AI Automation Is Shrinking IT Work—What SG Firms Do

A 6% one-day drop in Indian IT stocks is a loud signal—not because markets are always right, but because they’re reacting to a real structural shift. According to reporting on analyst commentary, rapid AI-driven automation (sparked in part by Anthropic’s latest push and similar claims from Palantir) is raising fears that high-margin application services—often 40%–70% of revenue for major IT services firms—could be delivered faster, with fewer billable hours, and at lower prices. That’s the core concern: when software work becomes “compressible”, the traditional labour-heavy services model gets squeezed.

If you run a business in Singapore, this isn’t just “an India IT story”. It’s a preview of how your vendors will price, how quickly your internal teams can ship changes, and how customer expectations will reset. In the AI Business Tools Singapore series, I’ve found the most useful framing is this: AI doesn’t simply replace work; it changes the unit of work, the cycle time, and the economics.

Here’s what’s happening, what’s likely to happen next, and a practical playbook for Singapore companies adopting AI business tools for operations, marketing, and customer engagement—without creating a governance mess.

Why analysts are worried: application work is becoming “deflatable”

Answer first: Analysts are worried because AI automation reduces the time and people required for common application services tasks, pressuring margins and shortening project timelines.

The Reuters/CNA piece captures a common market fear: if AI assistants can generate code, tests, documentation, and even migration scripts, then a lot of what used to be multi-month delivery turns into multi-week (or multi-day) work. Jefferies’ point is blunt: application services are a huge revenue pool, and if that pool shrinks, earnings and valuations follow.

A few specifics from the article worth anchoring on:

• Application services make up ~40%–70% of revenue for many large IT firms; some cited examples sit around 55%–60% exposure.
• Foreign investors reportedly sold US$8.5B of Indian IT stocks in 2025.
• One estimate (Motilal Oswal) suggests 9%–12% of industry revenue could be eliminated over four years due to AI-led disruption.

Whether the exact percentage is right is less important than the direction. If your company buys software services—or builds internal systems—AI is turning time into the new battleground.

The mechanics: why “time saved” becomes “revenue lost”

A lot of enterprise application work is priced implicitly on effort:

• tickets closed
• sprints delivered
• analysts + developers + QA hours
• change requests and integration work

When AI reduces effort, buyers don’t just say thanks. They renegotiate.

This is why the market reaction isn’t totally irrational. If vendors can deliver faster with fewer people, pricing models (and procurement expectations) adjust. Over time, “effort-based” contracts get replaced by outcome-based or capacity-based models, where the buyer captures some of the productivity gains.

The Singapore angle: your cost base may fall—but your bar will rise

Answer first: For Singapore firms, AI automation should reduce delivery costs and cycle times, but it will also raise expectations for speed, personalisation, and service quality.

Singapore companies sit in a unique position:

• You’re often regional HQs coordinating multi-country operations.
• You rely heavily on outsourced IT partners (many from India) for ERP, app maintenance, integration, and support.
• You also face intense pressure to improve customer experience quickly—especially in finance, retail, logistics, and healthcare.

So as AI compresses software work, the biggest practical changes for Singapore businesses are:

1. Vendor negotiations will change. Your partners will pitch “AI-enabled delivery”, and you’ll need a firm stance on pricing and governance.
2. Internal expectations will change. Business leaders will ask why a dashboard takes 6 weeks when “AI can do it in 2 days”.
3. Customer expectations will change. Faster iteration means competitors can ship new experiences more frequently.

There’s also a second-order effect: if IT service providers face margin pressure, they’ll aggressively pursue new revenue streams—AI implementation, data engineering, AI governance, model risk, and change management. That work is harder to automate and closer to strategic value.

What this means for your AI business tools strategy (and what to buy first)

Answer first: Treat AI as an operating model upgrade, not a single tool purchase—start with high-volume workflows where accuracy can be measured.

Most companies get this wrong by starting with a generic chatbot. The better approach is to map workflows where:

• volume is high (lots of repeated work)
• time-to-value is short
• risk is manageable
• quality can be measured

A prioritisation map that actually works

If you’re building an AI adoption roadmap in Singapore, I’d prioritise in this order:

1. Customer support & service ops

   • AI-assisted replies with human approval
   • auto-summarisation of tickets and calls
   • knowledge base drafting + freshness alerts

2. Sales & marketing operations

   • account research summaries
   • proposal drafting with controlled templates
   • campaign variant generation + compliance checks

3. Finance & admin workflows

   • invoice triage and exception handling
   • policy Q&A with citations to internal documents
   • month-end narrative generation (with source links)

1. Software delivery acceleration (internal and vendor)
   • AI-assisted testing and regression packs
   • code review support (standards + security)
   • migration planning and documentation

Why put software delivery at #4? Because it’s high-impact but also high-risk if you don’t have guardrails. If you start there without discipline, you get inconsistent code quality, security holes, and a mess of untracked changes.

