AI was hyped as the golden ticket for IT services, promising explosive revenue growth and soaring valuations. However, the reality on the ground paints a different picture. Instead of flourishing, the sector seems to be experiencing a slowdown, raising critical questions: Why are Global Service Integrators (GSIs) missing out on the AI revolution? And will AI cannibalize existing services more than it creates new ones?

The Valuation Disconnect

The GenAI age began in November of 2022 with the introduction of ChatGPT. Since then:

• NVIDIA’s stock is up 736%

• The Magnificent Seven (i.e., Facebook, Google, Amazon, Apple, Microsoft, NVIDIA and Tesla) are up 279%

• Enterprise Software Leaders (i.e., SAP, Oracle, Salesforce, Workday and ServiceNow) are up 159%

• The S&P 500 is up 33%

• The top GSIs (i.e., Accenture, CapGemini, Infosys, DXC, Cognizant, and Wipro) are down 4%. Relative to the S&P 500 Index this group has underperformed by 28%.

Minus 28%. Let that sink in for a moment. GSIs, in spite of a cumulative $10 billion in announced GenAI investments, have decoupled from the AI boom.

As an investment thesis, these firms have (historically) delivered consistent outperformance, acting somewhat like technology mutual funds run by “insiders,” for they enjoy:

• Unparalleled access, being privy to the tech strategies of both major corporations (buyers) and hardware/software leaders (sellers).

• The ability to ride big waves by capitalizing on the momentum of product companies, often generating a 5:1 service-to-product revenue ratios.

• The agility of malleable workforces, which can consistently grab the upside of market winners and mitigate any downside product risk.

But the concern is that GenAI will soon automate much work in the current model — pushing more work loads from people to software — thus eating into future cash flows.

Automation: The line between product and people is changing

This line between software and services — which has been stable for decades — is now being redrawn.

Automation of the Software Development Life Cycle (SDLC) is starting to occur — but not in the way many anticipated. GenAI is not eating the SDLC from its head (e.g., consulting, analysis, design and development); it’s eating it from the tail. And it’s left us with Automation and Confusion zones.

The Automation Zone

Quality Assurance (QA) and Testing, Application Maintenance and Application Modernization are now the focus of automation. These tasks are highly process-oriented, relying on repetitive procedures like checking for bugs across different platforms. Here, AI excels. LLMs can efficiently analyze mountains of code, identifying patterns and potential issues with impressive accuracy.

Clients are already demanding 15–40% efficiency gains via AI in this zone. In turn, the associated two+ million IT services workers suddenly face a stark reality; they must either upskill to become AI-powered “super consultants” or risk being displaced.

The Confusion Zone

Predictions of services automation at the front end of the cycle existed for good reason. After all, the VC community has poured more than $2 billion into code generation platforms (e.g., GitHub Copilot, Cursor, Replit Ghostwriter, Augment, Sourcegraph Cody, Codeium and Kite) and GSIs stand as their top prospective customers.

In looking at Copilot alone, Accenture has given it to tens of thousands of employees and claims 67% of developers use it daily. KPMG announced a $2 billion alliance with Microsoft. Cognizant now has 25,000 seats. And Infosys recently announced this as the “Age of Copilot.”

Yet for all of this effort there’s no recognizable revenue or margin lift. Additionally, clients now have POC fatigue, with many stating “we’re being POC’ed to death.”

A generation ago (with Client/Server then SaaS) massive application markets opened up with CRM, HRM, IT Management, ERP and Supply Chain. Such scale use cases have yet to emerge with AI, and it’s certainly not for lack of trying.

This may still change, and we may be in a gestation phase for enterprise AI at the app layer. But for now this wave is inverted from the Cloud wave, for as recent as five years ago app spending outweighed the supporting infra and semi spend 2:1. But thus far in the GenAI journey spend on semis (NVIDIA) and infrastructure (Azure, AWS, Google, OpenAI) is outweighing AI application revenue 20:1.

A final (random) thought: WIll somebody be willing to go full Elon?

Elon Musk cut about 80% of Twitter’s staff and…nothing much happened. OK, we did lose those delightful “Here’s my day as a Twitter employee” selfie videos from underutilized employees. But the platform kept rolling.

So, one wonders if any IT services leaders will have the courage to chase Moore’s Law to its conclusion and radically change the headcount vs. software model.

Never forget that winning IT services models correlate highly to what technology wants: EDS with mainframes, Accenture in Client/Server, TCS — Infosys — Cognizant — Wipro with the Internet, and EPAM and Globant with mobile.

18 months into the GenAI era investors, clients, partners and (most importantly) the 12 million employees in the industry are searching for a new model they can all bet on. It’s not here yet.

We’ll stay on the beat to see what emerges. And, as always, we welcome your comments.