How AI Coding Tools Are Reshaping Enterprise Software Buying

Nearly a third of companies surveyed by McKinsey this year didn't just delay a software purchase — they cancelled it outright, because their own teams could build the same thing with an AI coding assistant instead.

That's the headline number from McKinsey's State of AI in 2026 global survey, which polled 1,719 respondents across 97 countries between May and June this year. 32% of organizations say they've decided against buying at least one software product or feature specifically because agentic coding tools — AI systems that can write, test, and even deploy code with minimal hand-holding — made it cheaper to build it themselves.

The build-vs-buy math has flipped

For decades, "buy, don't build" was the default advice for anything that wasn't a company's core product. Custom software was expensive and slow, and a subscription to an established SaaS tool — software you rent and log into over the internet, rather than install and maintain yourself — was usually the safer bet. McKinsey's numbers suggest that calculus is breaking down fast, at least for companies with strong internal engineering benches.

The trend isn't even across industries. Some sectors are pulling away from off-the-shelf software much faster than others:

• Technology companies: 41% have skipped a software purchase to build in-house
• Healthcare payers and providers: 39%
• Professional services and energy: 38%
• Insurance: 19%
• Public and social sector: just 17%

The gap widens further among what McKinsey calls "high performers" — the 6% of respondents who say AI already contributes at least 5% of their company's profit. Nearly half of that group has skipped a software purchase in favor of building it themselves, against 31% for everyone else. Lieven Van der Veken, a senior partner at McKinsey, described the shift less as a simple cost decision and more as a change in mindset:

Organizations moving fastest are "becoming more deliberate about where to buy, where to build, and where to develop enough internal capability."

Productivity is up. Profit isn't, yet

Here's the part that should temper any triumphant narrative: 80% of respondents report real productivity gains from AI at the individual level, but only 37% see that translate into measurable profit impact at the company level — and that gap hasn't narrowed since last year's survey. Teams are shipping more code, faster, but a lot of organizations still haven't worked out how to turn that speed into actual business results. Building software in-house is cheap and fast now; running and maintaining it well is a separate problem, and one the survey suggests most companies haven't cracked.

What this means for India's IT industry

This shift matters more in India than almost anywhere else, because Indian IT services companies — TCS, Infosys, Wipro, HCLTech, Cognizant — built a multi-decade, multi-billion-dollar business on exactly the work that's now getting done in-house: custom enterprise software, built to a client's specification, usually cheaper than the client could do it themselves. If large US and European enterprises can increasingly spin up that same custom software with a small internal team pointed at an agentic coding tool, the traditional staff-augmentation and custom-development contracts that employ a huge share of India's roughly 5-million-strong IT workforce come under real pressure.

It isn't a one-way story, though. Indian IT majors have spent the last two years retraining engineers to supervise and review AI-generated code rather than write everything by hand, and several are now selling their own "agentic AI" delivery models to clients instead of resisting them. The companies that adapt fastest — treating agentic tools as something to sell expertise around, not just something clients use to cut them out — are the ones likely to hold onto revenue as this trend plays out.

The uncomfortable middle ground

What McKinsey's numbers really describe is a messy transition period, not a clean before-and-after. Companies are building more, buying less, feeling more productive, and still struggling to show it on a balance sheet. That's not a stable end state — it's a sign that most organizations, in India and everywhere else, are still figuring out what an AI-native software strategy actually looks like in practice, one skipped software purchase at a time.