Hajra Naz
For 24 years, Amanda Silver has worked at Microsoft, helping developers build better tools. In recent years, her focus has shifted to AI, particularly building tools that help enterprises deploy intelligent applications and agentic systems.
After her work on GitHub Copilot, Silver now serves as a corporate vice president in Microsoft’s CoreAI division, overseeing tools like the Foundry system inside Azure. Foundry acts as a unified AI portal for enterprises, giving Silver insight into how companies are actually using AI and where deployments succeed or fail.
I spoke with Silver about the capabilities of enterprise agents, the challenges of AI adoption, and why she believes this is the biggest opportunity for startups since the rise of the public cloud.
AI as a Watershed Moment for Startups
Q: How do you see AI impacting startups that aren’t traditionally AI-focused?
Silver compares this moment to the public cloud revolution. “The cloud changed the game for startups,” she explained. Previously, young companies had to invest heavily in physical servers and hardware infrastructure. Cloud computing made it cheaper and easier to launch new ventures.
“Now, agentic AI is doing something similar,” Silver said. “It reduces the overall cost of running software operations. Many tasks — from support to legal research — can now be automated. This means startups can launch faster, with fewer people, and potentially reach higher valuations with leaner teams. It’s an exciting world.”
Read More: AI Was Supposed to Make Work Easier. Instead, It’s Burning Us Out
AI in Action: Practical Applications
Q: What does AI-driven automation look like in practice?
Silver highlighted how multistep agents are transforming coding and operations tasks.
-
Code Maintenance: Developers must often update dependencies like older versions of .NET runtime or Java SDKs. Agentic systems can analyze entire codebases and bring them up to date, reducing effort by 70–80%.
-
Live-Site Operations: Maintaining websites or services often requires on-call staff to respond to incidents in the middle of the night. AI agents can now diagnose and mitigate issues automatically, reducing downtime and the need for humans to intervene in minor incidents.
These examples show how AI agents can improve efficiency while freeing humans from repetitive or stressful work.
The Slow Adoption Puzzle
Q: Why haven’t agentic AI deployments accelerated faster?
Silver points to unclear purpose as a primary barrier. Many teams building agents don’t fully define the business problem they want the AI to solve. She explained:
“You need clarity on the agent’s objective and the data it will reason over. Once that’s defined, the return on investment becomes clear.”
It’s less about technical limitations and more about organizational mindset. Teams must understand the outcomes they expect and structure agents accordingly.
Human-in-the-Loop: Balancing Automation and Oversight
Q: What about concerns over AI replacing humans entirely?
Silver emphasized that human oversight remains essential in many scenarios. For example:
-
Package Returns: Computer vision models can now inspect packages and process most returns automatically. Only borderline cases may require human intervention.
-
Critical Operations: Tasks like deploying production code or signing legal agreements still need some human review.
Even so, AI can automate the majority of the workflow, allowing humans to focus on the most important decisions rather than routine tasks.
Read More: Why Businesses Will Prioritize Other AI Skills Over Prompt Engineering in 2026
The Big Picture: Startups, Efficiency, and Opportunity
According to Silver, AI isn’t just a tool; it’s a catalyst for a new generation of startups. By lowering operational costs and enabling small teams to accomplish more, agentic AI could lead to faster launch cycles, leaner organizations, and higher-value ventures.
In her view, startups are positioned to leapfrog traditional constraints, much as cloud computing once allowed companies to innovate without massive infrastructure.
“We’re entering a phase where software operations are cheaper, faster, and more intelligent than ever before,” she said. “That’s transformative for startups and enterprises alike.”
