Hang Ten Systems CTO Believes AI Cannot Replace Some Indian IT Assets
By Ajay Rag ·
The billing model that has powered India's IT services industry for three decades, charging clients by the number of people staffed on a project, is running out of road as AI collapses the link between team size and software output, believes Navin Budhiraja, Chief Technology Officer of AI-native enterprise services firm Hang Ten Systems.
In an interview with AIM, Budhiraja says the shift is structural, not cyclical, and will force even the largest services providers to move away from large, multi-year implementation teams toward smaller teams delivering outcomes in weeks or months.
"The model that is based on the number of people that I needed to create software—I think that is going to be difficult to continue," he says. "The business model has to evolve, where it's no longer about large teams and multi-year projects. It has to be small teams delivering outcomes in a matter of months."
Barely three months old, Hang Ten Systems, founded by former Infosys CEO Vishal Sikka, has positioned itself as an "AI-native" enterprise services provider, focused on building, modernising, and running enterprise software for large clients using large language models paired with an in-house agentic delivery platform, rather than the large offshore implementation teams that have defined Indian IT services for decades.
The Palo Alto-based company has raised roughly $32 million, led by Mayfield with strategic investment from Aramco Ventures and participation from angel investors. It counts German healthcare firm Fresenius and Siemens Energy among its early clients.
Its work already spans SAP migrations, finance and HR transformation, and software engineering—segments long dominated by tier-1 IT majors such as TCS, Infosys and Wipro.
Separate Conversations
According to Budhiraja, people routinely conflate the pace of AI-driven change in how software gets built with its effect on jobs. "I think of these as two separate things."
First, the act of writing software has been decoupled from the size of the team writing it. "Something that may have taken a team of 10 people to do can be done by one or two people," he says.
That is exactly why the per-head billing model cannot hold. Budhiraja believes businesses built on multi-year, large-team contracts have to give way to short engagements built around outcomes, delivered by much smaller teams.
However, regarding jobs, Budhiraja pushes back on the idea that roles are simply disappearing. "It's not that the job changes. It is still a software job, but the task that you do as part of that may evolve," he notes.
The developer's role, in his view, is shifting from writing code to exercising judgement, understanding domain context, and reviewing what AI produces—the kind of work he believes AI will not replicate at the level of someone embedded in a business.
More Software, Not Less Demand
On whether AI will shrink the market, Budhiraja invokes the Jevons paradox, the idea that falling costs per unit can expand overall demand, to argue that cheaper, faster software production will make custom software economically viable for companies that previously couldn't justify the expense.
He describes this as an opportunity to deliver "bespoke SaaS"—tailor-made enterprise software at something closer to the economics of commercial, off-the-shelf platforms.
That, he says, also opens a large opportunity to modernise decades-old legacy systems that companies have long been reluctant to touch because rebuilding them was too expensive or risky.
Forward Deployed Engineers
On the growing trend of ‘forward-deployed engineers’, a model popularised by Palantir and now being explored by both OpenAI's deployment arm and Anthropic, alongside Indian IT majors, Budhiraja expects it to coexist with, not replace, offshore and nearshore delivery.
Early-stage, ambiguous problems benefit from engineers sitting with the customer, he says, while more mature, better-understood workstreams can move to distributed teams. "It's going to be a clear mix," he notes.
On enterprises' anxiety over AI compute costs, Budhiraja says the fixation on token pricing misses the point that not every problem needs a frontier model. Hang Ten Systems, he says, deliberately routes work to smaller, open-weight models where cost, data privacy, or sovereignty concerns make sense, reserving frontier models for harder problems.
He also pushes back on the narrative that enterprise AI pilots are broadly failing to reach production, framing it instead as an expectations mismatch.
"It's not just about the model—it's about the ability to get the right data to the model, the ability to build the right guardrails," he notes. Much of the "rest of the stack" like data pipelines, governance, security, applications hasn't kept pace with how fast the underlying models improved, he says, and enterprises often signed up expecting gains the tooling around the model wasn't yet built to deliver.
On what large Indian IT services company should be doing today, Budhiraja points to three shifts: blending forward-deployed and distributed delivery models, moving sales conversations away from multi-year contracts toward shorter, outcome-based engagements, and leaning on the deep client trust and domain expertise that established players have built over decades—an asset he said AI cannot replicate.