Software company Palantir built its business out of sending engineers to enterprises to set up its proprietary software—the forward-deployed engineering model that is now in high demand. But IT services firm Coforge wants to swim against this current.
“While the whole industry is marching towards the Palantir model, our model is about speed,” Coforge’s Chief AI Commercial Officer Anup Nair tells AIM, as he describes a strategy built around enablement suit, context engineering, and building projects that get cheaper with every iteration.
The pitch arrives as Coforge’s growth outpaces that of larger rivals.
An Enablement Suite, Not a Platform
Coforge is betting on its enterprise autonomy enablement suite, Nuuron. Nair is firm about what it is not. “It is not a platform,” he says.
Coforge chose not to follow the Palantis model because it believes enterprises are already integrating AI across various functions, whether in building agents, orchestration, and preventing data silos.
Coforge works with what a customer already has. Nair’s example is an insurer reimagining its claims process on top of years of backend systems. Coforge brings a methodology, architecture patterns, an execution model, and change management, and adds its own assets only where the client lacks them.
“Nobody wants a proprietary platform, especially in the core, because building software has become very cheap,” he adds.
Ravi Vasantraj, Global Delivery Head at IT services firm Mphasis, has argued along similar lines, saying owning the intellectual property is meaningless next to applying it to a client’s context.
The approach also suits a market where Indian IT firms and frontier labs are partners and rivals at once. Anthropic and OpenAI have each set up deployment businesses of their own.
Context as the Moat
Nair traces the strategy to context. Over the past 18 months to two years, Coforge has built about 120 assets, including EvolveOps.AI and CodeInsightAI, and he says all of them generate or manage context. “People are coming to the realisation now that context is king, but that has always been the charter,” he says.
The work has two parts. Creating context means capturing what sits in people’s heads, since “a lot of context is in the human brain,” Nair says, and having engineers encode it. Operationalising entails deciding which agent gets what context, in what quantity, and in what chunks.
Coforge also encodes industry knowledge into ontologies, adapts them inside a client’s environment, and uses them to fine-tune small language models. It does not sell the ontologies as products. Its AI Launchpad offering creates sovereign environments on open-weight models, and the company recently joined SpaceX’s partner network to build industry-specific Grok applications.
How well this works depends on the data and systems underneath, where fragmented data and legacy applications trip up many enterprises moving agents from pilot to production.
Nair expects context engineering to become an enterprise-level function, since a company cannot run 10 separate programmes. Whoever starts at an enterprise sooner keeps the moat, he says, though he is wary of calling it a first-mover advantage. “People who can pivot are the people who are going to be successful,” he adds.
Training Engineers to Talk to CEOs
Coforge is not trying to build thousands of forward-deployed engineers (FDEs), Nair says. Palantir coined the role, and Indian IT’s largest firms now have FDE targets running into the thousands. Postings in India grew nearly 19x in the year to June, according to Indeed data. A September note from brokerage PL Capital put Momentuum Blue, Coforge’s FDE unit, at about 70–75 engineers, with 150 targeted by March 2028.
Nair describes a rigorous programme. Training runs 12 weeks, with trainees working 80 hours a week and taking a test every week. Those who fail redo the course. In a one-week capstone, trainees build a customer relationship management system.
Coforge trains for three things: technical skills, including context generation and management; discovery, which Nair calls “by far the most important part”; and soft skills, because FDEs have to be able to talk to CEOs.
They work with business transformation consultants hired from the top five consulting firms, who identify the outcomes Coforge signs up for. Assigning the consultants by account, industry, or geography, he adds, helps Coforge build relationships inside an enterprise rather than finish a single project.
Priced by Autonomy
Coforge calls its repeat-project model the compounding flywheel: context and patterns built on one project carry over to the next. In legacy modernisation, Nair says, the cost is falling because Coforge has built context. “We made sure that the fifth project is cheaper than the first project,” he says.
He did not put a figure on the savings or say who keeps it.
Cheaper delivery also puts pressure on revenue across the industry. IT services firm HCLTech has flagged a 2–3% annual deflationary headwind from AI, since automation lets providers do the same work with fewer people.
Coforge has also begun pricing against autonomy. A workflow moves through levels, from L0 (human-assisted) to L3 (completely agent-driven, with human supervision). “You can’t suddenly switch off an existing workflow and make it completely autonomous,” Nair says. A few projects are priced this way so far.
In October, Coforge introduced an autonomous software factory where agents handle coding, testing, and review under human oversight. It has not disclosed customers or quantified gains.
AI-led engineering, data and integration, and cloud services contributed 86% of Coforge’s consolidated revenue in Q2 FY27. It is not a standalone AI revenue figure, and Coforge says it does not provide a breakdown because AI runs through every business unit.
Chief Financial Officer Saurabh Goel has said AI-led efficiencies in existing fixed-price contracts will show in margins from Q4. The uplift will come from those contracts, he has said, not from better pricing on future AI projects.
