A high-quality, cash-rich outsourcing firm that does the specialised back-office work big banks and brands do not want to do in-house. It is growing 22% and returns 35% on capital, and it has still de-rated, because the market keeps asking whether AI shrinks the very work eClerx gets paid for.
Mental model
eClerx is a specialised business-process firm, not a software company: it rents out trained brains at India cost.
Large banks and brands have oceans of repetitive, specialised operations work (trade processing, data management, product content) that is too costly and too distracting to run in-house. eClerx deserves to exist because it does that work cheaper, faster and with deep domain training, especially in financial-markets operations.
Why has no one else already won? Because the work is embedded and specialised. Once eClerx's people understand a bank's messy internal systems and processes, moving the work is risky, so clients stay. That stickiness is the moat, and it is exactly what AI could erode.
Mental model heatmap
★★★★★
Labour Arbitrage
The core edge is doing skilled work from India far cheaper than onshore.
★★★★★
Domain Expertise
Deep knowledge of derivatives operations and client systems is hard to replicate.
★★★★★
Switching Cost
Embedded processes make clients slow to move work elsewhere.
★★★★★
Cash Generation
Asset-light services throw off cash: FCF tracks profit closely.
★★★★★
Automation / AI Disruption
The per-person billing model is the thing AI threatens to compress.
★★★★★
Client Concentration
Top-10 clients are still about 59% of revenue.
Economic engine
Demand
Outsourced back-office work
Banks and brands wanting to offload repetitive operations.
Unit
A billed employee (FTE) or, increasingly, an outcome
Historically paid per person; now shifting toward outcome-based deals.
Volume
More seats, more processes, more clients
FY26 revenue up ~22%, top-10 concentration easing to 59%.
Gross take
Billing rate minus India delivery cost
The arbitrage spread is the gross engine.
Operating profit
~23% EBITDA margin
Wage inflation and pricing pressure are the swing factors.
Capital required
Very light
Capex is tiny (~Rs 130 cr on Rs 4,000+ cr revenue).
Returns
ROCE ~35%, ROE ~29%
Light capital plus a healthy spread equals high returns.
Strategic position
Generic BPO
Commodity seat-rental, easily undercut and easily automated.
↓
eClerx
Specialist in financial-markets operations and digital content, with domain depth generic players lack.
↓
Global majors (Accenture, Genpact, WNS)
Bigger, broader, more resources to build AI-led delivery at scale.
Why now
The stock fell about 14% over the past year while profit grew 30%, so the multiple compressed from the high-20s toward 23x. This year's numbers are not the worry. What the market is pricing is what AI does to the model over the next several years, and that is laid out in the AI fork below. If the market is overestimating the impact of AI, the lower multiple could prove attractive. If the underlying economics deteriorate, the lower multiple may not be cheap enough.
What the market is betting on
AI becomes a tool eClerx sells (productivity, outcome deals), not a wave that drowns the seat model.
The shift to outcome-based pricing protects revenue as per-FTE billing fades.
Domain depth keeps clients sticky through the transition.
Margins hold in the low-20s despite wage and pricing pressure.
Why it is winning
Revenue growing ~22% with profit up ~30% in FY26, so demand is clearly there.
Very high returns (ROCE ~35%) on an asset-light base that converts profit to cash.
Deep specialisation in derivatives and financial-markets operations is a genuine niche.
Actively moving to outcome-based and productised deals rather than clinging to per-seat billing.
Why it could stop winning
AI could automate the repetitive work eClerx bills for, compressing the per-person model.
Wage inflation and pricing deflation are already squeezing margins quarter to quarter.
Top-10 clients are ~59% of revenue, so losing one hurts.
Larger rivals have deeper pockets to build AI-led delivery and undercut on price.
Sector mental models
Pricing Power
Medium
Domain depth helps, but clients push for deflation and AI-led savings.
Switching Cost
Medium
Embedded processes are sticky, but not permanent.
Cash Generation
Excellent
Asset-light delivery, FCF close to profit.
AI Disruption
High risk
The seat-based model is squarely in AI's path.
Client Concentration
Medium risk
Top-10 at ~59%, improving but still meaningful.
One sentence to remember
eClerx is a superb operator of a business model with a real question mark over its head. Judge it on whether AI becomes its tool or its undoing, not on this quarter's margin.
The AI fork
AI will almost certainly reduce the human effort in this work. The real question is not whether, but who captures the value that reduction creates.
AI arrives → human effort per task falls
↓
The client captures the saving
Client needs fewer FTEs
eClerx bills fewer people
Revenue falls
Margins compress
eClerx captures the value
Same work, fewer people
Charges for the outcome, not headcount
Revenue per employee rises
Higher-value relationship
That is what we need to figure out: does eClerx capture the productivity gain from AI, or does the client?
