AI Automation Is Shrinking IT Projects—Here’s the Play

Indian IT stocks don’t drop 6% in a day because of a minor product update. They drop when the market senses a structural shift.

That’s what last week’s Reuters coverage (via CNA) captured: rapid AI automation progress—sparked by bold claims from firms like Anthropic and Palantir—is spooking analysts who think traditional, labour-heavy application services could see faster delivery timelines, pricing pressure, and ultimately lower revenue.

For the AI Business Tools Singapore series, this isn’t a “foreign markets” story. It’s a warning flare for every Singapore business that relies on outsourced development, long transformation projects, or manual back-office work. The reality? The value is moving from “hours billed” to “outcomes shipped.” If you’re running a company here, you can either pay for that shift—or benefit from it.

Why AI is a direct threat to “application services” revenue

Answer first: AI threatens application services revenue because it compresses the time and people needed to build, test, maintain, and support software.

In the CNA piece, analysts point out that application services account for roughly 40%–70% of revenues for many IT services firms. That’s the high-margin work: enhancements, integrations, ongoing maintenance, and the endless backlog of “small” changes that keep enterprise systems alive.

AI attacks this revenue pool in three concrete ways:

1. Timeline compression

   • If an AI coding assistant and automated test generation reduces a 12-week delivery cycle to 6–8 weeks, the same scope yields fewer billable weeks.

2. Team size compression

   • Automation doesn’t eliminate all roles, but it can reduce the number of developers, QA testers, and support analysts needed per project.

3. Deflation in legacy work

   • Jefferies’ view in the article is blunt: deflation in older service lines could outweigh new AI-related opportunities in the near term.

Motilal Oswal’s estimate—9%–12% of industry revenues eliminated over four years—is the kind of number that makes CFOs and boards stop treating AI like a “pilot.”

The bigger point for Singapore businesses

If AI reduces effort, your vendors will either:

• charge less for the same work,
• push for outcome-based pricing,
• or bundle AI “platform fees” while shrinking headcount.

Either way, procurement assumptions based on day-rates and headcount are going stale.

The myth: “AI will replace outsourcing” (it won’t)

Answer first: AI won’t remove the need for external IT partners; it will change what you pay them for.

One of the most useful lines in the Reuters/CNA reporting comes from the pushback: JPMorgan argues it’s illogical to assume enterprises will replace every layer of mission-critical software overnight. Kotak called the market reaction “plenty of panic over a little flutter.”

I agree with the spirit of that pushback. Mission-critical systems don’t disappear because a new model can write cleaner code. Real systems have:

• messy data,
• regulatory constraints,
• security requirements,
• legacy dependencies,
• and users who resist change.

So outsourcing doesn’t die. But commodity delivery gets priced like a commodity.

Here’s the stance I’d take if I were advising a Singapore SME or mid-market firm:

AI won’t eliminate services. It will eliminate uninstrumented services—work that isn’t measured, packaged, or tied to outcomes.

If your vendor’s value proposition is “we’ll staff 12 people,” AI makes that pitch weaker. If their pitch is “we’ll reduce incident volume by 30% and cut release cycle time from monthly to weekly,” AI makes them stronger.

What Singapore companies should do before your next IT renewal

Answer first: Update how you scope, contract, and measure work—because AI will change delivery economics faster than your annual budgeting cycle.

Singapore is positioning itself as an AI hub, and many teams already use AI for marketing, operations, and customer engagement. But plenty of organisations still buy IT services like it’s 2018. That gap is where costs creep in.

1) Rewrite your requirements in “outcome language”

Instead of writing requirements that describe labour, write requirements that describe results.

Old (labour-based): “Provide 2 developers and 1 QA for 6 months to enhance the customer portal.”

Better (outcome-based):

• “Reduce checkout drop-off by 15% through UI and performance improvements.”
• “Cut page load time to under 2 seconds for 95% of sessions.”
• “Achieve weekly releases with automated regression tests covering top 30 user journeys.”

Outcome language forces the vendor to apply automation (including AI) to hit targets efficiently.

2) Put “cycle time” into your contract

If AI shortens delivery cycles, you should explicitly benefit.

Add measurable metrics such as:

• lead time from request to production,
• deployment frequency,
• change failure rate,
• mean time to recover.

These are standard DevOps metrics, but most non-tech businesses still don’t use them in commercial terms.

3) Separate “AI tool cost” from “services cost”

Vendors increasingly bundle AI assistants, agent platforms, and code-generation tools into their fees.

