The global artificial intelligence conversation is undergoing a seismic shift. The initial frenzy over parameter sizes and foundational model capability is giving way to a far more pragmatic focus: real-world enterprise execution. “There was a time in the world of AI where the model debate was the only debate. Now we are getting into the value debate,” explains Amit Kumar, Managing Director, Digital Natives Business at Google Cloud India. “The model in itself will not derive the value, but the model is an important part of the architectural conversation.” This evolution isn’t just reshaping enterprise software, but also fundamentally transforming the talent landscape. As AI tools democratise software development, the traditional barriers to innovation are dissolving. “Engineering is not a degree anymore. It’s a mindset,” Kumar notes, pointing out that curiosity and hands-on experimentation are replacing formal coding degrees as the primary drivers of technological impact. Kumar works with startups and digitally driven companies, helping them use Google’s cloud and AI technology to grow their businesses. On the sidelines of Google’s first ‘Let’s Talk AI’ media tour stop in Bengaluru, Kumar spoke about how India has emerged as a high-velocity proving ground for global innovation and how enterprises and startups are navigating AI’s next frontier. Edited excerpts: YourStory: How is the conversation around enterprise AI evolving from initial experimentation to real business value, and why are agentic workflows central to this transformation? Amit Kumar: Decision-makers across manufacturing, banking, healthcare, government, ecommerce, and fintech have moved well past early experimentation and basic testing. We are now in a stage where CXOs are asking how these solutions scale at an enterprise level and deliver tangible ROI. However, to truly extract value from agentic platforms, it cannot be treated as just an IT overlay. You have to fundamentally redesign your internal workflows and rethink how you measure outcomes. An agentic workflow requires rethinking the architecture from the ground up: defining what information the agent can access, what context it retains, which systems it integrates with, and where its boundaries lie. At the same time, organisations face proliferation concerns; they want employees to use AI, but they must ensure security is not bypassed and data isn’t leaked. Real value comes from embedding AI directly into customer journeys and core operations. Look at digital leaders in India: ecommerce players like Meesho and Flipkart are enhancing customer search, improving contact centres, and introducing virtual try-ons and catalogue enhancements. MakeMyTrip is bringing multimodality into travel search, while media platforms like Pocket FM use AI for storytelling and video creation. Across banks, NBFCs, and manufacturing, companies are AI-enabling their workforces to drive productivity while establishing proper governance frameworks. YS: With discussions around a global AI slowdown, how is India positioning itself in AI adoption and solving complex, large-scale problems? AK: I do not use the word slowing down at all; if at all, India is leapfrogging. When you solve problems in India with all the natural complexity we have, there is a distinct advantage of doing things here. Our scale is completely different, and our language context is extremely dynamic with a unique mix of spoken languages and buying habits that vary across the country. Because of this complexity, Indian digital companies are solving global problems right here sitting out of India. Whether it is workforce productivity in banks, government enterprises, and manufacturing, or advanced consumer interfaces, we see a prudent yet rapid movement. Decision-makers across the board are not adopting AI simply because somebody else is doing it; that is not the stage we are in now. Most CXOs have personally tried these tools and appreciate the deeper architectural requirements around cybersecurity, governance, and data location. India is building solutions for global customers, and adoption is only scaling. YS: As data sovereignty and regulation become critical priorities, how does Google approach building sovereign AI stacks while ensuring flexibility for enterprises? AK: At the heart of Google’s philosophy are optionality and explainability. Optionality means offering choice: using industry-standard hardware like NVIDIA GPUs or Google’s own Tensor Processing Units (TPUs), and choosing between Google’s first-party models or open-weight, open-source models from other frontier labs. Sovereignty is about keeping specific regulated data within the country, but a single rigid solution does not fit every scenario. We offer Google Distributed Cloud as an air-gapped (completely isolated from external networks, including the internet) solution created on the customer’s premises under their control with no connection to the internet, which governments, defense, judiciary, and intelligence institutions are embracing. However, keeping everything strictly on-premises can curtail innovation and limit extended user benefits. That is why we believe a hybrid approach is best. For instance, a bank might keep sensitive core workflows on-premises while leveraging public cloud infrastructure for broader customer interactions. We see ourselves as an enabler, working closely with Indian partners, startups, and regulators to build sovereign solutions on our stack while adhering to strict safety guardrails and responsible AI principles. YS: With autonomous agents handling enterprise data, how should organisations construct security guardrails, and why is personal hands-on experience so crucial? AK: Architectures are being actively debated to manage dependency and risk. Depending on the sensitivity, the answer might be a human-in-the-loop requirement where decisions generated by AI are reviewed by a person before final approval, or a maker-checker model. Cybersecurity, governance, and data access must be rethought grounds up for agentic workflows. At a personal level, AI is no longer a conceptual discussion. The depth and pace of advancement are so intense that you cannot appreciate AI just by looking at a slide, reading a report, or watching a dashboard. Unless you build a demo yourself, you will not grasp what it takes. I sat through four days of enablement myself, building my own demos. The richness of human-like interaction is staggering. During an insurance sales demo, a founder suddenly tested our Gemini bot with an off-the-wall prompt: “Would you like to have a brick omelette with me?” The bot didn’t break stride; it responded smoothly, “Sir, enjoy your omelette, let me give you the next step of this.” Being hands-on separates those who truly understand AI from those who merely observe it. YS: As AI democratises software development and makes building tools accessible beyond traditional developers, how is the core definition of an engineering qualification changing, and what advice would you give young professionals entering this shifting landscape? AK: AI is democratising engineering because you no longer need a traditional coding degree or an engineering college background to write software. With tools like Antigravity, someone can automate a complex home loan origination workflow simply by providing the physical workflow inputs and writing plain prompts. For undergrads entering the workforce, formal degrees will matter less than your mindset. My advice is to focus on core human traits: curiosity, creativity, and the ability to deal with change and ambiguity. Above all, be hands-on with technology. Do not just watch others talk about it, but actually build things yourself. AI puts the world’s university at your fingertips, so doing the work yourself is the most critical factor for success. Edited by Affirunisa Kankudti