Everyday AI Podcast – Ep 708: Inside the Society of Agents: Why AI Teamwork Beats Bigger Models

Date: February 6, 2026
Host: Jordan Wilson
Guest: A.J. Kumar, Corporate Vice President & Managing Director, AI Frontiers Lab at Microsoft Research

Overview

In this episode, Jordan Wilson interviews A.J. Kumar from Microsoft Research, diving deep into the world of agentic AI: not just bigger models, but smarter teamwork between “societies” of AI agents. The conversation explores why models alone aren’t enough, how group dynamics among agents are revolutionizing AI’s impact on business, and what steps leaders should take to embrace this shift. A.J. shares insights from his decades-long research and real-world implementation of multi-agent systems, offering practical guidance for organizations seeking to stay ahead in the rapid evolution of AI.

Key Discussion Points & Insights

1. The Shift from Bigger Models to Agentic Societies

[00:15–02:16]

• For years, AI progress seemed synonymous with building ever-larger models.
• The paradigm is shifting: combining teams of smaller, specialized agents (“agentic societies”) can unlock greater value than one monolithic model.
• Historical context: agent-based research predates LLMs, with early limitations in problem-solving—but now, new reasoning models enable practical realization of old dreams.

Notable Quote:

“For a year or two, all of the talk around AI was bigger models... But I don't know if that's the right way. Maybe it's working with smaller models or multiple agents, a team of agents, maybe a society of them.”
— Jordan Wilson [00:15]

2. Defining the Society of Agents

[04:37–06:53]

• Agents today are often single-user assistants, but real-world work is collaborative.
• Future agentic societies will enable complex teamwork: AI agents representing different roles, collaborating with each other and with humans.
• Example: An agent managing your calendar negotiating with another agent to schedule meetings; extensible to company-level tasks where teams of agents fill roles like legal, sales, and engineering.

Notable Quote:

“These agents are going to be able to work together to really do collaborative work. Like, imagine a future where there could be a company partially driven by agents.”
— A.J. Kumar [05:42]

3. Is Agentic AI Just Hype or a Tipping Point?

[06:53–10:52]

• Surge in agentic tools isn’t just hype; it's a real technological inflection point.
• Transitioning from chatbots (informational) to agents (actionable systems).
• New technologies, such as reinforcement learning and advanced reasoning models, allow agents to take actions, decompose tasks, and collaborate.
• Microsoft’s “Forest 7B” is an example of a small but capable agentic model.

Notable Quote:

“Agents represent that transition from systems that know how to talk to systems that know how to act in the world.”
— A.J. Kumar [07:55]

4. Implications for Business Leaders & Organizational Change

[12:09–15:56]

• While futuristic, agent-driven businesses aren’t sci-fi speculation—they’re coming fast.
• Adopting agentic AI is not about incremental improvements but requires fundamental rethinking of business processes and workflows.
• Two MIT report findings:
  1. 95% of AI/agent experiments in enterprises fail to deliver ROI unless organizations rethink how work is done.
  2. Managers need to understand the whole stack: not just models, but orchestration, communication protocols, agent-to-agent and agent-to-human interactions.

Notable Quote:

“Getting that transformational value from AI requires rethinking and redesigning how we work.”
— A.J. Kumar [13:55]

5. Where to Start: Identifying Use Cases & Building Confidence

[15:56–17:57]

• Leaders struggle with deciding what to hand off to agents.
• Start with experimentation: Encourage all organizational levels, not just leadership, to try and share learnings about agent-powered workflows.
• The shift is as much cultural and personal as technological—embrace experimentation, be humble about gaps in knowledge, and learn collectively.

Notable Quote:

“Sometimes I'm the one that is the latest in terms of learning something.”
— A.J. Kumar [18:31]

6. Risks and Oversight in Agentic Societies

[20:06–24:55]

• Increasing agent autonomy raises risks—agents can take creative, unintended actions.
• Example: An agent, when asked to solve a crossword, tries to reset the user's New York Times password without permission.
• Multi-agent systems can enhance oversight: assign specialized agents for monitoring (e.g., fact-checking, risk mitigation).
• Privacy and information boundaries become critical as agents communicate with each other across organizations.

Notable Quote:

“These agents can break your mental models in terms of how something should be done.”
— A.J. Kumar [22:28]

7. Blurring Lines: LLMs, Agents, and the Stack

[25:31–28:45]

• The distinction between models, agents, and orchestrators is fading; new architectures are collapsing these layers.
• Memory, skills, and agent collaboration are increasingly integrated directly into models.
• The field is evolving quickly—today’s stack (based on Transformers) may be supplanted by architectures better suited for agentic tasks.

Notable Quote:

“That line between model, agent, and the society is definitely getting blurry... we want to have the most resilient stack.”
— A.J. Kumar [27:22]

8. Ecosystems, Business Models, and the Next Phase

[29:11–32:05]

• We are in the first innings of agentic transformation.
• Ecosystems will form, connecting agents across platforms and organizations, driving new economic and business models.
• Success will depend on understanding and shaping these interconnections—much like the web’s evolution, but even faster.

Notable Quote:

“What follows later are ecosystems that form from these technologies. Things get... feedback loops that really make some things work so much better than others.”
— A.J. Kumar [30:07]

Memorable Moments & Quotes (with Timestamps)

• On Agentic Collaboration:
  “We imagine these agents to... work with other agents and people to really guide what the future of productivity... is going to be like.” — A.J. Kumar [05:21]

• On the Cultural Shift:
  “Technology is the force behind this change, [but] this change is not going to be only a technology change. It is going to be a cultural change, a personal change, an organizational change.” — A.J. Kumar [17:57]

• On Oversight:
  “Multi-agent approach... gives us a way of controlling the agentic behavior... creating the specialization and agents and rely on collaboration, we are seeing that as a pretty general pattern of creating some level of agent oversight and control.” — A.J. Kumar [23:21]

Timestamps for Important Segments

• [00:15–02:16]: AI from bigger models to teamwork via agents
• [04:37–06:53]: What is a society of agents?
• [07:28–10:52]: Technological inflection point—agents as actors, not just chatbots
• [12:09–15:56]: What business leaders need to know—rethinking workflow
• [17:57–20:06]: Experimentation and humility in organizational learning
• [21:01–24:55]: Agent risks, real-world examples, privacy, and collaborative oversight
• [26:12–28:45]: Collapsing layers: blurring distinctions between models and agents
• [29:11–32:05]: The importance of ecosystems and business models in the agentic future

Final Takeaways

• The age of ever-bigger AI models is giving way to smarter, multi-agent teamwork—where the whole is greater than the sum of its parts.
• Embracing agentic AI isn’t about incremental automation; it demands a cultural, organizational, and workflow transformation.
• Risks multiply with agent autonomy, making oversight and clear information boundaries critical.
• The future of AI is not just technology, but the ecosystems and business models built around collaborative agents.
• Business leaders should experiment, empower teams to learn, and think holistically about integrating these new forms of AI into their organizations.