Why forward deployed AI engineering is the new capability in Indian financial services

There is a pattern emerging across financial institutions over the past two years. A bank or NBFC announces an AI initiative—fraud detection, credit underwriting, customer service automation. Pilots run. Results look promising. And then, somewhere between the proof-of-concept and enterprise-wide deployment, the momentum stalls. 

For instance, a global retail bank recently spent 18 months and a meaningful budget rolling out a generative AI assistant for its contact centre. The model was capable. The vendor was credible. But adoption stalled at under 15% because the assistant had no access to the agent’s actual case context, could not write back into the CRM, and routinely surfaced answers that conflicted with the latest product policy. None of these were modelling problems. They were deployment problems and they could not be fixed from outside the business. 

The gap between AI adoption and AI operationalisation is one of the defining challenges in financial services right now. What is less discussed is why that gap exists and what the emerging answer looks like. 

A potential solution is: forward deployed engineer. 

What is Forward Deployed AI Engineering? 

The concept was pioneered by Palantir, whose entire enterprise go-to-market model is built around embedding senior engineers inside client organisations—not to manage a project, but to make the technology work in the messy reality of that client’s systems, workflows, and constraints. 

A forward deployed engineer is not a data scientist with a different title. The role exists at the intersection of engineering, product, domain, and controls. 

In practice, the work looks like this. The engineer sits inside the business close to the underwriters, the fraud analysts, the AML investigators, the wealth advisors, the servicing supervisors. They learn how decisions are made, including the workarounds and the unwritten rules. They build, test, and iterate the AI system inside that real environment. They take responsibility for performance, drift, controls, and adoption. And critically, they transfer capability so the institution can keep building after they are gone. 

The model is now being explicitly replicated by the world’s leading AI labs. 

That is a fundamentally different engagement model from what the enterprise software industry has operated on for decades. Hiring interest in forward deployed engineer roles has grown 800% since January 2025, according to the Financial Times. 

Three distinct plays—at very different stages of readiness 

When we look at Indian financial services through this lens, three distinct sets of actors emerge—domestic banks and NBFCs, consulting firms, and global capability centres—each at a different stage of recognising and responding to what forward deployed AI engineering actually demands. 

Consulting firms are closer to the forward deployed model—their best senior practitioners on large financial services transformation programmes are effectively doing embedded deployment work already. The distinction is whether that is being deliberately structured, priced, and governed as a capability transfer engagement, or whether it is being delivered as traditional project-based consulting with AI bolted on. The difference matters because the goal of forward deployed engineering is not to complete a project and exit. It is to leave the institution with the capability to continue building. 

Why this shift matters disproportionately for GCCs 

Most GCCs in financial services are already, in practice, doing fragments of forward deployed work. They build, run, and govern critical systems for their parent institutions. They sit closer to production reality than almost any other part of the technology supply chain. But the role is rarely formalised, priced, or governed as a forward deployed engagement. It is structured as delivery. 

A GCC that organises itself around forward deployed AI engineering will have a fundamentally different conversation with its parent. Instead of being measured on cost arbitrage and ticket closure, it will be measured on production AI outcomes—risk reduction, decision speed, loss avoidance, advisor productivity, customer resolution. That is a different kind of mandate, and it commands a different kind of investment. 

The leaders who get this transition right will need to do three things deliberately. They will need to define forward deployed engineering as a distinct role, with a career path that rewards production ownership rather than technical depth alone. They will need to change the engagement model with the parent institution from project-based delivery to capability-based partnership. And they will need to build the supporting muscle—model risk, controls, MLOps, LLMOps, observability, and domain fluency—as first-class disciplines, not adjuncts. 

The capability India needs to build deliberately 

What makes a forward deployed AI engineer effective in financial services is a specific combination of skills that Indian engineering talent is well-positioned to develop. 

Indian financial services have the scale, the infrastructure, and the talent to be at the leading edge of that transition. The question is whether the institutions, the firms, and the centres that make up this ecosystem will organise themselves deliberately enough to get there. 

The engineer needs to be in the building. The more important question is whether the building is ready for them. 

(The author is Partner and Financial Services Technology Consulting Leader, EY India. Views are personal)