The strongest enterprise AI systems will not necessarily be those built around the largest model, feels Kalyan Kolachala, MD, SAIGroup India. Rather they will be the ones that organisations can operate reliably, securely and at scale, he tells in an exclusive conversation with Rajneesh De, Group Editor, APAC Media & CXO Media. AI will certainly become embedded in the software layer, but its larger impact will come from changing the division of work between people and technology.
What makes the current AI wave fundamentally different from previous shifts such as cloud, SaaS or mobile?
Having seen several technology cycles, I would say the biggest difference is that AI is moving beyond being an infrastructure or delivery model and becoming part of how work itself gets done. Cloud changed where technology ran, SaaS changed how businesses consumed it, while AI can influence how decisions are made, processes are executed and outcomes are improved.
What is particularly significant today is the combination of increasingly capable models with enterprise data, workflows and domain expertise. That creates an opportunity to build software that can understand context and take action, rather than simply present information or automate a predefined step.
What separates an enterprise that is genuinely AI-first from one that is simply adding AI features to existing software?
An AI-first company starts with the problem and asks what becomes possible if intelligence is built into the workflow from the beginning. It is not about adding a chatbot or a recommendation feature to an existing product and calling it AI. The real test is whether AI changes the way the product operates and whether that change produces a measurable improvement.
At Symphony AI Group, we look at whether technology can remove friction from complex processes, improve the quality or speed of decisions and create a better experience for the people using it. AI needs to be integral to the product, not an accessory.
Where do you see the biggest opportunity for agentic AI, and where should organisations exercise caution?
The strongest opportunity for agentic AI is in processes that involve multiple steps, decisions and systems. Enterprise workflows such as lending are a good example because they involve gathering information, validating it, interpreting documents and moving work between different stages. Agents can help connect those steps and reduce the amount of manual coordination required. But autonomy should not be confused with removing accountability.
Organisations need clear boundaries around what an agent can decide, what requires approval and how actions are monitored. In areas involving financial, clinical or other high consequence decisions, human oversight remains important. The objective should be controlled autonomy that improves the workflow without compromising trust.
Why can domain-specific or vertical AI deliver greater value than a generic AI model applied across industries?
The value of AI is closely tied to the context in which it is used. A generic model may understand language or patterns, but an enterprise workflow often depends on specialised terminology, regulations, processes and business rules. Domain expertise helps turn a capable model into something useful within that environment. This is particularly important in areas such as healthcare and financial services, where accuracy and context matter considerably.
At Symphony AI Group, our businesses are built around specific problems rather than trying to create a single solution for every industry. That focus allows the technology, workflows and intelligence to be designed around the realities of the people who use them.
What are the biggest barriers holding enterprises back from scaling AI into production?
The challenge is rarely the absence of an interesting AI use case. The harder part is integrating it into an organisation’s existing data, systems and processes. Data quality and accessibility remain important, but architecture, security and governance become equally critical once AI starts influencing real business operations. There is also an organisational challenge: teams need to change established workflows and build confidence in new ways of working.
Talent matters, but enterprises do not necessarily need more people experimenting with models; they need people who understand both the technology and the business problem. Scaling AI therefore requires a combination of sound infrastructure, governance and organisational readiness.
How should enterprises balance AI-driven decision-making and human judgement?
The right balance depends on the consequence of the decision. AI can operate with considerable autonomy where the task is repetitive, well defined and relatively low risk. As the consequences increase, the role of human judgement should become stronger. In healthcare, for example, AI can help clinicians identify patterns or bring relevant information together, but the clinician remains responsible for interpreting that information in the context of the individual patient.
The same principle applies in financial services and other sensitive environments. We should not think of humans and AI as competing decision makers. The better model is to let AI handle scale and complexity while people retain accountability for decisions that require judgement.
How do you identify which business problems are genuinely worth solving with AI?
We start with the problem, not the technology. A strong AI opportunity usually has a meaningful amount of complexity, repeated decision-making, large volumes of information or significant friction in the existing process. We then ask whether intelligence can materially improve the outcome rather than simply make the process look more sophisticated.
The business case has to be visible in measures such as time saved, accuracy, productivity, customer experience or decision quality. There are also situations where conventional software or process improvement is the better answer. AI should not be used simply because it is available. Its role is justified when it can change the economics or effectiveness of the underlying workflow.
What are the biggest opportunities for AI in healthcare, and what makes healthcare AI different?
