AI capability is soaring, but enterprise value lags. How do leaders convert technical brilliance into sustainable business returns?

Dr Sanjay Sarma

Professor of Mechanical Engineering | Massachusetts Institute of Technology

There is very little doubt that AI is incredible. Modern large language models rest on spectacular advances in backpropagation, accelerated computation, vector databases and neural networks whose roots reach back more than a century. Yet AI also has a habit of producing inflated expectations and subsequent winters. I like to say that artificial intelligence has taken “84 years to become an overnight success.” The achievement is real, but so is our tendency to confuse technical capability with immediate practical value.

One of the least appreciated breakthroughs is the vector database: the nuance of human language can be represented as vectors that obey mathematical operations. Specialized hardware is advancing too, including chips optimized for language models and the prospect of “frozen” chips with a model built into them for efficiency. These innovations explain why capability is moving so quickly. They do not, however, remove the organizational work. Every application still needs a defined purpose, reliable data, boundaries for decision-making and a response when the system is wrong. Technical brilliance increases the opportunity, but it also raises the standard of institutional responsibility.

That distinction is central to this book. You cannot take a magic technology and expect magic to occur; a technology must be absorbed. To incorporate AI into a workflow, you have to redesign the workflow around what AI can do, especially its ability to process unstructured data, and around what it can do badly, including hallucinating or making consequential mistakes. An event chatbot might answer a participant instantly, but who is responsible if it gives the wrong time or mishandles a dietary question? Unless responsibility, oversight and operating processes are redesigned, AI becomes unusable, expensive and potentially a liability.

I saw a simple illustration in Arizona when a Waymo vehicle appeared to break a traffic rule. How do you give a ticket to an autonomous car? Will it recognize a police siren and stop? If a ticket is placed under its wiper, who receives it? This is not merely a technical puzzle. It shows why companies, public policy and society must be rewired around new forms of agency. Otherwise, AI remains something we admire without extracting value from it.

The tension is visible. Data centres are being built in anticipation of demand, while companies know they must use AI but have not worked out where, why or how quickly. That gap fuels anxiety about costs and a possible bubble. The burden is the responsibility enterprises must accept to convert capability into results. I am convinced AI will soon shape traffic routing, registration, seating and event interfaces. It has not happened fully yet, which makes this an exciting moment to participate in the change. Satish’s book can launch the deeper thinking needed to make that participation worthwhile.

Mr Satish Viswanathan

Author, Advisor & Executive | Reinvigorating Enterprises with Data & AI, Generative AI

We are living between the promise of singularity and a practical boardroom question: when a CEO announces one hundred AI agents, which agent is responsible for the ROI? That gap is the essence of my book. I began with a white paper, drawing on years across HP, IBM and Accenture, but the field questions kept expanding. The result became a 500-page reference to help leaders locate themselves in a fast-moving AI economy.

Natural language has become the interface through which people engage algorithms, yet model capability is far ahead of enterprise value. I call this the “boardroom squeeze.” AI is fluid, while an enterprise is rigid with legacy systems, data, processes, incentives and technical debt. “Enterprise metabolization” makes those realities function together. It involves producers who build models, implementers who create services, consumers who apply them, and shapers such as academics, regulators and benchmark institutions who define success and responsible use.

Leaders need a more precise vocabulary than AI use case. They must assess what AI can and cannot do, map adoption complexity, choose partnerships and understand how value compounds across connected processes. Architecture matters, but human acceptance matters just as much. If the last employee or manager does not change how work is performed, investment will not deliver transformation. Measurement, governance and incentives must change too. We also need credible ways to calculate cost, consumption, risk and return.

The book therefore moves from diagnosis to practical architecture. Its frameworks help an enterprise classify adoption complexity, plot capability against consumption, and examine how cognition compounds across connected work rather than isolated pilots. I also wanted to address the technology stack, because architecture cannot be separated from governance, human adoption or economics. During my time at Accenture, I worked with MIT on an AI ROI model that could travel across industries and activities. The question remains fundamental: leaders must understand not only whether a model performs, but whether the operating system around it creates sustainable value at an acceptable cost.

