AI that Works: From Pilot Projects to Enterprise-Wide Payoffs

For today’s business leaders, AI is no longer a moonshot or a placeholder, it’s a mandate. But turning AI ambition into real impact is a classic case of 'strategy meeting execution'. It’s easier said than done, and measurable value still remains a challenge. There are many moving parts to achieving AI success. From aligning with strategic goals to building scalable data foundations and fostering the right culture, the road to enterprise-wide impact is layered and complex.

In this exclusive conversation with CiOL, Bharani Subramaniam, CTO, India & Middle East at Thoughtworks, deep dives into how the company is helping CxOs move beyond pilots and POCs and activate AI strategies that deliver lasting RoI, operational agility, and competitive edge.

How does Thoughtworks’ unique approach to technology strategy and agile delivery help companies in India maximize RoI on their AI investments?

Thoughtworks supports organisations in becoming intelligence-led enterprises through a combination of AI strategy, governance, and agile implementation. With a focus on scalable AI platforms, modern data architectures, and infrastructure automation, we enable clients to deploy production-ready AI that’s both robust and aligned to business needs.

One key offering is Haiven, an AI-enabled knowledge amplifier and team assistant that accelerates the adoption of Generative AI (GenAI) across the software delivery lifecycle. Haiven allows clients to experiment with and implement AI in a structured, scalable manner, especially within complex domains.

Backing this up is Thoughtworks' AI Research Lab (TAILS), which specialises in cutting-edge research around Large Language Models (LLMs), including areas like interpretability and governance. While engaged in foundational work, the lab’s insights directly shape real-world deployments, helping organisations transition from pilot phases to scaled production through effective model evaluation and governance frameworks.

How is it accelerating real-world AI adoption?

The integration of Haiven and the TAILS research lab into delivery processes exemplifies our approach to bridging innovation and implementation. These aren’t isolated tools but part of a continuous feedback loop that ensures clients benefit from the latest advances in AI while maintaining responsible, production-grade execution. The result is faster time-to-value, reduced deployment risk, and more consistent scaling of AI solutions.

What are the key factors businesses should focus on to ensure their AI initiatives align closely with measurable business outcomes?

Success begins with clarity. One should start with identifying specific business problems that AI can address with tangible, measurable impact. Many companies still struggle to translate AI hype into real-world value. In 2024, 74% of organisations failed to scale value from their AI initiatives, often due to deploying solutions without a clear use case or production roadmap.

To avoid this, AI strategies must align tightly with overarching business goals, be it improving customer satisfaction, reducing operational costs, or unlocking innovation. These goals should be continuously revisited as technology evolves.

Equally important is adopting a “human-at-the-centre” mindset. AI should augment, not replace, human capabilities. Its real power lies in amplifying expertise, decision-making, and creative potential. Moreover, organisations need to invest in cultural readiness - fostering innovation, accountability, and open dialogue. Governance frameworks should support responsible use while ensuring teams feel empowered to experiment and iterate.

In the evolving equation between humans and machines, how important is cultural readiness, and is it just as critical as technical readiness for successful AI adoption?

Technology alone doesn’t guarantee success. AI needs a fertile environment to thrive, one where experimentation is encouraged, and leadership actively supports transformation. Multidisciplinary teams that combine domain knowledge, technical skill, and design thinking are often the most effective. They help bridge the gap between technical possibility and real-world impact, ensuring AI serves actual business and user needs.

If you look at markets like India, what common pitfalls do companies face when trying to onboard or scale AI solutions, and how can they overcome them?

One of the most common missteps is accumulating data indiscriminately, under the false assumption that more is always better. Without purpose and governance, this leads to clutter, cost, and poor-quality insights. A better approach is to treat data as a product, with clear ownership and value drivers, enabling faster and more sustainable adoption.

Another pitfall is chasing innovation for its own sake. Adopting the latest technologies simply to stay ahead of the trend curve is not the right strategy. This can lead to rushed deployments, lack of ethical clarity, and a disconnect from real business objectives. AI should always be used to augment human potential, not as a gimmick or quick fix.

Overconfidence is also a risk. Assuming AI models are flawless or ignoring external threats can create dangerous gaps between strategy and execution. Grounding every AI initiative in real-world execution, with constant feedback, is essential.

On the flip side, some companies move too slowly. They are tied down by legacy systems or a culture that resists change. Balancing compliance with a spirit of calculated experimentation is key to unlocking innovation without compromising stability.

How can enterprises develop reliable frameworks to track and measure the RoI of diverse AI applications beyond immediate cost savings?

Organisations that lead in AI adoption are more than three times as likely to see tangible business gains, according to Thoughtworks’ recent State of Digital and AI Readiness Report. To capture these gains, enterprises must align AI investments with strategic objectives and use iterative methodologies that factor in user needs, journey mapping, and prioritised outcomes.

Measuring RoI requires clear, actionable roadmaps, ones that link AI efforts directly to outcomes like improved revenue, reduced churn, or sharper decision-making. These must include both hard metrics and soft indicators such as customer experience and operational agility.

What are the foundational steps needed to operationalise RoI measurement at scale?

Enterprises should begin by bridging the gap between experimentation and scaled deployment. This involves aligning projects with core goals, addressing skill gaps, and selecting high-impact use cases. Early pilots help establish benchmark metrics, which can be refined as the initiative scales.

Equally critical is modernising the data foundation. Treating enterprise data as a product - complete with ownership and usability- turns information into an asset that fuels strategic AI.

Reliable measurement also depends on the right tools. Real-time dashboards, analytics, and structured frameworks such as the AI Driven Business Value model enable continuous tracking of both operational and customer-centric metrics. A self-funding model, where savings from modernisation are reinvested into innovation, can further sustain momentum while reducing technical debt.

Looking at emerging AI trends, where do you believe AI will have the biggest impact on business performance over the next three to five years?

We are entering an era of AI-first software development, where AI becomes a fundamental mindset, not just a tool. Intelligent assistants are evolving rapidly, It's no longer limited to autocomplete, they are beginning to refactor code, understand entire systems, and handle complex tasks. This will reshape how software is built and accelerate innovation across industries.

At the same time, AI is becoming the digital nervous system of the enterprise. It is transforming operations, from logistics to risk management, by enabling smarter, faster decisions. This shift requires new governance frameworks to manage AI as part of core infrastructure, not peripheral experimentation.

Customer experience will also be significantly reshaped by AI. Hyper-personalisation at scale is fast becoming the norm, and companies that fail to meet this standard risk falling behind. Cloud-native AI, meanwhile, is becoming an expectation rather than a luxury, and with it comes the need for robust observability. As models grow in complexity, tools that detect drift, hallucinations, and prompt failures will be critical for maintaining trust and performance.

Finally I would say, foundational investments in data and workforce reskilling will determine the long-term success of AI efforts. With techniques like Retrieval-Augmented Generation improving contextual relevance, and with a growing need to treat data as a product, enterprises must simultaneously upskill their people and upgrade their infrastructure. Organisations that manage to do both will not only survive, but lead.