The counterargument
What if AI makes outsourcing more valuable, not less?
The bear case assumes AI shrinks the work. It might instead change it. Banks do not obviously need fewer outsourced operations in an automated world; they may need different ones. As workflows get automated, someone still has to implement the systems, monitor them, catch what they get wrong and manage the whole thing. If eClerx moves from supplying people to owning that workflow, the value it captures per client could rise even as the number of people it deploys falls.
This is a possibility, not a forecast. eClerx has to prove it in the revenue mix and the economics, not the concall. Management saying it is not the same as it happening.
01Company Overview
When an investment bank processes millions of derivative trades, or a luxury brand needs thousands of product photos edited and tagged for its website, that work is repetitive, detailed and expensive to do in-house. eClerx is the company you hand it to. It runs the unglamorous but critical back-office operations for large financial firms, telecom and media companies, and retail and luxury brands. Historically it charged mostly by the person: you need 100 trained people doing trade support, eClerx supplies and manages them from India at a fraction of the onshore cost. That per-person model built a 35%-return business. It is also the model AI now threatens, which is the whole debate around the stock.
No repackaging games. eClerx listed in 2007 and has traded for well over a decade. It did do a 1:1 bonus issue in March 2026, which doubles the share count and halves the per-share price optically but changes nothing about the business or your ownership.
02Business Model & Industry
Unit of revenue: Historically, one billed employee. Think of it like hiring a team: a client says it needs, say, 200 trained people to run its trade-support desk, and eClerx supplies and manages them from India, charging a rate per person that sits well below the onshore cost. The company is now shifting some of this toward outcome-based pricing, where it charges for a result delivered rather than heads deployed.
Model: Mostly staff-based (per-FTE) contracts today, migrating toward outcome-based and productised deals. Recurring, multi-year client relationships rather than one-off projects.
The whole business is one spread. Follow the money down the chain.
Client
A global bank needs, say, 200 trained people for trade support
↓
eClerx
Hires, trains and manages that team, delivered from India
↓
Client cost
Pays far less than the onshore cost of doing it in-house
The gap between what the client would pay onshore and what eClerx spends to deliver from India is the arbitrage spread. That spread, not software, is the gross engine of the whole company.
Financial Markets operations48%
The specialist core. Derivatives and trade operations for banks. Sticky, higher-value.
Digital / Customer Operations42%
Content, marketing ops, customer support for media, telecom, retail and luxury.
Analytics & Automation10%
The growth bet, now a ~USD 90m book. Higher-value, AI-adjacent.
Structure
Fragmented and competitive. Many BPM and KPO players, from global majors to niche specialists.
Competitors
Global majors like Accenture, Genpact and WNS are bigger and broader. Firstsource and Zensar are closer listed comparisons. eClerx competes on domain depth, especially in financial-markets operations, rather than scale.
Pricing power
Medium and under pressure. Domain expertise gives some pricing hold in the financial-markets niche, but clients are actively pushing for AI-led cost savings across the industry.
Demand driver
The desire of large firms to offload repetitive, specialised operations work at lower cost. That driver is structural, but the form it takes (people versus software) is changing. (Structural (outsourcing is not going away) but with a technology shift (AI) reshaping how the work gets done and priced.)
TAM
Large global BPM and outsourcing market growing at high single digits, but the value is migrating from headcount-based delivery toward automation and outcome-based models.
Penetration
Outsourcing is well established; growth here is share-gain and moving up the value chain, not a greenfield land grab.
Value-chain seat
The outsourced operations layer sitting behind a client's own systems. Valuable while the work is human-intensive, exposed if that work automates away.
As an operator, eClerx is genuinely excellent: ~22% revenue growth, ~35% ROCE, tiny capital needs, and cash flow that tracks profit. The niche in financial-markets operations is real and sticky. As an operator, there is very little to fault. The structural question (whether the pivot outruns the pressure on the FTE model) is real but separate, and the two-engine assessment and summary take it up.
03Valuation Snapshot
Market Cap
₹16,639 cr
52W High / Low
₹2,498 / ₹1,319
Stock P/E
22.8
₹1,769 / TTM EPS ₹77.6 (incl Q1FY27); ₹75.09 below is FY26 only
P/B
6.5
FY26 EPS
₹75.09
post 1:1 bonus (Mar 2026); TTM incl Q1FY27 ~₹77.6
Book Value
₹272
04Financial Performance (5Y, in Crores)
FY22
₹2,160net ₹418 · 19.3%
FY23
₹2,648net ₹489 · 18.5%
FY24
₹2,926net ₹512 · 17.5%
FY25
₹3,366net ₹541 · 16.1%
FY26
₹4,117net ₹706 · 17.2%
RevenueNet profit₹ crore · % = PAT margin
05Key Ratios
ROE
29%
ROCE
34.8%
PAT Margin
~17%
wage + pricing pressure
D/E
0.17
essentially net-cash
Debtor Days
59
normal for IT services
06Cash Flow Forensics (in Crores)
FY24
OCF₹526Capex₹64FCF₹462
FY25
OCF₹655Capex₹116FCF₹539
FY26
OCF₹873Capex₹132FCF₹756
This is a clean cash machine. Operating cash flow beats reported profit, capex is tiny (about Rs 132 cr on Rs 4,000+ cr of revenue) because a services business needs desks and laptops, not factories, and free cash flow of Rs 756 cr in FY26 is comfortably positive. Debtor days at 59 are normal for IT services, where clients pay on monthly billing cycles. There is no accounting worry here: the profit is real and it turns into cash. The risk in this business is strategic (AI), not financial.
