Liftoff with Keith Newman

Episode: Agentify: How AI Agents Are Reshaping Companies, Careers, and the Future of Work

Guest: Michael Palmer (CEO & Chief Scientist, Taos Research Corporation)

Date: December 30, 2025

Overview of the Episode

In this episode, host Keith Newman talks with Michael Palmer, the CEO and Chief Scientist at Taos Research Corporation and author of the new book Agentify: The Art, Science, and Engineering of Successful AI Agents. They explore the accelerating world of AI agents—how these technologies are transforming how companies are built and run, what skills and approaches are needed for the future, and how both new startups and established enterprises must adapt. The discussion weaves together business strategy, technical know-how, and the evolving “art” of making agents trusted partners in work and life.

Key Discussion Points & Insights

1. Rapid Pace of AI Change ([00:38]–[04:45])

• AI's Accelerating Evolution: 2025 is tagged as the “year of AI,” but Palmer and Newman agree it might be the third or fourth such year, depending on when people started following the trend.
• Personal Motivation: Palmer shifted from a large bank CTO role back to startup land due to the rapid pace:

  "The rate at which things were happening, I knew I just couldn't sort of sit inside a big company and... be in touch with the pace." (Michael Palmer, [02:03])

• Constant Workflow Changes:

  "My workflow for building things literally changes every... at least every month, if not every two to three weeks with just the new models coming and the new tools and the agents and what have you." (Michael Palmer, [02:46])

2. From “Vibe Coding” to Rigorous AI Engineering ([03:44]–[05:00])

• Palmer describes the transition from the experimental, improvisational coding era (dubbed “Vibe Coding”) to a more mature, high-quality, AI-assisted software development process as a key shift:

  “I actually have a chapter in the book called Beyond Vibes. Rigorous AI-assisted software development... Now I think we're entering an era where people are realizing, wait a second, if I'm going to do this right, software still has nuances and bugs and security issues.” (Michael Palmer, [03:56])

• The challenge: AI lets you prototype quickly, but quality, security, and sustainability require deeper discipline.

3. What Makes an AI Agent “Successful”? ([05:00]–[08:44])

• The book’s framing: Focus on foundational, lasting principles rather than tools that become outdated quickly.
• AI Agent Evolution: Palmer predicts a shift from passive, prompt-based agents to systems that take initiative—working more like employees than tools:

  “A lot of the agents that we see today... are going to be eclipsed in the next one to two years by agents that actually take initiative... actually function much more like an employee and pick up the ball without being told what to do.” (Michael Palmer, [06:31])

• Not AGI, But Simulated Agency:

  “With the current LLM technologies, we're not going to get to AGI... We're going to simulate more agent, true agency, true initiative. Even without AGI I think we get 80-85% of the way there.” (Michael Palmer, [07:44])

4. Misconceptions About AI Agents ([09:10]–[09:44])

• Biggest Misplaced Idea: That current agents—really just tools responding to prompts—are the end state.

  “Agents without agency... that’s really what an agent is going to be... but within a year or two we'll see much more agency and autonomy.” (Michael Palmer, [09:16])

5. AI in the Enterprise: Adoption, Orchestration, and Silos ([09:44]–[13:33])

• Most companies implement AI in point solutions (call centers, dev teams, marketing) but lack “holistic transformation.”
• Metrics & Impact: Right now, effectiveness is often measured with basic metrics—adoption rates, efficiency gains in software and call centers—but there’s a need for broader, more meaningful measurements.
• Many organizations are still in early stages of cross-functional AI orchestration, with broader transformation lagging behind:

  “Not yet much sort of holistic understanding of the enterprise and... overarching transformation.” (Michael Palmer, [11:03])

• Adoption will come from the bottom-up as well as enterprise mandates—mimicking past “consumerization of IT” trends.

6. Competing Against Legacy Vendors as an AI-First Company ([13:33]–[16:19])

• Challenging Behemoths:
  Enterprises are deeply entrenched with established vendors (Salesforce, Adobe, Oracle, Tableau, etc.), making it difficult for upstarts to break through.
• Advice for Startups:

  “Teams that want to start companies now and be successful have to be quite... laser-like, vertical, focused on a specific problem in specific areas.” (Michael Palmer, [15:00])

• Competing with generic agents means fighting giants; the path to success is through hyper-specialization, addressing specific industry or functional problems.

