Anthropic's AI push raises analyst concerns over Indian IT services revenues

Indian IT services firms have built a huge business on a simple promise: “We’ll supply skilled teams to build, run, and modernise your systems.” Generative AI flips that promise on its head. When a model can draft code, write test cases, summarise tickets, and propose cloud architectures in minutes, the value shifts away from headcount and toward outcomes.

That’s why analyst concerns about AI—especially the fast enterprise push from players like Anthropic—aren’t just a story about India’s IT giants. It’s a signal for Singapore businesses. If your company buys IT services, runs shared services, or has a backlog of process improvement projects, your cost structure and speed-to-market are about to change.

I’ve found the biggest mistake leaders make here is treating AI as “an IT upgrade.” It’s a procurement, operating-model, and productivity upgrade. The companies that act early will renegotiate vendor relationships, compress cycle times across marketing and operations, and build internal AI muscle while others are still debating policies.

Why Anthropic’s enterprise push worries IT services analysts

Answer first: If enterprises can do more work with fewer people using AI assistants, traditional IT services revenue—often priced by time and materials—comes under pressure.

Anthropic (maker of Claude) is part of a broader wave of enterprise-grade AI providers offering models that are safer, more controllable, and easier to deploy in corporate environments. As these tools get embedded into IDEs, ticketing systems, and knowledge bases, they shrink the “hours” needed for common tasks.

Here’s the business-model tension: many IT services engagements still rely on billing constructs tied to effort—FTE-based managed services, application maintenance contracts measured in hours, and large transformation programmes staffed by big teams. AI doesn’t eliminate the work, but it compresses the effort. If a task that used to take 10 hours now takes 4, someone’s margin gets squeezed unless pricing changes.

The real shift: from labour arbitrage to automation arbitrage

For two decades, global IT services grew through labour arbitrage—moving work to lower-cost talent pools. Generative AI introduces automation arbitrage: the winner is the firm that can convert AI capability into reliable delivery faster than competitors.

That’s why analysts worry about revenue: if services buyers push for productivity pass-through, vendors may deliver the same outcomes with smaller teams and lower billings.

Where AI hits first: the “repeatable middle”

AI tools are already strong at work that is:

• Document-heavy (requirements, SOPs, knowledge articles)
• Pattern-based (CRUD code, refactoring, unit tests)
• Ticket-driven (L1/L2 support, triage, resolution suggestions)
• QA-heavy (test case generation, regression scripts)

These areas are common inside large IT services portfolios. The more standardised the work, the faster AI deflates effort.

What changes for Singapore companies that rely on outsourced IT

Answer first: Singapore firms can buy outcomes cheaper and faster—but only if they change how they scope, govern, and measure vendor work.

Singapore businesses sit in an interesting position. Many run lean teams and outsource chunks of engineering, infrastructure, and business process work to regional partners (including Indian IT services). As those partners adopt AI, you’ll likely see two conflicting behaviours:

1. Some vendors will use AI to protect margins (same price, smaller team, faster delivery).
2. Others will be forced into price competition as buyers demand savings.

Your advantage comes from being deliberate: specify the outcomes you want, measure cycle time and quality, and structure contracts that share productivity gains.

Procurement and governance will matter more than model choice

Model debates (Claude vs GPT vs others) matter, but not as much as governance. The same AI tool can create either a productivity step-change or a compliance headache depending on how it’s deployed.

For Singapore firms, the practical question is:

“Can we let vendors and internal teams use AI on our code and data without creating IP leakage, regulatory exposure, or audit gaps?”

That requires clear rules on:

• Data classification (what can be pasted into AI tools)
• Approved AI environments (enterprise accounts, logging, retention controls)
• Human review requirements (especially for production code)
• Security testing and SBOM practices for AI-generated code

If you don’t define this, your vendors will define it for you—usually in ways that optimise their speed, not your risk posture.

How AI reshapes IT services pricing (and what to negotiate)

Answer first: Expect a move from time-and-materials to outcome-based pricing, with “AI productivity” baked into delivery assumptions.

The most important commercial change is pricing. AI doesn’t just improve productivity; it changes what fair pricing looks like.

Three pricing models you’ll see more of

1. Outcome-based bundles (e.g., “deliver X features per quarter”)
2. Platform + services (vendor provides an AI-enabled delivery platform and charges for access + delivery)
3. Shared savings (baseline cost vs AI-improved cost, savings split)

If you’re in Singapore buying managed services or software development, push for contracts that:

• Define throughput metrics (lead time, deployment frequency, defect escape rate)
• Tie payments to service levels and business outcomes
• Include transparency on AI usage (which tools, what controls, what audit trails)

A stance worth taking: don’t pay for avoidable rework

AI will draft code quickly. The failure mode is sloppy integration, brittle tests, and security gaps that create costly rework. I’m opinionated here: you should refuse to fund avoidable rework caused by AI-generated changes that weren’t reviewed or tested properly.

