IGI Logo

The Nifty IT index was down ~15% over the past 1M vs the broader market index. Nifty was flat on fears of advancement in AI disrupting IT/BPM services business. It is difficult to quantify the real impact of AI on IT Services business, as the timing of potential headwinds from AI-led productivity gains and tailwinds from modernization and new AI-led spending remain uncertain at this stage of the cycle. Also, it is overly simplistic to assume that AI can generate enterprise-grade code and replace IT Services companies. We expect IT services companies’ role to evolve and business model to shift toward outcomebased compared to predominantly input-based/effort-based today. We believe the market reaction is excessive, aligning well with Amara's law, “We tend to overestimate the effects of a technology in the short run and underestimate them in the long run.” Enterprise systems are complex and adoption is expected to remain gradual. IT Services companies have the advantage of contextual understanding of enterprises’ complex environment, domain knowledge, and clients trust; hence, they would remain relevant even in the AI era, in our view. Post correction, large-cap IT services companies are trading at FCF yield of ~5- 6%, and valuations imply ~5-6% terminal growth (Exhibit 1). We believe the correction largely reflects fear of the unknown. With the impacts hard to quantify, investors have trimmed terminal growth assumptions. Management commentary providing clarity on human + agent offerings and flow through the services component on spend on external models, business model transition toward outcome-based, and deal intake would be keenly watched, in our view, which provide confidence on earnings trajectory. A durable recovery, however, would take some time and depend on consistent earnings delivery.

Claude turning from helper (assistance) to doer (autonomy The latest leg of risk-off in software/SaaS/services has been catalyzed by high-velocity agent narratives: Anthropic’s Cowork has been positioned as a more general-purpose agentic workflow layer (beyond coding), and the plug-in concept makes it feel closer to a reusable enterprise worker than a chatbot, reinforcing the fear that routine knowledgework gets automated end-to-end. Cowork plug-ins could possibly standardize how work is done by connecting tools/data and enforcing consistent workflows, and Anthropic opensourced a starter set (with 11 plug-ins), which amplified concerns around faster diffusion.

amplified by a compressed timeline to implement ERP In parallel, Palantir’s commentary that its AI-forward deployment construct can compress complex SAP ECC to S/4 migrations from ‘years to as little as two weeks’ has further hardened the market’s belief that large implementation pools will structurally shrink, especially for vendors whose economics are effort-and-labor heavy.

driving fears of terminal growth and resulting in a sharp sell-off Superimposed, these developments were quickly mapped onto Indian IT Services’ perceived labor-dependent model and onto SaaS monetization risk. Indian IT Services firms have traditionally scaled revenue with headcount – more projects typically meant more people billed over longer timelines. If AI agents can execute meaningful portions of work autonomously, that link between effort, time, and revenue starts to weaken. The concern is not that demand disappears, but that the revenue per unit of work compresses as automation rises. With the rapid pace of developments by Anthropic, OpenAI, etc, this shift is unfolding faster than expected. Hence, the investors and the market at large have started questioning the relevance of these companies in future. Many SaaS businesses rely on seat-based pricing, where revenue grows with the number of human users. If AI agents become the primary actors interacting with software – running workflows, generating outputs, and making decisions – then the number of seats may no longer reflect actual value delivered. This raises the possibility that pricing models shift from per-user to per-usage, per-outcome, or platform fees, which could reset growth assumptions.

But what is being missed or incorrectly assumed? Amara’s Law offers a fitting perspective here: we tend to overestimate the short-term effects of technology while underestimating its long-term impact. Yet current market positioning appears to assume a largely linear, frictionless disruption curve. Implicitly, it underwrites several assumptions: i) Agent capabilities seen in demos will translate quickly into productiongrade autonomy, with limited friction from complex enterprise systems integration, controls, and exception handling. ii) Most routine delivery work will be executed by agents rather than by scaled delivery teams, meaning the scope of implementation shrinks rather than shifts. iii) Productivity gains will be passed through almost entirely to clients, compressing realization under T&M and forcing a rapid reset in pricing (even on complex programs). iv) The companies in the Indian IT Services space will struggle to pivot fast enough toward outcomebased constructs, integrating AI at a broader scale.

