Speaking at TechSparks 2026, Amazon Vice President of Machine Learning Rajeev Rastogi framed the next phase of artificial intelligence as a structural change in how enterprises run, not merely a new interface for software. His central argument was that the market is moving beyond the hype cycle around generative AI and into a more consequential era of agentic AI, where systems can interpret natural language instructions, break down tasks, and act across business processes with limited human intervention.
Enterprise Layer Shift
Rastogi's remarks point to a growing consensus among technology leaders that the most valuable AI applications will not be confined to content generation or customer-facing chatbots. Instead, the opportunity lies in embedding autonomous agents into the enterprise operating layer — the systems that govern workflows, approvals, support functions, and internal decision-making. In that model, AI is no longer a tool that responds to prompts in isolation. It becomes an operational participant that can coordinate actions, retrieve context, and complete multi-step tasks.
That distinction matters for startups, investors, and large technology buyers alike. The first wave of GenAI adoption was often driven by visible but narrow use cases: drafting text, summarising documents, and assisting with search. Agentic AI, by contrast, promises deeper integration into core business functions such as procurement, customer operations, finance, and IT support. If successful, it could shift AI from a productivity enhancer at the edges to a system that materially changes enterprise throughput and cost structures.
Rastogi's comments also reflect the practical pressure facing companies that have spent the past two years experimenting with large language models. Many enterprises are now asking a harder question: where is the measurable return? Agentic systems offer a clearer answer because they are designed to reduce manual handoffs, accelerate routine decisions, and automate workflows that currently require repeated human oversight. That makes them especially attractive in large organisations where operational friction is expensive and scale is difficult to manage.
From Prompts To Actions
The move from static software to autonomous systems is one of the most important shifts in enterprise technology architecture. Traditional software waits for users to click, fill, or submit. Agentic AI can understand an instruction in natural language, determine the sequence of actions required, and execute those steps across connected systems. In effect, it changes software from a passive repository of functions into an active layer of execution.
For Amazon, a company with deep exposure to cloud infrastructure, machine learning, and enterprise services, the strategic implications are significant. The cloud market is increasingly defined not just by compute and storage, but by the ability to offer AI systems that can be deployed safely, reliably, and at scale. Enterprises will demand guardrails, observability, and governance before allowing autonomous agents to touch sensitive workflows. That means the winners in this market will likely be those that can combine model capability with enterprise-grade controls.
The broader startup ecosystem is also likely to feel the effects. Venture capital has already begun shifting attention from generic AI wrappers toward infrastructure, orchestration, and vertical applications that can demonstrate operational impact. Agentic AI opens a new category of products, but it also raises the bar: startups will need to prove that their systems can work across real enterprise environments, not just in demos.
At the same time, the transition will not be frictionless. Autonomous systems introduce questions around reliability, auditability, and accountability. Enterprises will need to know when an agent can act independently, when it must escalate to a human, and how errors are traced and corrected. Those concerns are likely to slow adoption in regulated sectors, even as the technology matures.
Still, Rastogi's message at TechSparks 2026 was clear: the next competitive advantage in AI will come from systems that do more than generate language. The real prize is software that can understand intent, navigate enterprise complexity, and execute work. For businesses under pressure to do more with less, that promise may prove more durable than the initial GenAI frenzy.