I am less interested in chasing artificial general intelligence than in building artificial enterprise intelligence. There is no single roadmap already cracked by one company or one country; we will learn through repeated implementation. The United States has become a model capital and Europe a compliance capital. India’s opportunity is different. Over several decades, we have become the process capital of the world, with deep knowledge across industries and functions. We can now become the process ROI capital of the world: the place where AI returns are made real, repeatable and scalable.

That ambition includes education. A curriculum revised every two or three years cannot keep pace. We need faster learning systems. Lasting advantage will come from absorbing AI, redesigning work, developing people and turning intelligence into measurable enterprise value.

Mr Kapil Viswanathan

Member, Governing Council | Krea University

I approached this discussion with some nervousness because I do not claim to be an AI expert. The book clarified that its deeper subject is the enterprise and its capacity to adapt. Its layered frameworks connect technology, organization and execution. The question I keep returning to is concrete: for every dollar invested in AI, what comes back, and why? A serious discussion must connect transformation with practitioners working through adoption, process redesign, governance and measurable return.

Dr Chandramouliswaran V, PhD

Vice President – Artificial Intelligence | PayPal

I value this book because it captures the complexity an organization actually experiences. Our early approach was to ask each business unit for its leading use cases and give employees access to tools such as ChatGPT, Claude and Perplexity. That democratized learning and produced useful experimentation, along with duplication. It was a necessary starting point because a horizontal AI team does not automatically understand the depth of accounting, sales or every operating domain.

Coding and knowledge assistance are approachable because they are text-rich. Reconciliation across thousands of statements and hundreds of markets is harder. We must decide where AI assists people and where controls remain essential. Retrofitting AI into one process fragment is easy. Greater value comes from asking, “If I designed this process from scratch with AI available, what would it become?” That requires AI expertise, domain expertise and process ownership working together at enterprise scale.

Mr Lakshmi Narayanan

Chancellor; Chairperson, Governing Council | Krea University

The full benefit of AI requires us to reimagine a process end to end. Adding a model to one point in a value chain may produce a limited gain, but transformation demands design thinking, just as business process re-engineering was essential in the early days of computerization. AI is not merely another tool fitted into an old workflow; as agency increases, systems may make decisions within the authority, data and controls we define. We therefore have to choose the right model for the process rather than assume the largest model is always necessary.

Education must change with this reality. AI-native companies build products and workflows from the ground up and reach global audiences through capabilities such as translation. Students need creativity, critical thinking, logical reasoning, collaboration and just-in-time learning. Some will apply existing models; others will conduct deep research for Indian use cases. Universities must connect both groups so that ideas, research and enterprise capability grow together.

Mr Suresh Kuppuswamy

CEO, MD & Co-Founder | Zepto Logic Technologies

I am still learning because the scale of change is difficult to comprehend. Start-ups can adopt AI quickly, but established enterprises carry balance-sheet constraints, talent plans and complex architectures. They cannot stop recruiting and assume AI will run the organization. They must build foundational skills so tomorrow’s architects understand the entire lifecycle. The shift will take years as costs evolve. This is the time to begin with focused use cases, learn what improves design or industry processes, and build a framework for broader adoption instead of prematurely declaring an entire sector disrupted.

Dr Lakshmanan Nataraj

AI Researcher & Computer Vision Expert

AI governance cannot be defined only by the companies releasing frontier models. Their policies do not automatically translate to the needs of every enterprise or country. In India, governance must account for cultural and linguistic context, bias, safety and the realities of local adoption. The book’s category of “shapers” is useful here. Academics, regulators, AI safety institutes and responsible-AI centres must study outcomes, develop relevant cases and feed evidence back into model and policy design. Governance should evolve with adoption, identify where improvement is needed and make responsibility practical rather than merely declarative.