Healthcare offers an opportunity to use AI where the volume and complexity of information can make timely decision-making difficult. In precision care, AI can help bring together relevant patient information and support clinicians in identifying patterns that may otherwise be difficult to assess quickly. Patient engagement is another important area, particularly when digital tools can help people navigate care more effectively and stay connected with providers.
What makes healthcare different is the level of responsibility involved. A recommendation can have a direct impact on a person’s health, so accuracy, privacy, clinical context and human oversight are essential. Healthcare AI therefore has to be built around trust and outcomes, not technology alone.
How important will observability and AI governance become as AI enters critical enterprise infrastructure?
They will become fundamental. As organisations move from AI pilots to systems that influence real workflows, they need to understand not just whether a model is producing an answer, but why the system behaved in a particular way and whether it continues to perform as expected.
Governance provides the framework for security, accountability, data use and appropriate oversight, while observability helps teams identify performance issues and unexpected behaviour in production. I expect this to become a meaningful differentiator. The strongest enterprise AI systems will not necessarily be those built around the largest model. They will be the ones organisations can operate reliably, securely and at scale.
What changes are you seeing in engineering talent required to build AI-first products?
The profile of engineering talent is becoming broader. Traditional software engineering fundamentals remain extremely important, but AI-first products require people who are comfortable working across software, data, models and product requirements. Engineers increasingly need to understand how an AI system behaves in a real workflow rather than treating the model as an isolated component.
At our India innovation centre, this also means encouraging collaboration across disciplines rather than creating separate technology silos. With more than 200 engineers working in India, the opportunity is to build teams that combine strong engineering discipline with an understanding of enterprise problems. That combination is increasingly important as AI moves into production environments.
How will AI coding assistants and autonomous development agents change the role of software engineers?
Software engineers will spend less time on some of the mechanical parts of development, but that does not make engineering less important. If anything, the ability to define the right problem, design the architecture and evaluate the quality of an AI-generated solution becomes more valuable.
Engineers will need stronger judgement around security, reliability, performance and how different components work together. They will also need to understand the business context behind what they are building. AI can accelerate implementation, but someone still has to decide what should be built and whether it works properly in the real world. The role shifts from writing every line manually towards designing, directing and validating systems.
Will smaller, specialised and domain-trained models become more important for enterprise AI?
I expect enterprises to use a mix rather than converge on one model strategy. Large general-purpose models are powerful and useful for broad reasoning and language tasks, but they are not automatically the best choice for every enterprise workload. In many cases, a smaller or specialised model can provide the right level of performance with better control over cost, latency, privacy or deployment requirements.
The important question is not which model is largest, but which approach is appropriate for the task. For domain-specific applications, combining specialised intelligence with enterprise data and workflow context can be particularly valuable, especially where accuracy and reliability are critical.
What metrics should CEOs and technology leaders use to determine whether an AI investment is creating meaningful business value?
The first measure should be tied to the business problem the AI was intended to address. If the goal is to improve a workflow, leaders should look at cycle time, productivity, error rates or throughput. If the objective is better customer engagement, they should measure meaningful changes in customer behaviour and outcomes rather than simply usage of an AI feature. In healthcare, the measures may relate to patient engagement, care coordination or decision support. Cost reduction is important, but it is only one dimension. The more useful question is whether AI is improving the economics, quality or experience of a process in a way that can be sustained as adoption grows.
Over the next three to five years, will AI simply become another software layer or fundamentally change how organisations work?
I expect it to change how organisations work. AI will certainly become embedded in the software layer, but its larger impact will come from changing the division of work between people and technology. Today, many enterprise systems require people to move information between applications, interpret inputs and coordinate the next step.
Increasingly, AI will be able to handle parts of that chain itself. That does not mean organisations will become fully autonomous. Instead, people will spend more time on decisions, relationships, exceptions and areas requiring judgement, while AI manages more of the routine complexity. The organisations that benefit most will redesign workflows around that new balance rather than simply adding AI to existing processes.
Looking ahead to 2030, what will the enterprise technology landscape look like, and what role will SAIGroup and India play?
By 2030, I expect AI to be embedded into most enterprise workflows rather than treated as a separate technology category. Agents will handle increasingly complex sequences of work, while domain-specific intelligence will make those systems more useful in areas where context matters. The shift will be from asking software to provide information towards having software help complete the work itself.
Symphony AI Group’s role is to build businesses around those high-value opportunities and turn AI capabilities into products that deliver measurable outcomes. India has an important role to play in that journey, not only as an engineering base but as a centre for product development, innovation and the next generation of enterprise AI talent.
