Microsoft’s Amanda Silver: AI Agents Are the Next Big Startup Opportunity

The startup landscape is on the cusp of a dramatic shift, potentially as significant as the move to public cloud computing, according to Amanda Silver, corporate vice president at Microsoft’s CoreAI division. The catalyst? Agentic AI – AI systems capable of performing tasks autonomously – and its potential to drastically reduce operational costs for new ventures.

Silver, who has spent 24 years at Microsoft working with developers – most recently on GitHub Copilot – now focuses on tools for deploying AI applications and agentic systems within enterprises through Microsoft’s Foundry system, a unified AI portal within Azure. This vantage point provides a unique view into how companies are actually implementing AI and where those deployments are succeeding or falling short.

“I see this as being a watershed moment for startups as profound as the move to the public cloud,” Silver stated. “The cloud eliminated the need for startups to invest heavily in physical infrastructure – the real estate, the hardware, the cooling. Now, agentic AI is going to continue to reduce the overall cost of software operations.”

The cost reductions aren’t merely incremental. Silver believes AI agents can automate tasks previously requiring significant human capital, such as customer support and even preliminary legal investigations. This automation, she argues, will not only lower the barrier to entry for new startups but also lead to higher-valuation companies built with leaner teams.

But what does this look like in practice? Silver points to the increasing sophistication of “multistep agents” – AI systems capable of handling complex, multi-stage tasks. One example she cited is dependency management in codebases. Keeping software libraries up-to-date is a crucial but often tedious task for developers. “You can have these agentic systems reason over your entire codebase and bring it up to date much more easily, with maybe a 70% or 80% reduction of the time it takes,” Silver explained. “And it really has to be a deployed multistep agent to do that.”

Another area ripe for automation is live-site operations – the constant monitoring and maintenance of websites and services. Traditionally, this requires 24/7 on-call personnel to respond to incidents. Microsoft has developed agentic systems capable of diagnosing and mitigating many of these issues automatically, reducing the need for human intervention and dramatically shortening incident resolution times. “It used to be a really loathed job because you’d get woken up fairly often for these minor incidents,” Silver said. “We’ve now built a genetic system to successfully diagnose and in many cases fully mitigate issues…so that humans don’t have to be woken up in the middle of the night.”

Despite the clear potential, the adoption of agentic AI hasn’t been as rapid as some anticipated. Silver believes a key stumbling block isn’t technical uncertainty, but rather a lack of clarity around the specific business problems these agents should solve. “If you think about the people who are building agents, what is preventing them from being successful, in many cases, it comes down to not really knowing what the purpose of the agent should be,” she explained. “There’s a culture change that has to happen in how people build these systems. What is the business use case that they are trying to solve for? What are they trying to achieve? You need to be very clear-eyed about what the definition of success is for this agent.”

Silver argues that the return on investment for agentic systems is readily apparent, dispelling the notion of widespread hesitancy. She anticipates a future where agentic systems frequently operate with “human-in-the-loop” scenarios, handling the majority of tasks while escalating more complex or critical situations to human oversight.

A prime example of this hybrid approach is package returns. Computer vision models are becoming increasingly adept at assessing damage to returned items, reducing the need for human inspectors. “That’s a perfect example where actually now the computer vision models are getting so good that in many cases, we don’t need to have as much human oversight over inspecting the package and making that determination,” Silver noted. However, she acknowledges that certain operations, such as those with significant legal or financial implications – like incurring contractual obligations or deploying code to production – will likely always require a degree of human oversight.

The rise of agentic AI doesn’t signal the end of human roles, but rather a shift in focus. As AI handles routine tasks, human employees can concentrate on more strategic and complex challenges, driving innovation and growth. According to Silver, this represents an “exciting world” for startups and established enterprises alike, one where AI empowers leaner, more efficient, and ultimately more valuable organizations.