Infosys moved to steady nerves in a jittery IT market by striking a partnership with Anthropic to build enterprise-grade AI agents, integrating Claude models into its Topaz platform. The aim is to deliver autonomous, policy-aware systems that can execute complex workflows in sectors like banking, telecom, and manufacturing—where reliability, auditability, and compliance matter as much as raw model performance.
The announcement lands as investors debate how fast AI will reshape India’s $280 billion IT services engine. Shares of several Indian IT firms have swung on concerns that AI assistants could compress labor-intensive revenue pools even as they unlock new, higher-margin offerings. Infosys is signaling it intends to be on the front foot of that shift, not at the mercy of it.
Why This Partnership Matters For IT Services
Infosys disclosed that AI-related services generated ₹25 billion (about $275 million), or 5.5% of its ₹454.8 billion quarterly revenue. Rival Tata Consultancy Services has said AI accounts for roughly $1.8 billion annually, or around 6% of revenue. These figures show AI is no longer a side bet; it is already a measurable contributor—and an area clients are budgeting for even as they trim run-the-business spend.
At the same time, enterprise buyers want more than chat interfaces. They’re asking for agents that can reason over proprietary data, call tools, follow rules, and leave an audit trail. That is the crux of “enterprise-grade” in 2026: governed autonomy that plugs into existing systems and satisfies risk officers as well as developers.
What Infosys and Anthropic Are Building Together
Infosys plans to weave Anthropic’s Claude family across Topaz to orchestrate “agentic” patterns—task decomposition, planning, tool use, and multi-step execution. The stack will emphasize enterprise controls: role-based access, retrieval-augmented generation tied to approved knowledge bases, data loss prevention, human-in-the-loop checkpoints, and full observability for post-mortems and compliance reviews.
On the developer side, Infosys is rolling out Claude Code internally to write, test, and debug software, building hands-on expertise that can be ported to client work. That move tracks with broader industry evidence: a GitHub study found AI coding assistants helped developers complete tasks up to 55% faster in controlled trials, and large integrators increasingly anchor productivity cases on similar gains.
Anthropic, for its part, gains a scaled channel into regulated sectors where production-grade deployments hinge on integration know-how and governance. As Anthropic’s leadership has noted, the leap from an impressive demo to a safe, standards-compliant rollout in finance or telecom requires deep domain patterns and operational guardrails—areas where global IT services firms are indispensable.
Early Use Cases and Return on Investment Signals
Expect initial agents to land in high-volume, rules-heavy processes. In banking, that means automating portions of KYC refresh, adverse media checks, and credit memo drafting with human approval gates. In telecom, agents can triage service tickets, propose resolutions, and trigger workflows across OSS/BSS systems. In manufacturing, they can summarize quality anomalies from sensor data, draft corrective actions, and open work orders—while preserving traceability.
Clients will judge success on a tight set of metrics: time-to-resolution, first-contact containment, defect escape rates, and cost-to-serve. The design challenge is to blend autonomy with oversight—letting agents run routine steps end to end while escalating edge cases to human experts. That balance is where most ROI is realized without running afoul of risk thresholds.
India is also becoming a meaningful user base for Anthropic’s tools. The company has noted that India accounts for about 6% of global Claude usage, second only to the U.S., with a heavy skew toward programming tasks. That developer gravity can accelerate agent adoption inside delivery centers and client digital factories alike.
Market Turbulence And Competitive Moves
Volatility in IT stocks reflects a dual narrative: AI could compress legacy headcount-based revenues, yet it also opens premium consulting, data engineering, and platform operations lines. Infosys’s tie-up follows a pattern of alliances designed to capture the upside—HCLTech’s collaboration with OpenAI is one example—while cushioning against pure-play AI vendors disintermediating services layers.
Global peers are arming up too. Accenture has publicly committed billions to Data & AI, and Wipro has earmarked substantial investments over multiple years. The competitive edge will hinge less on model access—now broadly available—and more on reference architectures, safety tooling, integration accelerators, and outcome guarantees embedded in contracts.
Key Risks to Watch and the Road Ahead for AI Agents
Enterprise AI agents still face known risks: hallucinations, tool misuse, privacy leaks, and unpredictable failure modes under distribution shift. Mitigations include stronger retrieval pipelines, red-teaming, model ensembles, and continuous evaluation against domain-specific test suites. Emerging eval frameworks from academia and industry are helping, but procurement teams will demand transparent benchmarks tied to their data.
Cost governance is another fulcrum. Agent architectures can spawn runaway token and API calls without careful budgeting, caching, and policy constraints. Expect Infosys to bundle FinOps-style controls and multi-model routing to balance accuracy, latency, and spend.
Infosys has not disclosed timelines or financial terms for Claude-powered agents, but the signals are clear: clients want governed autonomy, and vendors that combine cutting-edge models with enterprise muscle will have the inside track. For investors, the question is execution—can services firms convert AI anxiety into durable, higher-margin revenue before pure software players swallow the prize?
