In today’s deep dive, we will explore how AI is changing India’s SaaS, IT-services and KPO businesses. Research firms like Evalueserve are automating work once done by analysts, IT-services companies are building AI products, and SaaS startups are hiring engineers to implement their software. Through real examples, we will examine what this means for margins, valuations, jobs and the increasingly blurred line between software and services.

For years, banks paid teams of analysts to do work that was important but painfully repetitive. They would go through borrowers’ financial statements, extract hundreds of numbers, check them for accuracy and enter them into lending systems. The documents came in different formats and languages, making even routine financial analysis a time-consuming exercise. For companies like Evalueserve, this was good business. Global banks needed the work done, and outsourcing it to skilled analysts in India was considerably cheaper than doing it themselves. But something interesting is happening now. The very companies that built their businesses around this kind of work are developing software that can do much of it without those analysts.

Evalueserve, one of the better-known names in India’s knowledge process outsourcing (KPO) industry, helped automate this process using its software, Spreadsmart. According to its published case study, financial data extraction became 70% faster, with 99.6% accuracy. Integrating the software with the bank’s existing lending platform reduced overall processing time by 21%. Additional AI features improved efficiency further.

What's interesting is that Evalueserve had built much of its business by providing the analysts, research capabilities and financial expertise required for exactly this kind of work. Now it was selling technology that reduced the need for manual effort.

This isn't an isolated example. Across India's outsourcing industry, companies that once earned money by supplying skilled people are building software to automate their own services. At the same time, software companies that were supposed to sell products with relatively little human involvement are hiring engineers to customise, integrate and operate those products for enterprise customers.

The distinction between a software company and a services company is becoming harder to maintain. And for India, where outsourcing and software have developed into two large but very different industries, the consequences could be substantial.

For more than two decades, India’s KPO industry benefited from a simple arrangement. Companies in the US and Europe needed financial research, market analysis, competitor tracking, forecasting models, investment reports and business intelligence. Hiring large teams locally was expensive, while outsourcing the same work to India offered considerable savings.

Companies such as Evalueserve, SG Analytics, The Smart Cube and Netscribes built businesses around this demand. They employed graduates, MBAs, finance professionals and researchers who became extensions of their clients’ teams.

An investment bank could outsource company profiles and financial models. A consulting firm could commission market-sizing studies and competitor reports. A consumer goods company could employ an external team to track pricing, product launches and changes in market share.

The work ranged from routine information collection to sophisticated analysis. But a significant portion involved collecting data, organising it, maintaining spreadsheets, updating presentations and producing standardised reports.

The economics depended heavily on people. A KPO would employ analysts in India at a lower cost than its overseas customer, charge the client for the work and earn the difference after salaries, infrastructure and management expenses.

As demand increased, the company hired more analysts. Experienced employees supervised larger teams, while new graduates performed much of the initial research and data collection.

AI is now challenging the assumption that expanding the business requires expanding those teams.

Financial information can be extracted from annual reports in minutes. Earnings calls can be transcribed, summarised and compared with previous quarters. AI tools can produce initial company profiles, update financial models, identify changes in competitor strategies and prepare drafts of industry reports.

They still make mistakes, and financial analysis requires judgment. But much of the preparatory work that once occupied junior analysts can now be completed far more quickly.

This creates a problem for KPO companies. Their customers still need research, but they may no longer need to pay for the same number of research hours.

The difference matters because outsourced research was never valuable merely because somebody spent three days preparing it. The customer wanted reliable information to make a decision. If that information can be delivered in three hours, the economic value of those three days of labour becomes much harder to defend.

Evalueserve's response has been to develop reusable software that can automate common research and analytical tasks while retaining its industry expertise and services business.

Its products now include Insightsfirst for market and competitive intelligence, Spreadsmart for financial data extraction, and RiskSight for risk assessment. Its AIRA technology supports research automation, financial information extraction, document analysis and other industry-specific applications.

But the more revealing part is how the company describes its change in strategy.

In a detailed account of Evalueserve's product development, CTO Rigvi Chevala explained that the company previously built many applications in response to individual customer requirements. Some solutions could have been useful to other clients, but they were not always designed for reuse.

