Indian IT’s FDE Bet Will Not Be Proven by Headcount

Indian IT can train and retitle thousands of people inside a year. That has never been the hard part. The hard part is proving the new team costs less to run.

FDE Work Runs on the Scarcest Part of India’s Talent Pyramid

India’s services industry was built on a shape. Large graduate cohorts entered at the base, learned by doing repetitive build-and-test work, and moved up under smaller groups of architects and managers. Infosys still plans about 20,000 campus hires this fiscal year. The shape is the business model. That is also what separates Indian IT from a Western consultancy that mostly hires mid-career professionals.

FDE work values the opposite end. It needs engineers who can ship production software, read a P&L, sit with a client’s operations head and decide without three layers of escalation. Those people are scarce everywhere. In India, they are scarce within firms whose training systems were designed to produce something else.

The US-based research and advisory firm, the Technology and Services Industry Association (TSIA), identified the same shortage in April. Its research traces FDE from bespoke work by a few specialists toward a repeatable capability. Along the way, it finds too few people who combine technical depth with customer-facing work. TeamLease reports that large Indian firms are now retraining cloud, platform, and solution engineers because the external talent pool is too small to draw from. Tredence expects about half its FDE pool to come from inside.

Mohta expects the base of the pyramid to thin as repetitive coding, testing and implementation are automated. His worry is not only the job count. Those tasks were how large graduate cohorts acquired judgment, and nothing has replaced them.

Maya Nair, executive director at Grafton Recruitment, expects the adjustment to show up in skill mix rather than headline headcount. Demand will sharpen for people who possess both technical and commercial knowledge. She locates the model’s appeal in the same place. Put senior expertise directly into the execution team and remove the layers in between. Decisions come faster, and ownership is clearer.

Korla expects conventional engineering roles to survive regardless. FDE-style teams may move deployments forward, he said, but somebody still has to make the resulting systems secure, scalable and reliable.

None of this would matter much if pricing held. It is not holding. HCLTech puts AI-related deflation across its portfolio at 2%-3%. Infosys narrowed its full-year guidance in July and flagged weaker-than-expected pricing gains. Its AI-first work reached 8.2% of revenue in the June quarter. Clients have learned to ask for a share of the productivity, and they are asking now.

This is where the FDE case gets awkward. Its selling point is that fewer, more expensive people do the work of many cheaper ones. If the price falls in proportion, the firm has bought a costlier workforce and kept none of the gain.

Reuse and Repricing Decide Whether the Bet Pays

Two things decide this. Headcount is not one of them.

The first is whether an engagement leaves behind something the next engagement can use. Gartner made this the central argument of a June report. Assets created by FDEs must feed back into the product, so work for one customer reduces the effort required for the next. Kim and Hwang’s definition rests on the same loop. Without it, an FDE practice is a consulting bench with a new name and a higher wage bill.

The second is whether the contract allows the provider to retain some of the value created by those productivity gains. TSIA argued in March that traditional measures such as utilization work against FDE teams. Pricing, it said, must shift from hours logged to business outcomes.

Put those two factors together and four business models emerge. When little work is reused, and the provider captures little of the value created, FDE becomes consulting under another name: custom projects billed by the hour and delivered by expensive specialists. It is the costliest model to operate. When reuse is high, but value capture is low, the provider creates the leverage while the client keeps most of the savings. When reuse is low, but value capture is high, scarce expertise commands a premium, but the number of available specialists constrains expansion.

Only the fourth model offers durable scale: reusable assets paired with outcome-linked pricing. Call it the deployment dividend.

IBM has put a number on that fourth position. Its Forward Deployed Units, announced on May 14, are pods rather than individuals. Business specialists, architects and engineers work alongside AI agents that handle coding, testing, evaluation and documentation. The pods draw on reusable assets from IBM’s consulting platform. IBM says a six-person unit does work that previously took 30 people, and names Riyadh Air, Nestlé, Heineken and Pearson among its customers. The productivity figure is IBM’s own.

Tredence is aiming at the same position from a smaller base. It is building around compact multidisciplinary teams, reusable intellectual property and reference architectures, not a headcount expansion. Abhijeet Kate, co-founder and principal consultant at the Salesforce consultancy digiCloud Solutions, measures outcomes directly. He tracks new revenue, operating efficiency and margin after AI costs. On some complex engagements, his firm has seen efficiency gains of 20% to 30%, though those are its own numbers.

Indian providers are chasing that position while their suppliers move in the opposite direction. OpenAI launched a majority-owned Deployment Company on May 11, backed by $4 billion of initial investment. It also agreed to acquire Tomoro, which brought about 150 experienced FDEs in at launch. AWS committed $1 billion to forward-deployed engineering and, on June 30, extended the model to consulting partners. Salesforce chose partners outright: its FDE Partner Network, launched April 15, includes more than 30 firms, including TCS and Cognizant.

Kate does not read that as displacement. “Enterprises increasingly use several AI models alongside old applications, cloud platforms and proprietary systems,” he said. That leaves integration, workflow redesign and change management to the services firms. Korla makes the same case, adding governance, security, compliance and legacy modernization to the list.

Krithivasan is betting on exactly that. Deep knowledge of the customer’s environment is what differentiates TCS, he told Reuters, and it has nothing to do with cost arbitrage.

Implications by Role

C-suite. Stop reporting FDE progress primarily through headcount. Two measures matter more: the share of each engagement’s output reused in subsequent work, and the share of revenue linked to outcomes rather than effort. Establish a baseline for both this quarter and report the results to the board by the end of the fiscal year. A firm that cannot report either is running professional services with AI-level salaries.

Functional leaders. Delivery and HR should plan the talent pyramid at least 2 years in advance. Automation is removing much of the entry-level work through which today’s architects learned their trade, yet few firms have built an apprenticeship to replace it. Sales and finance should also reconsider utilization targets, which can penalize pre-sales and discovery work that successful FDE deployments depend on. Pilot at least one outcome-linked contract before committing to train the next thousand engineers.

Boards and governance. Ask for a comparison that published research has yet to provide: deployment time, cost and business outcomes for FDE-staffed projects versus conventionally staffed projects within the company’s own portfolio. Ask what happens to gross margins if clients retain the entire productivity gain. Then establish who owns the reusable assets when a partner-led program involving AWS, Salesforce or a model provider ends.

India’s IT firms may well hit their FDE targets. Training and retitling people at scale is something the industry knows how to do. But a target met is not an advantage won. Announced totals such as 5,900, 6,000, 1,000 or 200 demonstrate scale, not a working business model. The real proof lies in two places the industry has so far said little about: what each engagement leaves behind and who keeps the money it saves.