7. The Power of Integration and Partnerships ([16:19]–[17:49])

• Full-Solution Wins:

  “Enterprises... don't want to buy building blocks that they've got to assemble together. The thing that comes with a full solution tends to win.” (Michael Palmer, [16:51])

• Strategic partnerships with big players' ecosystems (Microsoft, Google) are increasingly important—startups need a “go to market” plan within established platforms.

8. Future Trends: Solopreneur Unicorns & Fully Autonomous Agents ([17:49]–[20:16])

• Palmer predicts the rise of the “solopreneur billionaire unicorn”—a single individual, armed with AI agents, scaling a high-value business:

  “I think we’re going to see the Solopreneur billion unicorn. I think the first kind of literally person running their operation with a lot of agents.” (Michael Palmer, [18:11])

• Fully Autonomous Agents:
  Expect to see more scenarios (outside highly regulated industries) where AI agents get near-total task autonomy, functioning as holistic employees that plan and act within guardrails.

9. Building Next-Gen AI Companies: What Matters? ([20:16]–[25:08])

• Key Capabilities for AI Agents:
  • Self-directed goal management
  • Continuous observation of their environment (not just waiting for prompts)
  • Adaptive planning and acting in real time
  • Integration with company operations and teams, not just individual workflows

  “You have to build in these loops... where the agent is not just waiting for human input, but is actually observing.” (Michael Palmer, [21:00])

• Company Structure Evolution:
  AI coordination and alignment roles may emerge, focusing on dynamic, outcome-driven performance rather than static org charts. Agile and traditional software dev roles are changing as product cycles accelerate.

  “All of these cycle times just become much, much faster because every role is empowered to do stuff that they could never do before. So the role definitions kind of have to change.” (Michael Palmer, [26:14])

10. The “Art” of AI Agents: Behavioral Design & Trust ([27:55]–[30:04])

• From Code to Character:
  • Product leaders and behavioral psychologists are needed to shape agent “personalities”
  • Differentiation among agents will be based not just on intelligence, but the “trusted partner” feel they invoke
  • Avoiding “sycophancy”—agents just telling users what they want to hear—is critical for building real trust

  “How do you make your agent feel like a really trusted partner? And that means it has to behave much more nuanced way than I think what we're seeing.” (Michael Palmer, [29:24])

Notable Quotes & Memorable Moments

• On startup disruption:

  “The smaller companies... will quickly show that level of efficiency and then the enterprise will probably either have to buy them or get disrupted by them.” (Michael Palmer, [27:38])

• On designing agent personality:

  “There’s this term sycophancy in AI... that’s an opportunity for founders and designers to really think about what builds trust. How do you make your agent feel like a really trusted partner?" (Michael Palmer, [29:24])

• On transformative potential:

  “The ultra lean startup is now in the imagination. I think we're going to see some examples that really shock people as to how small the company is.” (Michael Palmer, [18:11])

Important Timestamps

• 00:38: Introduction to Michael Palmer and why he wrote Agentify
• 02:17: Palmer’s background and motivation to leave a big bank for startup life amid AI’s fast pace
• 03:51: The shift from “Vibe Coding” to structured, rigorous AI software engineering
• 06:31: Rise of initiative-taking agents—future of “employee-like” AI systems
• 09:16: Misconceptions about what AI agents can (and will) be
• 11:03: How AI is being adopted within large enterprises
• 13:33: Why legacy vendors are hard to unseat and how startups must go “laser vertical”
• 16:51: The need for full solutions, not building blocks, in enterprise AI adoption
• 18:11: The coming era of “solopreneur” unicorns and radically lean startups
• 20:27: Critical capabilities for AI agents in the next few years
• 24:29: Agents as company coordinators—helping align goals, performance, and accountability
• 26:14: Changing cadence and role definitions in AI-powered organizations
• 29:24: The new art of agent design—building trust and behavioral intelligence

Conclusion

Through a lively, in-depth conversation filled with practical insights and big-picture thinking, Michael Palmer makes the case that the AI agent revolution has only begun. Companies will need to rethink structure, metrics, and design paradigms—while founders must make some bets on rapid specialization and authentic behavioral design. Trust, continuous learning, and close attention to both technical rigor and human experience are central to “agentifying” the future of work, careers, and even what it means to build a business from scratch.

Check out Michael Palmer’s book Agentify and his Substack via Taos Research for more on these evolving frontiers.