That approach started changing around 2019, well before the current enthusiasm around generative AI. Evalueserve wanted to build products for problems common to multiple customers, rather than developing separate applications every time a client requested something.

The company calls this approach Product-Led Services.

One example mentioned by Chevala involved a customer investing $2 million annually in a customised platform that combined customer and market intelligence. This is substantially more than the price of a typical software subscription, but it also involves more work than simply giving somebody access to an online product.

Evalueserve provides the underlying technology, builds it around the customer's requirements and combines it with its own analysts and industry specialists.

Its expansion into data engineering, cloud migration, software integration and enterprise AI implementation adds another dimension. These are activities traditionally associated with IT-services companies rather than research outsourcing firms.

Evalueserve is no longer competing only with other KPOs to produce better reports. It is also competing with software companies and technology consultants to become part of the customer's data and research infrastructure.

That could make its customer relationships more valuable, particularly in industries such as banking where years of experience with credit models, financial statements and regulatory requirements are difficult to replicate.

But there is also a commercial risk. A customer that once paid for a team of analysts may now expect the software to do most of the work at a lower price. Unless Evalueserve creates enough additional value, automating its services could reduce revenue even as productivity improves.

SG Analytics offers another useful example. The company began with investment research before expanding into financial services, market intelligence, data analytics and AI engineering.

Its AI Invest Studio combines specialised research software with financial analysts and experienced investment professionals. It covers activities such as company research, financial modelling, portfolio monitoring and investment reporting.

SG Analytics claims that its approach can reduce research turnaround times by 40% to 60% and investment operating costs by around 30%. These are company-reported results, not independently established industry averages, but they indicate the level of automation it is offering customers.

The company has also developed AI Deal Studio for investment banks and private equity firms. The software assists with company profiles, investment presentations, financial models, valuation work and transaction documents, while experienced professionals review the results.

Interestingly, SG Analytics itself does not claim that AI can replace all the work involved. In a September 2026 assessment of investment-banking documents, it estimated that AI could handle roughly 40% of the work involved in preparing a confidential information memorandum, while the remaining work required human judgment.

That distinction is important. AI may be able to extract a company's revenue, identify competitors and draft a business overview. Deciding which information matters to a prospective buyer, how to explain weak margins or whether management's growth projections are credible remains a different task.

C5i, formerly Course5 Intelligence, is pursuing a similar strategy through products such as Compete, which continuously collects and analyses market and competitor information. Instead of delivering another competitor report every quarter, the company can provide customers with a platform that keeps monitoring the market.

The Smart Cube, acquired by WNS in 2022, has followed a comparable route in procurement intelligence. Its Amplifi PRO platform combines research with software that helps businesses track suppliers, commodities, prices and market developments. By August 2023, its database included more than 1,150 intelligence reports and dashboards, along with price data covering over 9,000 commodities.

These companies were developing technology long before the current AI boom. What has changed is the amount of work their software can perform and the price customers may now be willing to pay for tasks previously handled by analysts.

Consider a hypothetical KPO with a ₹1 crore annual contract to provide financial research.

Suppose it spends ₹60 lakh on analysts, ₹20 lakh on supervision and other expenses, and retains ₹20 lakh as operating profit.

Now imagine AI allows the company to reduce analyst costs to ₹30 lakh. It spends another ₹10 lakh on software, cloud infrastructure and related running costs. Its total cost falls to ₹60 lakh, increasing operating profit to ₹40 lakh.

Initially, this looks attractive. The company has doubled its profit without acquiring another customer.

But the customer also understands what AI can do. During the next contract renewal, it may ask why the service still costs ₹1 crore when much of the work has been automated.

If the customer negotiates the annual fee down to ₹70 lakh, the KPO's operating profit falls to ₹10 lakh under the same cost assumptions. It is now more productive than before, but earns less money.

And this simple example doesn't include the substantial upfront cost of developing the software.

There is another possibility. The customer could decide that its own employees, equipped with AI tools and access to internal information, no longer need an external research provider for that work.

This is why productivity improvements cannot automatically be treated as future margin improvements.

In August 2026, Reuters reported that India's large IT-services companies were facing similar pressure. Customers were demanding lower prices, shifting towards contracts linked to completed work rather than employee hours, and bringing some activities in-house.