07Growth
Sales CAGR 5Y
21%
Profit CAGR 5Y
22%
Profit growth TTM
28%
Q1 FY27 PAT
₹164 cr
+16% YoY, ~7% down QoQ on wage hikes
Q1 FY27 revenue
₹1,152 cr
+24% YoY; EBITDA margin 24.2%
Stock CAGR 1Y
-14%
profit grew 30% same window
08Management
Founder-led by PD Mundhra and Anjan Malik, who built eClerx from 2000 and still anchor the promoter group (about 54.5%, with zero pledge). The scorecard is an operator's: two decades of high-return, cash-generative growth without diluting shareholders or piling on debt. The current test is the hardest of their careers: navigating an AI shift that threatens the per-person model they built the company on. To their credit, they are meeting it head-on in the concalls, talking openly about deflationary pressure, moving to outcome-based and non-FTE deals, and building the analytics and automation book. At least so far, they appear to be addressing the problem rather than dismissing it. Judge them over the next few years by whether the value-added mix grows faster than the seat business shrinks.
Domain depth in financial-markets and derivatives operations
Embedded processes inside client systems (switching cost)
Long-standing Fortune-500 relationships
Asset-light, high-return delivery model
The moat is real but narrow, and it is being tested. eClerx's edge is knowing a client's messy operations so well that moving the work is risky, plus genuine specialism in derivatives processing that generic BPOs lack. That has kept clients for years. The honest caveat is the direction the moat does not defend: AI does not out-compete eClerx on service, it shrinks how much human work there is to bill for (the 'shrinking island' lens below unpacks this). Narrow, real today, and on notice.
11The Story So Far
eClerx has quietly compounded for two decades: revenue from Rs 2,160 cr in FY22 to Rs 4,117 cr in FY26, profit from Rs 418 cr to Rs 706 cr, all on tiny capital and essentially net cash (D/E 0.17). FY26 was a strong year, revenue up ~22% and profit up ~30%, capped with a 1:1 bonus issue. And yet the stock fell about 14% over the year. The reason is not the results, it is the theme. eClerx sits in the BPO/KPO sector, the corner of IT most directly in the path of AI, and the market has spent the year re-rating anything that bills by the hour or the head. So a business that grew earnings 30% saw its multiple compress instead of expand.
Price action (12M): Roughly ₹2,050 down to ₹1,769 over twelve months (about minus 14%), well off the ₹2,498 high but comfortably above the ₹1,319 low. The de-rating, not the results, drove the move. Why the multiple fell while earnings rose is the two-engine assessment below.
12Risks
AI disruption of the core model. eClerx is paid largely for human effort on repetitive work, which is exactly what AI automates. If clients use AI to cut headcount-based spend faster than eClerx moves to outcome-based revenue, growth and margins compress. High.
Margin pressure. Indian wage inflation pushes costs up while clients push billing rates down (deflationary pressure). Margins have drifted and can drift further. Medium-High.
Client concentration. Top-10 clients are about 59% of revenue. Losing or shrinking one large relationship would hurt. Medium.
Competition from larger players. Accenture, Genpact and WNS have deeper resources to build AI-led delivery and undercut on price. Medium.
Currency. Revenue is largely in USD against India costs, so a sharp rupee move swings reported margins. Medium.
13Where the Numbers Could Mislead
✓
Profit up, cash flow not
OCF beats profit; FCF Rs 756 cr in FY26
✓
Receivables blowing up
Debtor days 59, normal for IT services
–
Project / POC accounting
Recurring services billing, not project POC
✓
Promoter pledging
Zero pledge; founders hold ~54.5%
!
Structural disruption risk
AI directly threatens the per-FTE model
!
Margin trend
PAT margin drifted from ~19% to ~17% on wage + pricing
!
Client concentration
Top-10 clients ~59% of revenue
Sector checklist
✓
Cash conversion
Asset-light; FCF tracks profit
✓
Return on capital
ROCE ~35%, ROE ~29%
!
