Tech IPOs Show What Apparel Supply Chains Must Become

Zepto is preparing draft papers for an IPO valued at about US $ 1.3 billion. That number matters less than what it represents: public markets are still willing to back operational speed, but they’re getting stricter about how that speed is achieved.

For Sri Lanka’s apparel industry—where margins are negotiated hard, compliance is non‑negotiable, and lead times decide who wins—this is a useful signal. Quick commerce lives or dies on prediction, automation, and ruthless visibility. Apparel manufacturing is different, but the pressure is starting to rhyme: buyers want faster turns, fewer errors, clearer proof, and lower waste.

This post sits inside our series on “ශ්‍රී ලංකාවේ වස්ත්‍ර හා ඇඳුම් කර්මාන්තය කෘත්‍රිම බුද්ධිය මඟින් කෙසේ වෙනස් වෙමින් තිබේද”. The aim isn’t to talk about a retail startup for the sake of it. It’s to translate what Zepto’s IPO moment says about where global expectations are heading—and what Sri Lankan manufacturers can do with AI in apparel manufacturing and supply chain automation right now.

Zepto’s IPO is a signal: growth is fine, but efficiency is priced in

Public investors aren’t buying a dream anymore; they’re buying a system that performs under scrutiny.

From the RSS report: Zepto plans to raise roughly Rs. 11,000 crore in fresh shares and list around July–September next year, using a confidential filing route that keeps room to resize the offer. Ahead of that, it has already made visible moves: reducing 800–900 roles through layoffs and vacancies, cutting customer acquisition costs, and trimming overheads.

Here’s the part Sri Lankan apparel leaders should notice: in sectors that depend on speed, the market is increasingly skeptical of “growth at any cost.” Quick commerce is cash-intensive, and Zepto faces competitors supported by listed parents. It’s entering public markets in a competitive lane where even leaders have to justify expansion and unit economics.

Translate that to apparel: global brands are also tightening. They’re less tolerant of:

• Excess inventory caused by weak forecasting
• Rework from inconsistent quality
• Delays hidden until the last week
• ESG claims without verifiable data

AI becomes valuable not because it’s trendy, but because it produces proof: forecasts, detection rates, audit trails, and predictable throughput.

Quick commerce runs on AI-grade operations—apparel can borrow the blueprint

Quick commerce’s real product is not “10-minute delivery.” It’s demand sensing + micro-fulfilment discipline.

The RSS content highlights the scale: Blinkit with more than 1,800 dark stores, and Instamart and Zepto around 1,000–1,100 outlets each. Rivals like Flipkart and Amazon are scaling fast too, with Flipkart Minutes nearing 800 dark stores and Amazon Now adding micro-warehouses daily in major metros.

That density forces a specific operating model:

• Predict demand at hyper-local level
• Allocate inventory correctly
• Pick/pack without errors
• Replenish continuously
• Detect issues early (shrinkage, wrong picks, stockouts)

Sri Lankan apparel manufacturing has its own “dark stores”: cutting rooms, sewing lines, washing, finishing, packing, and export documentation. The parallel is direct.

Where the parallels show up inside a garment factory

Answer first: If you want shorter lead times and fewer surprises, you need the same three pillars quick commerce uses—prediction, visibility, and automation.

1. Prediction (demand → production plans)

   • AI forecasting that blends buyer history, style attributes, and seasonality
   • Early risk flags for styles likely to face fabric delays or quality issues

2. Visibility (real-time truth)

   • Line-level WIP visibility (not yesterday’s spreadsheet)
   • ETA confidence scores per PO based on current bottlenecks

3. Automation (repeatable execution)

   • Automated inspection using computer vision
   • Auto-generation of compliance and traceability artifacts

If your factory still “finds problems” only in end-line or final audit, you’re operating like a quick commerce warehouse that checks stock only once a week. That business wouldn’t last a month.

AI in Sri Lanka’s apparel industry: the use cases that pay back fastest

Most companies get this wrong by starting with the fanciest tool (or the cheapest pilot) instead of the highest-friction process.

Below are AI use cases that typically show ROI because they cut cost drivers that buyers actually punish: rework, delays, and uncertainty.

1) Computer vision for fabric and garment quality control

Answer first: Computer vision reduces subjectivity and catches defects earlier—before they become rework and shipment risk.

Instead of relying only on human inspection, a camera-based system can flag:

• Fabric defects (holes, stains, barre, slubs)
• Stitching inconsistencies
• Shade variation (with proper lighting and calibrated workflows)
• Measurement deviations for key points

This matters because rework isn’t just labour. It burns capacity, disrupts line balancing, and increases airfreight risk. AI-based QC works best when paired with a clear “stop-the-line” rule and a defect taxonomy your team agrees on.

2) AI-assisted line balancing and production scheduling

Answer first: AI scheduling improves throughput by reacting to constraints in minutes, not meetings.

