Spend five minutes reading the commentary surrounding the IT services sector and you could be forgiven for thinking the industry’s obituary has already been written. Every earnings miss, every AI announcement and every productivity gain is being interpreted as further evidence that the traditional services model is entering terminal decline. Earlier this year, almost $280 billion was wiped from the market capitalisation of IT services and software companies. Since then, India’s largest IT services firms have lost close to half their combined market value from their August 2024 peak, while the Nifty IT index has recorded its worst first half since 2003.

The market has undoubtedly recognised something important. AI is not another incremental technology cycle. It has the potential to fundamentally reshape how enterprise technology is delivered, commercialised and consumed. On that point, I think investors are right. Where I think they’ve got ahead of themselves is in assuming that technological capability and commercial reality move at the same pace. They rarely have.

Much of today’s valuation reset appears to assume a relatively simple chain of events. AI becomes capable enough to automate work, enterprises rapidly reduce demand for traditional delivery, headcount falls and revenues follow shortly afterwards. That may ultimately prove to be the destination, but I don’t believe that’s how the journey unfolds. Enterprise technology has never worked like that.

The market is pricing the end state before it has priced the transition.

Every major technology shift over the past twenty years has followed a similar pattern. Robotic Process Automation was expected to remove large numbers of operational roles. Cloud was expected to commoditise infrastructure services. Machine Learning promised to automate significant parts of application support and business operations. Each prediction contained a great deal of truth. Each fundamentally changed parts of the industry. None of them happened at the speed financial markets initially expected, largely because enterprise adoption moves at the pace of organisational change rather than technological capability.

AI will almost certainly follow the same pattern. The technology is moving extraordinarily quickly, but large enterprises don’t replace operating models overnight. Existing contracts still need to be honoured. Critical systems still need to be modernised. Data still needs to be governed. Cyber security, compliance, procurement, architecture, legal approval and change management don’t disappear because a new foundation model has demonstrated another performance improvement. The constraint is no longer whether AI works. The constraint is how quickly organisations can absorb it.

That distinction matters because there is an enormous transition economy sitting between today’s delivery model and tomorrow’s AI-native enterprise. Before organisations can realise AI-driven productivity at scale, they need strategies, governance frameworks, data remediation, application modernisation, platform integration, operating model redesign, workforce transformation and implementation support. None of that represents a reduction in services demand. In many cases, it represents entirely new demand.

History suggests that major technology shifts often create more consulting and transformation work before they eliminate it. Cloud is a good example. It didn’t remove systems integration. It created one of the largest enterprise transformation markets the industry has ever seen. AI has the potential to follow a similar trajectory. The work changes long before it disappears.

There is another assumption embedded within today’s market narrative that deserves more scrutiny. Increasingly, investors appear to be comparing AI-enabled services businesses with software companies and expecting them to converge on similar economics. I don’t think that’s realistic. Software scales because the same product can be sold repeatedly with relatively little incremental effort. The services game solve multitudes of bespoke solutions for multiple organisations, and industries, operating under unique regulatory environments, commercial constraints and technology estates. AI undoubtedly reduces delivery effort, but it doesn’t eliminate implementation complexity or commercial accountability.

That’s one reason the incumbent providers still possess advantages that are difficult to replicate. Large enterprises don’t award billion-dollar transformation programmes simply because one organisation has access to a better model. They buy execution capability, governance, reputation, industry expertise and confidence that somebody will take responsibility if something goes wrong. Those characteristics remain valuable in an AI-enabled world.

Recent earnings reinforce this more nuanced picture. AI is already contributing meaningful revenue for providers such as TCS, while productivity gains are beginning to influence commercial discussions with clients. HCL continues to announce major transformation wins. Across the sector, providers are talking about AI-led productivity while continuing to secure large, complex engagements. Those aren’t contradictory signals. They’re exactly what you would expect during the early stages of structural change.

None of this is an argument that the industry can continue as it always has, it can’t. Traditional FTE-led commercial models will come under increasing pressure. Outcome-based pricing will become more common. Delivery organisations will become leaner and productivity expectations will continue to rise. Some providers will adapt successfully and others won’t.

Will AI changes IT services, ofcourse. The question though, is whether financial markets are correctly pricing how quickly that change translates into lower industry revenues, breakdown of contracts and structural upheaval. My view is that they aren’t. Markets have correctly identified the direction of travel, but they’re assuming enterprises transform at the same speed as technology evolves. History suggests they never have.

The market isn’t wrong about AI, it’s wrong about IT services.