For KPOs, the risk could be particularly acute because many of their traditional services involve processing information, precisely the kind of work AI has become better at performing.

However, not every KPO activity is equally exposed. Access to proprietary databases, specialised financial models, confidential company information and experienced industry professionals can still matter enormously.

A generic competitor summary may be easy to reproduce. A detailed assessment of a pharmaceutical company's clinical pipeline, or a credit model incorporating a bank's underwriting policies and historical defaults, is considerably harder.

The difference between those two kinds of work may determine which KPO businesses survive and which lose relevance.

The same technology is changing another part of India's outsourcing industry, although the pressure is arriving from a different direction.

Traditional IT-services companies have spent decades designing applications, maintaining enterprise systems, managing infrastructure and supplying engineers to global customers.

Like KPOs, their growth has been closely connected to headcount. More contracts generally required more engineers, while improving profitability involved better staff utilisation, higher billing rates and lower delivery costs.

AI allows these companies to automate portions of software development, testing, maintenance and data processing. But it also means customers may question why they should continue paying for the same number of engineering hours.

One response has been to develop proprietary software that makes delivery faster and can be reused across customers.

Quantiphi, founded in 2013, illustrates this approach. The company built its business around AI engineering, data analytics and enterprise technology projects. It now offers proprietary platforms including baioniq, an enterprise generative AI platform, and Codeaira, which helps automate software development activities.

Rather than building every enterprise application from the beginning, Quantiphi can use common technology across customers and customise it for their requirements.

Tredence has adopted a similar approach through ATOM.AI, its collection of reusable data engineering and AI tools. In one customer example, it reported a 90% reduction in the time required to collect and organise information for a private equity firm's work, with estimated annual savings of $8 million to $10 million.

These platforms can make services businesses more productive. But they are not necessarily equivalent to independently sold software products.

A platform that helps consultants deliver projects faster may improve margins while leaving the company's revenue almost entirely dependent on services contracts.

MathCo provides an especially revealing financial example.

Formerly TheMathCompany, MathCo has developed NucliOS, a proprietary platform used to build AI applications and business analysis tools. The company reported more than 100 applications of its technology and claimed over $800 million in business value generated for customers.

Yet its FY25 financial statements, reported by Entrackr, show an overwhelmingly services-driven business.

Operating revenue was ₹501.5 crore, barely above the previous year's ₹499 crore. The company's reported income came entirely from services. Employee expenses accounted for ₹374 crore, or around 84% of total expenses.

MathCo did improve its profitability considerably. Net profit increased from ₹15 crore to ₹64 crore as expenses declined, including a 7% reduction in employee costs. But these numbers do not establish how much of the improvement came from AI rather than broader cost controls.

The interesting point is that a company can possess sophisticated proprietary technology while remaining financially a services business.

Fractal, which listed in February 2026, offers a larger example. It has built Cogentiq, an enterprise AI platform designed to integrate business data and automate tasks across customers. For FY26, Fractal reported approximately ₹3,300 crore in revenue, a 47% gross margin and an 18% adjusted EBITDA margin.

It also invested 6.4% of revenue in research and development, including operating and capital expenditure.

Fractal demonstrates what is required to build proprietary technology alongside a substantial services operation. The company must fund product development while continuing to employ the engineers, researchers and industry specialists who deliver customer work.

Whether this investment eventually produces a larger share of separately sold software revenue is a crucial question.

While services firms are investing in products, an almost opposite change is taking place among SaaS companies.

The traditional attraction of SaaS was that a company could develop software once and sell subscriptions to thousands of customers. Additional customers did not necessarily require a proportional increase in engineering staff.

That model still works for many products. A small business can subscribe to accounting software, configure it and begin using it without employing a consulting team.

But complex enterprise AI products are different.

Consider Leena AI, which sells AI assistants for employee support, HR and IT operations. Its technology can automate requests that would otherwise be handled by internal support staff.

For Godrej Industries, Leena AI helped deploy a digital assistant that could connect with the company's existing systems and handle employee IT queries.

Getting such a product to work involves much more than activating a software licence. The assistant needs to understand company policies, connect to internal systems, recognise employee permissions and respond reliably to requests.