Garment production is full of micro-constraints: absenteeism, machine downtime, SMV variance by operator, and style complexity. AI models can recommend:

• Operator allocation and reallocation
• Bundle release timing
• Predicted bottlenecks by operation

I’ve found the biggest win here is cultural: when planners trust the numbers, the “heroic last week” becomes less common.

3) Automated compliance workflows and audit-ready documentation

Answer first: AI helps build consistent, searchable evidence trails—exactly what global buyers want as scrutiny rises.

Compliance teams drown in repetitive documentation: certifications, test reports, supplier declarations, and corrective actions. With AI and workflow automation, factories can:

• Extract key fields from PDFs and emails
• Auto-create checklists per buyer standard
• Track expiries and non-conformities
• Produce audit packs faster

This aligns tightly with the series theme: අනුකූලතා ක්‍රියාවලීන් ස්වයංක්‍රීය කිරීම is no longer optional when buyers want faster approvals and clearer traceability.

4) Forecasting for raw material planning (the hidden profit lever)

Answer first: Better material planning reduces deadstock and emergency purchasing.

Quick commerce lives on avoiding stockouts and avoiding overstock. Apparel can adopt the same discipline:

• Predict fabric demand by colour/attribute, not only by style
• Plan trims using consumption variance from historical production
• Flag POs likely to miss because supplier lead times are slipping

For Sri Lanka—where import lead times and currency exposure can bite—this is one of the most practical places to apply AI.

5) AI-generated digital content for buyer communication

Answer first: Generative AI speeds up communication and reduces avoidable back-and-forth.

This doesn’t mean “spam buyers with AI emails.” It means:

• Faster creation of tech-pack clarifications and change summaries
• Cleaner incident reports with images, timelines, and CAPA drafts
• Internal multilingual summaries for teams (merchandising, QA, production)

It supports the series goal of ජාත්‍යන්තර වෙළඳ නාම සමඟ සන්නිවේදනය ශක්තිමත් කරන ඩිජිටල් අන්තර්ගතය.

What Zepto’s cost-cutting teaches: AI isn’t a project, it’s governance

Cost-cutting before an IPO can look like a headline, but the deeper lesson is governance: you can’t scale chaos.

Zepto’s preparation phase includes lowering acquisition costs and overheads, and competing in a sector where leaders are publicly debating whether capital markets will keep funding expansion. That dynamic forces a discipline Sri Lankan apparel firms should adopt proactively: every major process needs measurable unit economics.

Here’s a practical way to frame an AI roadmap in apparel manufacturing:

A 90-day “AI readiness” checklist for apparel manufacturers

Answer first: Start with data discipline and one high-impact workflow, then scale.

1. Pick one measurable pain

   • Example: reduce rework minutes per line, improve FPY, reduce shade claims, improve OTIF

2. Standardise your data capture

   • One defect dictionary
   • One downtime reason code list
   • One PO status definition across merchandisers

3. Create a single operational dashboard

   • WIP by line
   • ETA confidence
   • Top 5 defect types and cost

4. Pilot one AI use case with clear guardrails

   • Define who acts on the output
   • Define what changes when the model flags risk

5. Train supervisors, not just analysts

   • If the floor can’t use it, it’s theatre

This approach avoids the common trap: buying tools that look impressive but never change decisions.

The real opportunity for Sri Lanka: sell “predictability,” not just capacity

The strongest commercial stance for Sri Lanka’s apparel sector is predictable delivery with verifiable quality and compliance.

Quick commerce startups are raising and listing because they can translate operational capability into market confidence. Apparel exporters can do something similar with buyers: make your reliability measurable.

A Sri Lankan factory that can say:

• “We catch 70% of critical defects before end-line”
• “We provide PO-level ETA confidence updated daily”
• “Our audit pack is generated within 24 hours of request”

…isn’t selling a commodity service. It’s selling reduced risk.

And risk reduction is one thing brands still pay for, even when they negotiate price aggressively.

Next steps: turn AI from a buzzword into a buyer-facing advantage

Zepto’s US $ 1.3 billion IPO plan is a reminder that markets reward operational systems, not hype. For Sri Lanka’s apparel industry, the same principle applies with global brands: proof beats promises.

If you’re mapping your next 6–12 months, I’d start with two tracks in parallel: (1) computer vision or process analytics to reduce quality loss, and (2) compliance/document automation to reduce approval friction. Those two alone improve OTIF and buyer confidence—fast.

Our broader series on ශ්‍රී ලංකාවේ වස්ත්‍ර හා ඇඳුම් කර්මාන්තය කෘත්‍රිම බුද්ධිය මඟින් කෙසේ වෙනස් වෙමින් තිබේද keeps coming back to one idea: AI works when it changes decisions on the floor and in the merch room.

So here’s the question worth ending on: when your next big buyer asks for faster lead times and stronger traceability, will you respond with a promise—or with dashboards, detection rates, and audit-ready evidence?