65% of the Fortune 500 use us, and not one of them is going to say ‘Come in here AI startup and run my back office and financial controls,’ ” Workday CEO Carl Eschenbach said in an interview. Companies such as Workday and Salesforce, which disrupted the more traditional on-premise software industry not too long ago, are working furiously to adopt AI themselves within their own products. 

Agents—AI chatbots that are enabled to take certain actions on behalf of people—are a big part of this effort. Nearly every softwareas- a-service company is now selling agent tools to their customers, though RBC analysts argued in a report this week that the real opportunity will come from “multiagentic” systems that can operate across different software applications, which no one has really cracked yet. “For example, a sales leader should be able to use natural language to onboard a new salesperson,” RBC’s report read, executing across HR identity systems, expense cards, training programs and so on. 

Vibe coding is unlikely to pull that off. In his own report earlier this month, longtime software analyst Brent Thill of Jefferies cited the “meaningful complexity” across current business software workflows that AI cannot yet replace. 

“In our view, the intricacies of enterprise architecture make full AI disintermediation of software unlikely,” he wrote. Stumbles by AI companies will help make that case. OpenAI faced a flood of criticism following the GPT-5 launch, with users complaining about inaccurate answers from a chatbot that Altman touted as the equivalent of “having a team of Ph.D. level experts in your pocket.” Meta Platforms has also struggled with taking its Llama 4 model to the next level. AI chatbots may have quickly disrupted high-school term papers. Replacing billion-dollar software systems won’t come so easily.