“AI-first delivery” doesn’t mean “no humans”

The article quotes JPMorgan’s view that it’s illogical to assume companies will replace every layer of mission-critical software overnight. That’s correct—and it’s the mindset Singapore firms should adopt.

A sensible target for 2026 is AI-assisted delivery:

• humans make the final decisions
• AI drafts, checks, and accelerates
• telemetry and audit trails are non-negotiable

This is where AI business tools become practical: copilots, testing automation, retrieval-augmented generation (RAG) on internal docs, and workflow automation.

How to renegotiate IT services in an AI era (a Singapore buyer’s checklist)

Answer first: Shift contracts from effort-based billing to measurable outcomes, and require transparency on AI usage, quality gates, and audit logs.

If your outsourcing partners are adopting Anthropic-style automation internally, you should treat it like any other delivery transformation: beneficial, but only if you control for quality and risk.

Here’s a checklist I recommend for Singapore procurement and tech leaders.

1) Move from “hours” to “outcomes”

Replace open-ended effort metrics with outcome metrics:

• release frequency (e.g., weekly vs monthly)
• mean time to restore (MTTR)
• defect leakage rate
• test coverage thresholds
• performance SLOs

A single sentence you can use internally:

If AI reduces effort, we shouldn’t pay the old effort price—we should pay for outcomes at a new speed.

2) Demand an AI delivery disclosure

Require the vendor to document:

• where AI is used (coding, testing, documentation, support)
• what data is exposed to the tool
• whether prompts or artifacts are retained
• how IP and confidentiality are protected

This isn’t paranoia. It’s basic governance.

3) Put quality gates in writing

AI can produce correct-looking nonsense. Your contract should specify:

• mandatory security scanning
• code review rules
• testing standards (unit + integration + regression)
• acceptance criteria for documentation

4) Protect your architecture from “AI-speed sprawl”

Faster changes can create more chaos. To prevent that:

• enforce API standards
• centralise reusable components
• maintain a living architecture map
• implement change control that’s lightweight but real

“People also ask” (and the answers you can reuse internally)

Will AI reduce the need for IT vendors?

Yes for routine work, no for complex accountability. Maintenance tickets, report changes, and basic integrations will shrink. But governance, architecture, and risk ownership still need accountable teams.

Should we pause digital transformation until the AI dust settles?

No. Pausing is usually the most expensive option. The winning move is to redesign delivery so you can adopt AI safely: clear standards, measurable outcomes, and strong data controls.

What AI capabilities matter most for Singapore SMEs?

Focus on AI for operations and customer engagement first:

• ticket triage and response drafting
• quoting and proposal workflows
• SOP and policy assistants grounded in your documents
• simple analytics narratives for management reporting

A practical 30-60-90 day plan for Singapore businesses

Answer first: Start with one workflow, one dataset, one success metric—then scale only after you can measure quality and risk.

First 30 days: pick a narrow win

• Choose one workflow (e.g., support ticket summarisation)
• Define a metric (e.g., 25% reduction in handling time)
• Decide on guardrails (human approval, redaction, access control)

Next 60 days: operationalise and integrate

• Connect AI outputs into your existing tools (CRM/helpdesk)
• Add feedback loops (thumbs up/down, correction capture)
• Create a “do not use” policy for sensitive data classes

Next 90 days: scale with governance

• Expand to adjacent workflows
• Create an AI playbook for vendors and internal teams
• Establish logging, evaluation, and periodic review

If you do this well, you’ll be in a position to benefit from the same AI-driven productivity gains that are pressuring IT service providers—while avoiding the common failure mode: lots of pilots, no operational impact.

What to watch through 2026: the second-order effects

Answer first: Expect faster delivery cycles, more outcome-based pricing, and a shift in vendor value from “builders” to “operators and governors.”

Three predictions I’m comfortable making for the Singapore market:

1. Outcome-based contracting becomes mainstream for application work that’s easy to benchmark.
2. AI governance and data readiness become board-level topics as regulators and customers expect stronger controls.
3. Competitive advantage shifts to iteration speed—who can ship improvements weekly without breaking production.

The market panic described in the source article may be “overdone” in the short run, but the underlying direction is clear: AI automation is compressing routine work. Singapore firms can treat that as a threat—or as an opportunity to redesign how work gets done.

If you’re planning your next phase of AI adoption, the most useful question isn’t “Which model is winning?” It’s this: Which workflows in our business are still priced like time—and how quickly can we reprice them as outcomes?