In an exclusive interaction with Samrat Pradhan, Managing Editor at Finance Outlook India, Kaushik Sarkar, Senior Vice President, CFO and Board Member at Bosch Global Software Technologies, shares how AI is reshaping the economics of R&D in software and GCC-led businesses. Drawing on nearly three decades of experience across finance and business leadership, he explains why the real focus must now move from mere scale to disciplined value creation through smarter build-versus-buy choices, stronger governance, and AI-led productivity.

Kaushik Sarkar highlights Indian GCC’s story is not all about the cost arbitrage anymore, it is about building distinction for the future-proofing of the global organization with a better understanding of core competency, deep engineering, location choice and “buy commodity, build distinction.” He also outlines how Bosch is leveraging its own proprietary mobility developer platform and AI-driven workflows to boost productivity, enhance decision making and build a competitive edge going forward. He reinstates that today, a GCC is no longer running on the cost savings methodology, it is measured by the impact it brings in.

With AI-led development and cloud-native architectures rapidly reshaping software engineering, how are Indian firms rethinking R&D cost structures while ensuring cost predictability and sustained innovation velocity?

AI-led development and cloud-native architecture have fundamentally changed the shape of software R&D economics. Earlier, most leaders could think of software R&D primarily as a talent-cost equation. Today, the cost structure is far more layered—talent, cloud consumption, model usage, platform engineering, observability, cybersecurity, data pipelines, and compliance all sit within the same economic framework.

Cloud-native operating models have driven a significant change in how technology investments are made, from mostly capital spend to a much more variable operating expenditure (OPEX). AI has added yet another layer of uncertainty on top of that with the usage of GPUs, inference costs, token usage and burst compute needs. Today, the key to predictability is not freezing budgets for long, but instrumenting them with a much greater degree of precision. This reflects a broader industry shift toward formal AI FinOps practices and real-time cloud governance.

The shift we are witnessing in India is from headcount-based budgets towards product/ platform/ workload economics. Leading engineering companies have begun to differentiate between investments that support the foundations of their dev teams, like developer platforms, shared data architecture and test automation, DevSecOps tooling, and reusable services, and their spend on experimentation. This protects innovation velocity because teams are no longer rebuilding the same plumbing repeatedly.

At Bosch Global Software Technologies (BGSW), for instance, we have institutionalized this through our proprietary mobility developer platform, our in-house AI-enabled engineering platform that helps developers code faster, test better, and reduce repetitive effort. This platform is helping us drive 15-20% cost savings, while ensuring engineering throughput remains high.

Ultimately, finance and engineering must now operate together and be responsible for metrics such as cost per release, cost per workflow, defect rates, and reuse ratios. It's not that the AI era makes R&D cheaper, it simply makes costs more variable, so governance needs to become more real-time and intelligent.

 

 

As the build-versus-buy debate intensifies with mature SaaS ecosystems, how are companies evaluating total cost of ownership and long-term R&D ROI before allocating capital?

The build vs. buy argument is not always a simple one. Too many companies think that they've done the math when they compare a first year SaaS license fee to direct payroll costs for the team involved in deploying the solution. They have not.

There is a minimum of 9 layers of cost to consider for a serious TCO assessment: Infrastructure, Operations, License, Migration, Integration, Security, Compliance, Change Management, Switching/Exit costs. In addition to these, businesses need to deal with the subtle problem of complexity build-up, maintenance overhead and technical debt, that snowballs over time. Research on the industry consistently demonstrates that the indirect technology costs can significantly affect the actual cost/benefit of the product throughout its life cycle.

With AI, ROI calculations have become even more complicated. The benefits in terms of efficiency are becoming apparent in repetitive coding scenarios and in content automations, but there are reliability and governance concerns that must be carefully considered in mission-critical or safety-critical environments.

Across my years in enterprise software, one principle has remained constant: the most expensive code is often not what you write on day one, but what you are forced to maintain for years without differentiation.

My rule is simple: “Buy commodity, build distinction.”

Commodity capabilities-horizontal infrastructure and mature non-differentiating services-should generally be bought. Strategic differentiators tied to intellectual property, customer experience, data advantage, control, security posture, or margin expansion should be built.

This discipline is even more important today in the age of AI as consumption-based pricing models ask if innovation is leading to long-term value, and not just momentum during initial release.

Also Read: Navigating Inflation and Financial Uncertainties

With GCCs driving up engineering costs, how are you optimizing R&D spend through workforce structuring, offshoring strategies, or productivity-linked cost controls?

