Infosys + Anthropic: Agentic AI moves from demos to production
Infosys and Anthropic are integrating the Claude family of models - including Claude Code - with the Infosys Topaz platform to build agentic AI for telecoms, financial services, manufacturing, and software development. The rollout starts in telecoms with a dedicated Anthropic Centre of Excellence, then expands to finance and manufacturing.
The goal is simple: deploy agents that handle multi-step work reliably under regulation, not just one-off prompts in a sandbox.
Why this matters for regulated industries
There's a gap between impressive demos and production-grade systems that stand up to audits, SLAs, and policy. As Anthropic CEO Dario Amodei put it: "There's a big gap between an AI model that works in a demo and one that works in a regulated industry - and if you want to close that gap, you need domain expertise."
Infosys brings sector depth and delivery muscle; Anthropic brings model capability and an agent SDK built for persistent, long-running tasks. That combination is the headline.
What's actually being built
Using Anthropic's Claude models, Claude Code, and the Claude Agent SDK alongside Infosys Topaz, teams will create agents that run end-to-end processes: coordinate tools and data sources, ask for approvals when needed, and keep state over long workflows. This covers more than code assistance - think claims handling, compliance review, and legacy modernization at scale.
Infosys CEO Salil Parekh frames the intent clearly: use AI to modernize financial services with risk and compliance workflows, enable AI-led design and manufacturing, and compress time-to-value for enterprise programs.
Sector-by-sector: priority use cases
Financial services- Risk detection across transactions and portfolios with clear audit trails and rationale.
- Automated compliance reporting: policy mapping, evidence gathering, and draft narratives for review.
- Personalized client interactions that factor full account history and live market context.
- Network operations: fault triage, ticket routing, change plans, and post-incident reviews.
- Customer lifecycle: next-best action in care and retention with policy-safe guardrails.
- Service delivery: order orchestration, provisioning checks, and proactive remediation.
- Product design and simulation to shorten R&D cycles and explore more variants.
- Quality documentation, supplier risk analysis, and standards compliance workflows.
- Production support: work instructions, root-cause narratives, and maintenance planning.
- Claude Code for high-quality generation, tests, debugging, and refactoring.
- Agent-driven regression runs, integration checks, and release note drafting.
- Legacy modernization: code translation, interface mapping, and data migration accelerators.
What finance and product leaders should do now
- Pick 1-2 high-friction workflows with measurable KPIs (cycle time, error rate, cost per case). Baseline them.
- Lock down data governance: PII handling, retention, lineage, and least-privilege access.
- Define agent guardrails: scope of actions, approval thresholds, and human-in-the-loop checkpoints.
- Run a constrained pilot: limited surface area, staged deployment, A/B against current process.
- Instrument evaluation: accuracy, intervention rate, drift, latency, and cost per completed task.
- Plan change management: role impacts, RACI updates, training, and incentives.
- Vendor diligence: update cadence, incident playbooks, audit support, and uptime SLAs.
Risk controls that actually reduce exposure
- Independent validation datasets and red-teaming before production access.
- Immutable logs, signed outputs, policy-based access, and full traceability for audits.
- Continuous monitoring with rollback paths, canary agents, and kill switches.
Rollout and what to expect
Telecoms leads via the new Centre of Excellence, followed by finance and manufacturing. Infosys is already using Claude Code inside its Exponential Engineering group, building internal expertise ahead of client delivery. Early client wins will likely land in back-office and legacy-heavy areas before expanding to customer-facing flows.
Resources
Want to build internal capability around Claude and agent workflows? Start here: Claude. If you lead in financial services, this collection is useful: AI for Finance.
Your membership also unlocks:
