Podcast Summary: How I Invest with David Weisburd

Episode: E295 – Why AI Agents Will Quietly Replace 80% of Investment Teams
Date: February 2, 2026
Guest: Ali (Founder, Micro One)

Episode Overview

In this episode, David Weisburd sits down with Ali, founder of Micro One, to discuss how AI agents are transforming investment teams by automating complex workflows, why the biggest leaps in AI still depend on human expertise, and how tomorrow’s investment landscape will fundamentally restructure around human-AI collaboration. Ali shares his inside perspective on building AI data infrastructure, reveals practical case studies already improving investor alpha, and candidly addresses common fears and misconceptions about “AI takeover.”

Key Discussion Points & Insights

1. What Is Micro One and Who Are Their Customers?

• Ali introduces Micro One as “the AI platform for human intelligence” — sourcing highly skilled PhDs, professors, and industry experts to generate high-quality training data for frontier large language models.
  • “We vet highly skilled people...in medical, finance and legal, but also many other domains. And we help train frontier large language models.” — Ali [00:02]
• Customers include:
  • Leading AI labs building foundational models
  • Fortune 500 enterprises developing AI agents for production use
  • “Customers are...frontier labs that build foundational models and we also have enterprise customers...building enterprise agents that we help them evaluate.” — Ali [00:35]

2. Human-Centric AI Training

• AI models aren’t self-improving: Despite hype, human expertise is still essential to improving model performance—especially as models advance from internet-scale data to specialized, expert-labeled data.
  • “It's almost entirely humans that teach the models in some way or another. Of course that started with the pretraining phase...afterwards where the models really got useful is when humans...preference labeling...now we're in this kind of expert data training where humans are creating really complex data from scratch.” — Ali [01:17]
• Stages of AI learning:
  • Pre-training via internet-scale data
  • Human preference labeling
  • Direct expert-generated data and workflows

3. The Core Bottleneck Blocking AI Agents from Broad Adoption

• Enterprise evaluation frameworks are the missing piece: To move AI agents from impressive demos to production-ready tools, rigorous, granular evaluation is needed—far beyond traditional binary QA testing.
  • “There needs to be an evaluation framework for each of the actions that the probabilistic software needs to do...because the agent, there's no notion of the agent works or doesn't work. It's instead, what is the action space of this agent?” — Ali [03:15]
• Why demos don’t equate to production:
  • “If the agent works 1 out of 5 times or 1 out of 10 times, you...record that 1 out of 5 and it looks really impressive. But then it doesn’t work four out of five times, and you cannot have that in production.” — Ali [04:01]

4. The “Killer App” for AI Agents

• Coding is the breakout vertical:
  • “The obvious example is coding...the only use case that is very useful in production now...we’ve seen the immense amount of speed it’s added to programming....Imagine applying that to every other domain.” — Ali [06:19]

5. How Investors Are Using AI for Competitive Alpha

• Automation of financial modeling (e.g. LBO models):
  • “Makers and private equity investors...creating LBO models...models are getting quite good at that...allows them to work on the higher level of thinking that any investment requires.” — Ali [07:06]
• Human-in-the-loop workflows:
  • AI gets trained in simulated collaborative investment environments (multi-expert, peer-reviewed tasks) to mirror actual investor workflows

6. Rethinking the Investment Team of the Future

• Not “coworkers,” but new systems:
  • "I don't think AI agents are going to be coworkers...AI agents are going to be systems that actually change the domain of any given function...the human will only focus on that 10% that really requires human creativity and focus. And the rest will be taken by that agent." — Ali [09:38]
• Example for 2028/2030:
  • Humans focus on industry targeting, relationship-building, and high-level decision-making
  • AI handles repetitive modeling, diligence, and operational tasks

7. Preparing for an AI-First Future

• Work will be more meaningful, not less:
  • "It's going to be a really nice future. ...What this means is that there's essentially going to be net new functions created pretty rapidly by humans in every domain." — Ali [12:12]
• Humans will create new value on top of AI-automated foundations

8. Addressing Fears: Loss of Meaning and the “Terminator” Scenario

• Low probability of runaway AI:
  • “It’s a very unlikely case where models...completely learn on their own and also have...the ability to kind of create versions of themselves...it’s very, very unlikely.” — Ali [13:49]
  • Focus on constantly updated safety evaluation frameworks
  • Government (e.g., Trump administration) is accelerating AI development, but must invest more in safety science

9. Founder Lessons: The Importance of Focus and the Right Market

• Key to scaling Micro One:
  • “Market matters a lot…Once we made that decision of just focusing on this kind of one application where the market was really hot...the company more than 30x in one year.” — Ali [15:45]
• Firing customers and focusing the team:
  • Directing all resources toward lab-specific data infrastructure
  • "We had to stop serving...startups...and only focus on the AI labs...and that kind of changed the product roadmap." — Ali [17:41]
• Paradox of focus:
  • Must balance absolute focus with handling many data domains for key customers
  • “We have to build the product in a paradoxically focused way where the focus is actually to be able to vet all types of skill sets.” — Ali [19:04]

10. CEO Practices Amid Uncertainty

• Cancel meetings ruthlessly:
  • “I try to cancel as many meetings as I can...I actually end up canceling like roughly 30% of meetings every single day by just questioning them.” — Ali [20:15]
• Talk to customers directly and often:
  • "There's no alternative than the CEO and really the whole exec team talking to customers very frequently. I am practically an account executive at Micro1, and it needs to stay this way for a while." — Ali [21:26]

Memorable Quotes & Timestamps

• "It's almost entirely humans that teach the models in some way or another." – Ali [01:17]
• "There needs to be an evaluation framework for each of the actions that the probabilistic software needs to do..." – Ali [03:15]
• "The obvious example [for AI agent killer app] is coding...that's actually an exciting thing." – Ali [06:19]
• "I don’t think AI agents are going to be coworkers...they’re going to be systems that actually change the domain of any given function." – Ali [09:38]
• "Once we made that decision of just focusing on this kind of one application...the company more than 30x in one year." – Ali [15:45]
• "There's no alternative than the CEO and really the whole exec team talking to customers very frequently...I am practically an account executive at Micro1." – Ali [21:26]

Important Timestamps

• [00:02] – Introduction to Micro One
• [01:17] – How humans train AI models; stages of AI learning
• [03:15] – The QA bottleneck and new evaluation frameworks
• [06:19] – Coding as the AI killer app
• [07:06] – Case studies: how investors use AI for alpha
• [09:38] – Why AI agents won't be "coworkers," but new types of systems
• [12:12] – Preparing for an AI-first future
• [13:49] – AI existential risk: why it’s unlikely, but safety matters
• [15:45] – Lessons in focus and scaling from Micro One
• [20:15] – Founder time management and ruthless prioritization
• [21:26] – The necessity of talking to customers

Final Thoughts

Ali and David’s conversation powerfully reframes the coming shift: AI agents aren’t just another productivity tool—they’ll surge forward to handle vast swaths of investment work, moving humans into ever-more-creative, high-impact, and meaningful roles. For investment professionals and founders alike, the path forward calls for a clear-eyed understanding of AI’s current limitations, laser focus on market opportunity, relentless customer engagement, and—above all—the willingness to reinvent work alongside AI’s rapid evolution.