Leena AI's own hiring provides evidence of how the business is adapting. In 2026, it advertised for Forward Deployed Engineers, technical employees responsible for working directly with enterprise customers, building integrations, configuring products and taking implementations into production.

These are activities that resemble IT consulting, even though Leena AI primarily sells its own software.

Sarvam AI offers a similar example. The Indian AI startup has recruited engineers to work directly with enterprises, integrate its models into their internal systems and make sure the resulting applications function reliably.

Its job description covers everything from writing integration code to deploying applications inside customer environments. Sarvam identifies companies such as Tata Capital, SBI Life, CRED and LIC among its enterprise partners.

Yellow.ai's work with Bharat Petroleum shows what this means in practice.

The company developed Urja, a multilingual customer-service assistant for BPCL, supporting 13 languages and a wide range of use cases involving LPG bookings, customer support and other activities. The project required integration with BPCL's private-cloud infrastructure and internal systems.

Yellow.ai had a software platform, but the actual customer deployment required implementation work specific to BPCL.

For these companies, selling the product and making it work are closely connected. Every major customer may have different databases, security requirements, internal processes and ways of measuring success.

That can create stronger customer relationships, but it also creates expenses that a conventional software subscription may not cover.

The important test is whether the company becomes better at deploying its product with each new customer. If implementation costs fall as its customer base grows, the software model remains attractive. If every new contract requires another large team of engineers, growth begins to resemble a services business.

SquadStack is an interesting example because it comes from yet another part of India's outsourcing industry.

The company historically combined technology with human teams handling sales and customer conversations. It has since moved further into AI-powered calling through its VANI platform.

For IndiaMART, SquadStack developed a voice AI system that contacts buyers, understands requirements, follows up on enquiries and connects relevant leads with sellers.

The complexity goes beyond answering standard questions. IndiaMART's customers may switch languages, change their requirements halfway through a conversation or communicate from noisy factories and shops.

According to the companies' reported results, the system now handles more than one lakh conversations daily. It reportedly delivers 20% higher conversions than manual calling and reduces the cost per confirmed lead by approximately 15%.

These are company-reported comparisons, but the underlying change is significant.

SquadStack once relied heavily on people making calls. It now owns technology capable of handling a substantial volume of those conversations, while human employees remain involved in exceptions and other tasks.

Its commercial opportunity is increasingly connected to the result delivered rather than the number of employees deployed.

That could prove more attractive than simply selling software licences or charging for call-centre staff. However, its profitability will depend on the cost of running AI conversations, customer-specific engineering requirements and how much of the savings customers retain.

For years, investors have generally valued SaaS businesses differently from IT-services companies.

The reason is fairly straightforward. A software business with high recurring revenue can potentially serve many more customers without increasing expenses at the same rate. A services company usually has to employ more people, even if it can improve employee productivity.

Freshworks demonstrates why this distinction matters.

In 2025, the Indian-founded SaaS company generated $838.8 million in total revenue. Of this, professional services contributed just $9.4 million, approximately 1.1%. Its gross margin was 85%.

These figures, disclosed in its US regulatory filings, show that enterprise software can still operate without becoming heavily dependent on services revenue.

Fractal's 47% gross margin provides a useful contrast, although the two companies have different business models and accounting structures. Fractal's combination of products and services cannot be valued simply by applying the same assumptions used for a predominantly subscription-based business.

This becomes important as KPO and IT-services companies begin describing themselves as AI product companies.

Imagine two businesses earning ₹100 crore annually. The first generates ₹80 crore from recurring software subscriptions and ₹20 crore from implementation work. The second earns ₹20 crore from software and ₹80 crore from customised projects.

They have identical revenue, but the economics could be very different.

The first company may be able to double its customer base without doubling employee expenses. The second may need to keep hiring engineers, even if its proprietary software helps them work faster.

Their ability to retain customers, expand margins and generate cash would therefore differ. Investors should not value them identically merely because both possess AI platforms.

There is another complication. Proprietary software requires continuous investment. Companies must maintain integrations, improve security, fund research and pay for computing infrastructure. AI tools also face the risk that competing technology becomes cheaper or more capable.