Put it in the contract:

• PR review coverage targets
• Minimum automated test thresholds
• Security scanning requirements
• Rollback and incident response playbooks

Quality is how you stop AI speed from turning into operational drag.

Practical playbook: using AI business tools in Singapore now

Answer first: Start with high-volume workflows in marketing and operations, then extend to software delivery—while building governance as you scale.

This post is part of our AI Business Tools Singapore series, and here’s the recurring pattern: AI adoption works when you treat it as a workflow redesign, not a tool rollout.

Step 1: pick “volume + friction” workflows

Choose processes that are frequent, measurable, and painful. In Singapore SMEs and mid-market firms, these are common starting points:

• Customer support: auto-drafting replies, summarising cases, routing tickets
• Sales ops: call summaries, CRM updates, proposal drafting
• Marketing ops: campaign briefs, landing page variants, ad copy testing
• Finance ops: invoice classification, policy Q&A, exception handling
• HR ops: onboarding Q&A, policy lookups, document drafting

A realistic target: 20–40% cycle-time reduction in the first 6–10 weeks for text-heavy workflows when you combine AI with better templates and approvals. (The number varies by process maturity, but that range is achievable in many teams.)

Step 2: build a “safe AI” setup before scaling

You want speed and control. A minimal enterprise setup looks like:

• Approved AI accounts (enterprise plans where possible)
• A private knowledge base/RAG layer for internal documents
• Logging of prompts/outputs for audit in sensitive workflows
• A redaction policy for customer data
• Role-based access and prompt guidelines

If you’re in regulated industries (finance, healthcare, telecom), treat this as non-negotiable plumbing.

Step 3: extend into software and IT delivery

Once governance is stable, software delivery is where AI can compound. Practical use cases:

• Accelerated backlog grooming (user story drafts + acceptance criteria)
• Test generation and maintenance
• Refactoring assistance for legacy systems
• Runbook drafting for ops teams
• Faster incident summaries and postmortems

The win isn’t “AI writes code.” The win is shorter lead time from idea to production, with fewer defects.

What happens to Indian IT services—and why Singapore shouldn’t wait

Answer first: Indian IT firms will adapt, but the transition will be messy—creating a buyer’s market for a period. Singapore companies can use that window to modernise faster.

Indian IT services companies aren’t going away. They have distribution, talent, domain knowledge, and enterprise relationships. But revenue pressure can show up when:

• Buyers demand AI-driven rate reductions
• Large transformation programmes shrink in size
• Managed services contracts get repriced on productivity
• In-house teams use AI to “pull work back” internally

That uncertainty creates opportunity for Singapore buyers. When vendors are retooling, they’re also more willing to:

• Pilot new pricing models
• Commit stronger SLAs
• Fund proof-of-value projects
• Offer AI accelerators (templates, agents, delivery platforms)

My advice: don’t wait for vendor maturity to peak. If you move now, you learn faster, negotiate better, and build internal capability while the market is still sorting itself out.

People also ask: “Will AI replace outsourced IT?”

Not fully. AI will replace chunks of tasks, and it will compress team sizes for repeatable work. But large-scale delivery still needs architecture decisions, stakeholder management, domain context, and accountability. The outsourcing relationship changes from “staffing” to “managed outcomes.”

People also ask: “Is Claude/Anthropic specifically the threat?”

Anthropic is a symbol of the trend: enterprise-grade models with strong safety and governance features. The bigger force is competition among model providers plus rapid integration into everyday tools (IDEs, service desks, CRMs). The model you use matters less than how quickly you redesign workflows around it.

What to do next (especially if you’re in Singapore)

The primary keyword here is AI impact on IT services revenue, but the action isn’t academic. It’s operational.

Start with two moves:

1. Audit where you pay for effort, not outcomes. Identify managed services or development scopes that are ripe for AI productivity. Prepare to renegotiate.
2. Run one AI pilot in operations and one in marketing. You’ll learn different lessons—ops teaches governance and reliability; marketing teaches iteration speed and brand controls.

The companies that win in 2026 will treat AI like a new operating layer—something you bake into vendor management, internal workflows, and KPI design. If you’re planning budgets right now, the most expensive choice is pretending this will wait until next year.

What would change in your business if your team shipped the same work with 30% fewer handoffs—and your vendors priced for results instead of headcount?