Ask for a split:

• Tooling (licenses, model usage, platform fees)
• Services (implementation, testing, change management)

This makes it harder for “AI” to become a vague surcharge.

4) Build a small internal AI capability (even if you outsource)

You don’t need a huge AI team. You do need enough internal competence to evaluate claims.

In practice, that means:

• a product owner who understands how AI changes scope,
• an operations lead who can map processes for automation,
• and someone who can do basic vendor due diligence on data handling.

In Singapore, this is often the difference between “AI got approved” and “AI delivered savings.”

The new outsourcing model: smaller teams, faster sprints, tighter governance

Answer first: The winning delivery model combines AI-assisted build with strong governance, because speed without controls creates risk.

The CNA article frames the threat to labour-intensive models. That’s accurate. But it’s not the full picture. AI speeds up output; it also increases the chance of:

• inconsistent code patterns,
• security oversights,
• brittle integrations,
• and “hallucinated” assumptions in documentation.

So the future model is not “let AI run wild.” It’s “use AI to move faster, and tighten the guardrails.”

Practical model I’ve found works

If you’re a Singapore business buying services, push partners toward a model like this:

1. AI-assisted delivery

   • coding assistants, automated test generation, scripted environment setup.

2. Human review on high-risk layers

   • security review, architecture decisions, data access patterns.

3. Continuous validation

   • monitoring, automated regression, and clear rollback paths.

4. Documentation as a deliverable

   • system diagrams, runbooks, data dictionaries. If it’s not documented, it’s not done.

The point is to get the efficiency upside of AI without paying for avoidable incidents later.

Where AI business tools help right now (beyond IT)

Answer first: Singapore businesses can use AI tools today to reduce manual work in ops, marketing, finance, and customer support—without waiting for a giant IT transformation.

This matters because many firms hear “AI disrupts IT services” and think it’s irrelevant to them. It’s the opposite: if AI compresses software work, it also compresses business process work.

Here are practical, low-drama plays that fit the AI Business Tools Singapore theme:

Marketing & customer engagement

• Draft first-pass campaign copy and variants; keep final approval human.
• Summarise call transcripts and tag objections automatically.
• Generate personalised follow-up emails based on CRM notes.

Operations

• Turn SOPs into searchable knowledge bases.
• Auto-generate weekly ops reports from spreadsheets and dashboards.
• Flag anomalies in order processing (missing fields, unusual delays).

Finance

• Classify expenses and extract invoice fields.
• Draft variance explanations for month-end packs.
• Summarise procurement contracts for renewal risks.

Customer support

• Suggest replies, cite internal policy, and hand off to agents.
• Auto-triage tickets by intent and urgency.
• Build “known issue” summaries from repeated tickets.

Notice what’s consistent: you’re not trying to “replace the department.” You’re removing the repetitive layers that slow everything down.

A simple checklist before you buy (or renew) outsourced IT services

Answer first: Use a checklist that tests whether your partner is selling outcomes, not headcount—and whether their AI claims translate to your risk and governance needs.

Use this as a pre-renewal checklist:

1. What % of the work is application maintenance vs new build?

   • (The CNA piece highlights application services exposure as a key risk.)

2. How will AI change the delivery plan and price?

   • Ask for side-by-side: “with AI” vs “without AI.”

3. What’s the testing strategy if AI generates code faster?

   • Speed without automated testing is a hidden cost.

4. Who owns the prompts, scripts, and automation assets?

   • Treat these like IP. Clarify rights.

5. Where does your data go?

   • Get explicit answers on data residency, retention, and access.

6. What are your measurable outcomes?

   • Tie payment milestones to outcomes where feasible.

If a vendor can’t answer these cleanly, you’re buying uncertainty.

What to watch next (and how to stay ahead in Singapore)

AI automation is already impacting market expectations—CNA notes that foreign investors sold a record US$8.5 billion of Indian IT stocks in 2025, and the sector has faced persistent pressure.

But for operators (not traders), the lesson is simpler: assume AI will reduce the cost and time of many software and business workflows, then plan your contracts and teams accordingly.

If you’re building your 2026 execution plan now, pick one area—support, finance ops, sales follow-ups, analytics reporting—and implement a practical AI workflow that saves real hours. Then apply what you learned to bigger vendor discussions.

The question worth sitting with: when AI makes output cheaper, will your business keep paying for effort—or start buying outcomes?

Source: https://www.channelnewsasia.com/business/anthropics-ai-push-raises-analyst-concerns-over-indian-it-services-revenues-5909371