Simply automating a service does not create a highly profitable software business. The company must be able to sell its technology repeatedly, retain control of its intellectual property and deliver additional customer work at a progressively lower cost.

For private companies such as Evalueserve, Quantiphi and MathCo, the most useful disclosures would be the percentage of revenue generated from software subscriptions, the cost of implementing each new customer and revenue growth relative to employee growth.

Without these numbers, claims about becoming a product company remain difficult to evaluate.

India's outsourcing industry has long offered a fairly predictable career path for graduates.

Freshers joined as analysts, software developers, testers or support executives. They handled repetitive tasks, learned through experience and eventually moved into specialised or managerial roles.

This system was also central to the industry's growth. Companies could recruit large batches of graduates, train them and deploy them across customer projects.

AI is putting pressure on the work that sustained many entry-level positions.

In a KPO, preparing the first version of a financial model or collecting competitor information may no longer require the same number of junior analysts. In IT services, routine coding and testing can increasingly be assisted by software.

That doesn't mean every affected position will disappear. Companies may handle greater volumes of work, enter new markets or create jobs that did not previously exist.

But the skills employers need are changing.

A financial analyst who can only collect information and prepare standard reports faces a more difficult future than somebody who understands credit risk, can challenge financial assumptions and knows when an AI-generated conclusion is wrong.

Similarly, software developers who understand business operations, customer data and enterprise integration may become more valuable than employees focused entirely on repetitive coding.

Hiring patterns already show some movement. According to Indeed data reported in October 2026, Indian job postings for Forward Deployed Engineers increased 19-fold between June 2025 and June 2026. The increase is from an unspecified base, so it does not tell us how many jobs exist, but it indicates growing demand for technical employees who work directly with customers.

The difficult question is how companies will develop experienced specialists if they require fewer beginners.

Traditional KPO and IT-services businesses trained people through junior-level work. If machines perform much of that work, companies will need different ways to develop the judgment and experience they expect from senior employees.

This may become one of the industry's less visible problems over the next few years.

The convergence of software and services creates an unusual competitive situation.

Evalueserve may compete with an IT-services company to implement a banking AI system. Quantiphi may compete with a SaaS startup that sells similar software. Leena AI may compete with a systems integrator offering to build a customised employee support application.

The customer may no longer care whether the supplier calls itself a KPO, SaaS startup or IT-services company. It wants reliable results at a reasonable price.

That changes the importance of traditional competitive advantages.

A large workforce, which once helped outsourcing companies win substantial contracts, may become less attractive if a smaller competitor can deliver the same work with fewer employees and better software.

At the same time, owning an AI platform may not provide much protection if competitors can build similar products using widely available models and cloud infrastructure.

The more defensible advantage could come from combining software with specialised industry knowledge, proprietary data and years of experience handling complicated customer operations.

Evalueserve's understanding of financial research may help it build better credit and investment tools. Fractal's experience working with large companies may give it an advantage in deploying complex analytical systems. Leena AI's experience integrating with enterprise applications may make its software more useful than a generic assistant.

But none of these advantages is permanent. Customers will keep comparing suppliers, technology will improve and prices will remain under pressure.

This is why the next stage of India's outsourcing industry may be less about hiring thousands of employees and more about how much valuable work those employees can perform with technology.

The winners won't necessarily be the companies with the largest workforces or the most impressive AI demonstrations. They will be the ones that can make their customers more productive while retaining enough of the resulting savings to grow their own businesses.

For KPOs, this is particularly urgent. They spent two decades convincing global companies that research could be done more cheaply in India. Now AI is encouraging those same customers to ask whether some of that research needs to be outsourced at all.

Evalueserve and its competitors are trying to answer that question by building software, expanding into technology services and using their existing industry knowledge in new ways. SaaS companies, meanwhile, are discovering that technology alone may not be sufficient to solve complicated enterprise problems.

India's software and outsourcing companies may eventually look more alike, but their financial success will still depend on something quite basic: how much customers are willing to pay, how much the work costs to deliver and whether revenue can grow without expenses growing just as quickly.

The old Indian outsourcing model was built on doing the same work with cheaper people. The next one will have to demonstrate that Indian companies can do more valuable work with fewer people, while persuading customers that the outcome is still worth paying for.