AI Plug-ins Are Squeezing IT Services—SG Playbook

A 6% one-day drop in India’s big IT services stocks isn’t “market noise.” It’s a signal that the staffing-heavy way of delivering tech work—lots of people billing lots of hours—is starting to look fragile when AI agents can do chunks of that work on demand.

That’s what made last week’s news stand out: Anthropic released plug-ins for its Claude “Cowork” agent that automate tasks across legal, sales, marketing, and data analysis. Reuters reported the announcement helped trigger a selloff that pushed India’s IT sub-index toward its worst day since March 2020, with names like Infosys, TCS, Wipro, and HCLTech all falling sharply.

This matters for Singapore because our businesses buy the same kinds of services—software development, analytics, marketing ops, customer support—and we run into the same constraint: there’s never enough time or headcount. The opportunity isn’t “replace everyone with AI.” The opportunity is to redesign workflows so output scales faster than payroll.

What happened in India—and why the market reacted

The direct answer: investors fear that agent plug-ins reduce the need for large vendor teams by automating routine, billable work.

Indian IT firms are global powerhouses, but the core model has long been labour-intensive. Clients pay for teams to build, test, maintain, and support systems. If AI plug-ins and agents absorb more of that “routine throughput,” fewer people-hours get billed—and margins compress.

Reuters quoted Systematix Group analyst Ambrish Shah warning that as enterprises integrate Claude into critical coding workflows, dependency on large vendor teams may fall, “squeezing billable hours and margins.” He also pointed out the entry-level talent pool is exposed because routine development and testing are exactly where automation hits first.

Here’s the important framing for Singapore leaders: AI is not just a productivity tool; it’s a pricing-model disruptor. When output can be generated by an agent at near-zero marginal cost, buyers stop accepting “more hours = more value.”

The real disruption: from staffing to systems

If you’re still measuring delivery capacity by headcount, you’re already behind.

AI plug-ins represent a shift from:

• People as the unit of work (assign more staff, bill more hours)
• to systems as the unit of work (design workflows once, run repeatedly)

That’s why stock markets react so quickly. They’re not rating the quality of the plug-ins. They’re repricing the future of a business model.

What Singapore businesses can learn (even if you’re not in IT)

The direct answer: the same AI forces that pressure Indian IT exporters will pressure any Singapore business that relies on manual workflows, repetitive analysis, and “human middleware.”

In this AI Business Tools Singapore series, I keep coming back to one point: most companies don’t have an “AI problem,” they have a workflow design problem. AI just makes that obvious.

Here’s how the India story maps to common Singapore scenarios:

• Marketing teams that spend days compiling performance reports, writing variants, and updating landing pages
• Sales teams that lose hours to research, proposals, CRM hygiene, and follow-ups
• Ops and finance teams that reconcile data across spreadsheets and systems
• Customer service teams that answer the same intent clusters repeatedly

If agent plug-ins can connect AI to the tools you already use (email, docs, CRM, analytics, ticketing), then AI isn’t a sidekick. It becomes part of your operating model.

Myth-bust: “We’ll adopt AI later when it’s stable”

Waiting sounds safe. It isn’t.

The stable part is already here: using AI to automate bounded, repeatable tasks with clear inputs/outputs. You don’t need perfect AI. You need a process where:

1. The work is well-defined
2. The result can be checked quickly
3. Errors are contained

That’s why plug-ins matter: they narrow the gap between “AI can generate text” and “AI can complete a workflow.”

A practical playbook: move from billable hours to measurable outcomes

The direct answer: redesign work around outcomes, then deploy AI tools to reduce cycle time and headcount dependence.

Even if you’re not selling IT services, you still “pay” for work in internal hours. The best Singapore adopters I’ve seen do three things consistently.

1) Pick one workflow and make it boring

Boring is good. Boring scales.

Choose a workflow that happens weekly (or daily) and has a clear definition of done. Examples:

• Weekly marketing performance reporting across channels
• First-draft sales proposals for a standard service
• Product FAQ updates from support tickets
• Vendor invoice categorisation and anomaly checking

Then write it down in 10–20 steps. If you can’t describe the workflow, you can’t automate it.

2) Use AI for the “middle 60%” of the work

AI is strongest when it accelerates the heavy lifting, not when it’s forced to own the final decision.