Revenue durability (AI)
Per-FTE model squarely in AI's path
!
Pricing / margin
Wage inflation up, billing rates under deflation pressure
!
Client diversification
Top-10 at ~59%, improving but still high
14Two-Engine Assessment
Earnings engine
The earnings engine is running well and is cash-backed: FY26 profit up ~30%, TTM profit growth ~28%, ROCE ~35%, and free cash flow that tracks profit. Q1 FY27 (June 2026) is the near-term wobble in one number: revenue up ~24% YoY to Rs 1,152 cr, but profit of Rs 164 cr grew only 16% YoY and fell about 7% QoQ as the April wage hike and new delivery centres (Lima live, Cairo starting) pushed EBITDA margin to 24.2% and PAT margin to 14%. Growth is fine; the squeeze is on margin, exactly the tension this report flags. The catalyst from here is the analytics and automation book (now ~USD 90m) scaling; the drag is pricing deflation.
Multiple engine
The multiple has already compressed against the stock. Over the past year profit grew ~30% while the stock fell ~14%, so the PE de-rated from the high-20s toward 23x. That is the market pricing AI risk, not a business miss. At 23x for a 35%-ROCE, debt-free compounder, the multiple is not demanding by its own history, but it is not a giveaway either given the structural overhang.
On the mechanics, earnings are rising while the multiple has compressed and returns are stable. That creates re-rating potential, but only if the market's AI concern proves overstated. The compression here is not pure sentiment the way it would be for a company with no structural threat; it reflects a real, unresolved question about the durability of the revenue itself. So the optionality is genuine but conditional on the AI transition going well. The evidence to watch is whether the value-added mix grows fast enough to offset pressure on the seat-based business.
15Mental-Model Lenses
The price fell while earnings rose
Profit grew about 30% in FY26 but the stock fell about 14%, so the multiple compressed rather than expanded, and ROE held near 29% the whole time. In a business with no structural threat that would be a clean re-rating setup. Here the compression is the market pricing the AI question, so read it as conditional, not free.
A moat around a shrinking island
eClerx's stickiness comes from doing embedded, repetitive work no one wants to move. But AI attacks from the flank: it does not beat eClerx on service, it shrinks how much human work there is to bill for. A moat is only as good as the thing it protects, and here that thing is under technological pressure.
Loyalty measured before the alternative existed
Years of high client retention were earned in a world where the only realistic option was human-run offshore delivery. AI is the first genuine alternative to that model. Historical retention tells you clients were sticky when there was no substitute; it tells you less about how sticky they stay once one appears.
Operator meeting the threat head-on
This is the encouraging side. Management is not dodging the AI question in concalls; it is talking openly about deflation, moving to outcome-based and non-FTE deals, and building the analytics book. That is operator behaviour. The thesis is not that the threat is fake, it is that this team may be good enough to navigate it. Judge them by the value-added mix, not the rhetoric.
15.1AI Transition Scorecard
You do not have to guess which way the fork breaks. These are the numbers that tell you, quarter by quarter, whether eClerx is capturing the AI gain or losing it to clients.
MetricGood signBad signWhy it matters
Revenue per employeeRisingFlat or fallingTests whether growth is decoupling from headcount, the core AI question.
Non-FTE / outcome revenueGrowing shareStagnantShows the shift from billing people to billing for results.
Analytics & automation bookScaling past ~USD 90mStallsShows movement toward higher-value, AI-adjacent work.
Headcount vs revenueRevenue grows faster than headsBoth rise togetherTests how dependent growth still is on adding labour.
Pricing / billing rateStablePersistent deflationShows who captures the AI productivity gain, eClerx or the client.
EBITDA marginHolds low-20s or risesKeeps decliningTests whether the productivity gain is retained, not competed away.
Large-client spendStable or risingShrinking walletsTests whether AI is shrinking how much clients need to spend at all.
eClerx does not disclose all of these cleanly each quarter, so read the trend across the ones it does. Revenue per employee is the single clearest window into whether the economics are moving away from pure headcount.
17Summary
eClerx pairs high growth (22%), high returns (35% ROCE), a net-cash balance sheet and clean cash flow, and after a 14% fall while earnings rose 30% it trades at about 23x. Those two facts sit awkwardly together. The reason is the labour-arbitrage model: AI is built to compress exactly the human-effort billing eClerx lives on. Management is pivoting to analytics, automation and outcome deals, and it is too good an operator to wave away, but whether the pivot outruns the pressure is genuinely unknown. The old model clearly works. Whether it keeps working once the work itself becomes easier to automate is the open question, and the next few quarters of revenue mix, revenue per employee and pricing should start to answer it. This is not a buy or sell call. Do your own work and talk to a SEBI-registered adviser.
Figures are a point-in-time snapshot as of 8 Aug 2026 and may be stale.