The opportunity pool remains large, with net-new work aided by AI A coherent framing is that AI creates deflationary pressure in certain legacy revenue pools (routine maintenance, repetitive enhancements, standardized testing/documentation, etc), while simultaneously enabling net-new programs that were previously uneconomical or too risky. Before any AI works, companies need their data in the cloud, their apps modernized, and their systems integrated. Hence, the opportunity set includes legacy code modernization (including large estates, such as COBOL), building customizable to composable solutions, etc. Even where AI shortens parts of delivery, the end-to-end enterprise journey still requires redesigning data models, controls, workflows, and integrations; the services burden often shifts rather than disappears.

Our view is anchored on the following points

* Complexity of enterprise workflows remains the binding constraint Enterprise delivery is constrained less by code production and more by end-to-end workflow complexity across apps, data, controls, security, and stakeholder approvals. Even when AI accelerates discrete tasks (drafting code, generating tests, summarizing requirements, etc), IT services firms bring the context – understanding legacy tech stacks, business processes, and regulatory environments – that allows solutions to actually work in production.

* Enterprise readiness is lagging the pace of foundational AI advancement AI is progressing on a curve that looks exponential, while enterprises change on curves that are organizational, budgetary, and human. This mismatch creates a persistent deployment gap: models get better, cheaper, and more capable every passing day, yet many companies still struggle to move beyond pilots. While AI adoption among users is the fastest of any technology, enterprise value realization lags due to legacy systems, data silos, governance needs, change-management complexity, and connecting AI to such unstructured data, and legacy workflows. As a result, the realized value compounds more slowly, increasing scope for SIs to act as a bridge in making enterprises AI ready (Exhibit 3).

* SLMs (Small language models) are strategically important for enterprises Enterprises increasingly evaluate AI through the lenses of cost, latency, deployment control, data residency, and domain fit, which often favor smaller or specialized models over the largest frontier models. The trend toward SLM adoption plays directly to the strengths of IT Services providers who possess deep domain expertise, access to proprietary industry data through decades of client engagements, and established relationships with enterprise clients who require customization rather than off-the-shelf solutions. The value chain will be distributed across companies that can integrate, customize, deploy, and maintain AI systems within complex enterprise environments.

* Legacy modernization and cloud migration remain a large and resilient spend category

Legacy modernization continues to represent a substantial and growing portion of the IT services opportunity landscape, providing a significant buffer against near-term AI displacement fears. An estimated ~800bn lines of COBOL code remain active in daily use on production systems, powering critical financial infra, as per MicroFocus and Vanson Bourne.

The opportunity of IT Services players in this context is expanding rather than shrinking, with AI acting as a key enabler. Most large enterprises still run critical business processes on mainframe systems, decades-old ERP implementations, custom COBOL applications, and monolithic architectures that were never built for cloud or AI-first environments. Modernizing these systems involves enormous technical debt and operational risk, which makes enterprises cautious and drives demand toward experienced service providers. AI helps in reducing costs and failure rates, and addresses the talent shortage issue in modernizing legacy code. While AI is accelerating parts of the modernization process (such as code analysis, translation, and testing), it is also increasing the overall scope of transformation by enabling more cloud, data, and AI-led re-architecting programs (Exhibit 9).

A meaningful portion of enterprise workloads remains on-prem, and shifting these workloads to cloud would be imperative in the AI-first landscape. In this context, Indian IT Services players are well-positioned to benefit from scale, cost efficiency, longstanding client relationships, and growing investments in cloud and AI capabilities, positioning them as preferred partners for large, multi-year modernization and transformation programs.

However, our view carries some downside and hence needs to be assessed in conjunction with the following

* Reinvestment shortfall, if savings are harvested and not redeployed The constructive case assumes a decent portion of productivity gains would be reinvested into more modernization, data, cloud hardening, and automation initiatives. The downside is a procurement-led outcome, where clients bank the savings to offset macro pressure or fund non-services spend (spend with model providers, infrastructure compute platforms/cloud commitments, security products, internal hires), limiting the volume offset that IT Services vendors need to keep toplines stable.

* Timing mismatch between effort deflation and new growth opportunities and scope expansion If AI materially improves code understanding, test generation, refactoring, and incident triage in production settings, a larger share of application work can be delivered with fewer billable hours and smaller teams, especially in run/change work. This would risk the monetization per unit of work getting compressed, as clients see tangible productivity and push for immediate pass-through at renewals, making revenue deflation visible before new AI-era workstreams scale.

For More  Emkay Global Financial Services Ltd Disclaimer http://www.emkayglobal.com/Uploads/disclaimer.pdf & SEBI Registration number is INH000000354