The days of thinking of GCCs as being a labor arbitrage destination are over. As the world moves towards digital transformation, GCCs are expected to become innovation orchestrators, leading digital transformation, developing IP and fulfilling strategic engineering requirements.

At the same time, salary environments remain firm, with average wage inflation continuing across engineering talent pools. The answer cannot be indiscriminate cost-cutting or endless hiring. It must be smarter workforce design.

In practice, this means optimizing R&D through a deliberate mix of:

• Pyramid structuring, maintaining a compact senior layer while building strong mid-level depth
• Global delivery diversification across India, Vietnam, and Mexico
• Platformization
• Sharper productivity measurement, with annual utilization levels exceeding 92%

Most importantly, productivity cannot be measured merely by hours logged or headcount deployed. As someone responsible across finance, operations, strategy, and governance functions, I evaluate performance through quality, value creation, and business impact.

This keeps teams lean while ensuring everyone clearly understands their contribution to enterprise outcomes. India is no longer simply a cost-center story in software, it is increasingly a value-creation story.

As AI experimentation cycles become longer and more resource-intensive, how are finance teams managing uncertainty in R&D outcomes while ensuring capital efficiency and return visibility?

AI experimentation requires finance teams to evolve from budget gatekeepers into disciplined capital allocators. They need to be in tune with today’s world where they too need to have a risk appetite.

Enterprise AI adoption is now becoming widespread, which tells us experimentation is necessary, but blind scaling is dangerous.

The right model is staged funding.

• The first tranche should buy learning-data readiness, model accuracy, feasibility validation, and governance design.
• The second tranche should buy evidence-workflow improvement, cost efficiency, user adoption, and operating stability.
• The third tranche should fund scale, but only when economics are visible and repeatable.

This staged discipline prevents organizations from mistaking experimentation activity for commercial progress.

I strongly advocate reusable evaluation frameworks, common guardrails, and shared data infrastructure rather than allowing every team to independently rediscover at controlled cost targets.

Also Read: Capital Raising and Initial Public Offering: A CFO Perspective

With investors pushing for profitability over aggressive growth, how are SaaS companies prioritizing R&D investments without compromising long-term differentiation?

Investor expectations have become significantly more disciplined. Growth is still critical, but efficient growth is now the benchmark. The correct response is not broad-based R&D cuts. That may improve short-term optics, but it weakens strategic resilience.

The stronger approach is to aggressively protect investments that deepen competitive moats:

• Core architecture modernization
• AI-enabled workflow transformation
• Proprietary data capabilities
• Security and compliance infrastructure
• Engineering platforms that accelerate portfolio-wide delivery

At the same time, leaders must be ruthless about eliminating low-adoption features, duplicated tooling, vanity experiments, and unresolved technical debt with no measurable payoff.   Good strategic finance does not trim muscle, it hardens the core that protects the vitals.

The most effective organizations use finance to sharpen ambition, not suppress it, ensuring every R&D rupee contributes to pricing power, retention, expansion, or lower cost-to-serve economics.

Looking ahead, how will evolving pricing models like usage-based and AI-driven monetization influence how R&D investments are planned, justified, and recovered?

Pricing models are about to reshape R&D spending more fundamentally than many companies currently appreciate.

The rise of consumption-based and hybrid monetization models means engineering investment decisions can no longer be separated from pricing design. Software firms are increasingly blending subscription access with usage-linked monetization, particularly for AI services where cost scales dynamically with customer activity. This trend has more than doubled across software companies over the past decade and is reshaping how providers align value delivery with recoverable economics.

This has direct implications for R&D.

Engineering teams can no longer justify AI-heavy investments without simultaneously designing:

• Metering architecture
• Billing visibility
• Usage throttling controls
• Margin guardrails

Finance increasingly asks three design-stage questions:

• What is the unit of value?
• What is the unit of cost?
• How quickly can investment recovery scale with adoption?

In AI, answers may involve tokens consumed, workflows completed, agents invoked, time saved, or measurable business outcomes delivered.

The winning monetization model will likely be hybrid:

Subscription for platform access, usage for scalable value, and premium pricing for high-impact automation.

In this environment, pricing design and R&D design are rapidly becoming the same conversation.

Key Strategic Takeaways

• The AI era does not make R&D cheaper by default; it makes costs more variable.
• Buy commodity, build distinction.
• India is no longer just a cost-center story-it is a value-creation story.
• The first tranche of AI funding should buy learning, not scale.
• Good finance protects the moat; it does not trim the muscle.
• In AI monetization, pricing design and R&D design are becoming the same conversation.
• Drive innovation, without losing track of disciplined execution.