🎙️ Episode Snapshot

In this deep-dive conversation with Aman Khan (Director of Product, Arize AI), Deepak (CEO Adopt.ai) explores the emerging "agent-to-care revolution" that's reshaping how we build and think about products. Aman breaks down what AI agents really are, how they're evolving from simple chatbots to sophisticated systems that can reason, select tools, and take actions on our behalf. From reimagining product roadmaps to the transformation of the PM role, this conversation offers a practical framework for navigating the agentic future that's already unfolding around us.

Topics Discussed

• Definition and components of AI agents (reasoning, tool selection, action-taking)
• The shift from human users to agents as primary consumers of web services
• How product roadmaps are evolving in the AI era (strategy over rigid plans)
• Key components for building enterprise-grade AI products
• Open source vs. enterprise tooling decisions for AI implementation
• Evolution of product management in the age of AI
• Developing customer empathy as a north star for AI product development

Key Quotes + Ideas

"An agent generally is a system that comprises of three parts - reasoning, tool selection and action"

Aman provides a practical breakdown of what makes an effective agent, using trip planning as a relatable example. The agent must reason about what you're asking, select appropriate tools (like flight and hotel booking systems), and then take the correct actions. This three-part framework helps demystify how agents work and why they sometimes fail - often one broken link in this chain can lead to complete failure, like booking a flight to San Diego instead of San Francisco. Understanding this architecture is crucial for product teams designing for an agentic future, as each component presents different challenges and opportunities.

"I kind of wonder like, what does a roadmap even look like anymore? I just think that a strategy is being adaptive to what is the latest technology in AI, trying it out for yourself and trying to understand, is this something that could impact my business?"

Aman challenges the traditional product roadmap approach in the AI era. The rapid evolution of AI technologies means rigid roadmaps quickly become obsolete. Instead, he advocates for an adaptive strategy focused on experimentation and upskilling. This represents a significant shift in how products are built - from predictable feature development to continuous exploration and adaptation. For product teams, this means prioritizing learning velocity over feature completion and developing a culture of rapid experimentation.

"The new internet actually starts to look like agents, which are intermediaries between me wanting to purchase something or solve a problem... There might be multiple agents along the layers of me actually solving a problem."

This perspective reimagines the entire internet architecture as layers of interacting agents. Just as communication evolved from messengers to telegrams to the internet, we're now entering an era where agents will serve as intermediaries across digital services. The implication is profound: companies need to consider not just how humans interact with their products, but how other agents will. This creates a cascade effect where products must be optimized both for human users and for agent interaction - a dual-interface challenge that few companies are fully preparing for.

Key Takeaways

1. Every product needs an agent strategy: It's no longer optional - companies must consider how their products will function in an agent-driven ecosystem, both as creators of agents and as services that agents will interact with.
2. Evolve from rigid roadmaps to adaptive strategies: The AI product development cycle requires flexibility and continuous experimentation rather than fixed feature sets planned months in advance.
3. Build enterprise AI with three core components: Framework (the agent's reasoning system), memory (data access layer), and observability (monitoring how the agent makes decisions) form the foundation of production-grade AI systems.
4. Start with supervised agents in lower-risk domains: Begin with human oversight in non-critical applications before attempting high-risk use cases like healthcare or finance where failures have serious consequences.
5. Product managers must upskill through hands-on building: The best way for PMs to develop AI intuition is through direct experimentation - building prototypes, testing tools, and developing a firsthand understanding of the technology's capabilities and limitations.
   
   
   ‍

If any of these insights resonated with you, check out the full episode to hear Aman's complete perspective on navigating the agent revolution that's reshaping our digital landscape.