A simple rule that works:

• AI drafts (analysis, summary, first-pass recommendations)
• Humans approve (final checks, judgement calls, compliance)
• Systems execute (publishing, sending, logging)

This model is exactly what agent plug-ins push toward: AI does the time-consuming processing; humans steer and verify.

3) Change your KPIs before you change your org chart

If your KPI is “tickets handled per agent” or “reports produced per analyst,” you’ll optimize for volume.

Switch to KPIs that force outcome thinking:

• Marketing: cost per qualified lead, speed to launch campaigns, variant velocity
• Sales: time-to-proposal, follow-up SLA, conversion rate by segment
• Ops: cycle time per process, exception rate, rework rate

One-liner to remember: AI doesn’t fix weak KPIs; it amplifies them.

Where to apply AI plug-ins first: marketing, sales, analytics

The direct answer: start where work is repetitive and data is already digital—marketing ops, sales ops, and reporting.

These areas match the plug-in promise described in the Reuters report (legal, sales, marketing, data analysis). They’re also where Singapore SMEs and mid-market firms can see results in weeks, not quarters.

Marketing: from “content production” to “campaign throughput”

AI tools can take over the repetitive parts:

• Drafting ad variants and landing page sections
• Summarising campaign performance and pulling insights
• Generating audience-specific messaging angles
• Creating standard operating checklists for launches

What you keep human-led: positioning choices, brand voice decisions, and anything that could create compliance risk.

Sales: reduce the invisible admin tax

Sales productivity often dies in the cracks:

• Account research
• Meeting notes and next steps
• Proposal drafts
• CRM updates

An agent that can prepare briefs, draft follow-ups, and structure proposals cuts the “admin tax” dramatically. The goal isn’t fewer salespeople. It’s more selling time per salesperson.

Analytics: stop building the same report 12 times

Reporting is a perfect automation candidate because it’s frequent and structured.

A realistic target for many teams is to reduce a weekly reporting cycle from 6–10 hours to 1–2 hours by automating:

• Data extraction and cleaning
• Trend detection
• Narrative summaries
• Slide first drafts

Then humans focus on what actually matters: deciding what to do next.

Risk and governance: how to adopt AI without creating a mess

The direct answer: set usage boundaries, protect sensitive data, and design verification into the workflow.

AI agents feel powerful, which is why governance has to be practical—not a 40-page policy no one follows.

A simple three-layer approach works well:

1. Data rules: what can and can’t be pasted into AI (customer PII, contracts, pricing, credentials)
2. Tool rules: which tools are approved and how access is managed
3. Check rules: what must be reviewed by a human before it goes out

If you want one habit that prevents most problems: make AI cite its inputs. Require it to reference the source doc, the dataset snapshot, or the ticket IDs it used. That way, verification is fast.

Snippet-worthy stance: If you can’t verify it quickly, don’t automate it fully.

“People also ask” (and what I tell clients in Singapore)

Will AI plug-ins eliminate jobs?

They’ll eliminate tasks first, especially repetitive junior work. The companies that win will redeploy people to higher-value activities: customer conversations, solution design, partner management, and quality control.

Should we wait until our competitors adopt AI?

No. By the time it’s obvious, the early adopters have already rebuilt processes and trained teams. Catching up becomes expensive.

What’s the fastest way to get ROI from AI business tools?

Pick one workflow with clear volume and clear quality checks. Automate 30–50% of it, measure cycle time reduction, then expand.

The Singapore takeaway: AI forces a business model upgrade

India’s market reaction wasn’t really about Anthropic. It was about what plug-ins represent: a shift from staff-based delivery to system-based delivery. That shift is coming for every business function that looks like “smart people doing repetitive work in front of a screen.”

If you’re building your 2026 plan right now, treat AI business tools as operational infrastructure—like CRM or accounting software—not as an experiment. Start small, but start with seriousness: workflow mapped, metrics set, humans in the loop.

What part of your business still scales mainly by hiring—marketing output, reporting, sales follow-ups, customer support—and what would change if you could scale it by improving the system instead?

Landing page URL: https://www.channelnewsasia.com/business/anthropics-ai-plug-ins-shake-indias-staffing-intensive-it-sector-stocks-dive-6-5905851