The Startup Ideas Podcast
The Startup Ideas Podcast

How to win with AI Agents in 2026

April 29, 2026

AI Summary

5 min read

Howie Liu, co-founder and CEO of Airtable, has been building a new product called Hyper Agent, and he believes the opportunity in AI agents is far larger than most people realize. In this conversation, he walks through the state of the market, demonstrates how Hyper Agent works, and explains why the biggest barrier isn't the technology—it's that people aren't putting in the time to get good at using it.

The opportunity is bigger than the charts show

Liu reacts to a Sequoia chart showing AI agent deployment concentrated in software engineering (nearly 50%), with back office, marketing, and sales each below 10%. He argues this reflects "under penetration" rather than a ceiling. The frontier models are already smart enough to handle almost any white-collar task—the bottleneck is deployment. "If you took frontier agents today and deployed them into every one of these categories, you should get to 100%," he says. Even the 50% figure for software engineering is misleading, because most companies are still using AI for autocomplete rather than fully autonomous development workflows. Liu describes his own shift: he now runs 30 cloud code instances in parallel, each coupled to a browser, fully autonomous. The total addressable market isn't a trillion dollars—it's "the whole GDP of all white collar labor," which is many tens of trillions.

The unit economics reframe

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

What you'll learn

  • 1 (00:00) **Episode Introduction & Sponsor Offer** - Host Greg Isenberg introduces Howie Liu, co-founder and CEO of Airtable, and teases a $1,000 Hyper Agent credit giveaway for the first 1,000 listeners.
  • 2 (01:52) **The Under-Penetration of AI Agents** - Howie reacts to Sequoia's chart showing AI agent deployment by domain (software engineering at ~50%, most others below 10%).
  • 3 (04:42) **Copilot vs. Autopilot: The Real Opportunity Size** - Howie argues the TAM is not just $1 trillion but "the whole GDP of white collar labor" (many tens of trillions).
  • 4 (06:59) **Why Now is the Best Time to Start a Company** - Howie and Greg discuss the urgency of building with agents.
  • 5 (08:29) **Reframing AI Costs vs. Human Value** - Howie addresses the "sticker shock" of frontier model token costs.
  • 6 (11:11) **The Fastest Enterprise Adoption in History** - Howie reacts to the AI adoption curve and the two angles for building a valuable AI business.
  • 7 (14:48) **The Agent Command Center Vision** - Howie explains why agents will map to human job roles, not a single omnipotent AI.

+ Full timestamped outline available in the app

Show Notes

Limited BONUS: First 1,000 builders get $1,000. Claim yours while supplies lasts.: https://startup-ideas-pod.link/hyperagent

I sit down with Howie Liu, co-founder and CEO of Airtable, to talk about the agent economy and the launch of HyperAgent. We walk through Sequoia's charts on AI agent deployment, the economics of token-based work versus human labor, and why frontier agents have crossed a threshold that changes how companies get built. Howie then does a live show-and-tell of HyperAgent, including a custom "Greg Isenberg contrarian AI" skill he spins up in real time. This one is for anyone building a solopreneur business, operating a fleet of agents, or trying to figure out where to place their bet in the agent ecosystem

Timestamps

00:00 – Intro

02:22 – Sequoia's AI agent deployment chart reaction

04:41 – Copilot vs Autopilot territory and the $1T+ opportunity

08:13 – Agent economics vs human labor costs

11:12 – Fastest enterprise adoption curve in history

14:48 – The agent command center and fleet of 20 agents

18:03 – What is HyperAgent?

19:43 – Live demo: hyperlocal real estate market reports

22:38 – HyperAgent as the founder, not just the developer

23:21 – Street View, Zillow redesigns, and visual tool power

24:15 – Command center view across a fleet of agents

25:48 – Skills as the key primitive for frontier agents

26:30 – Building the Greg Isenberg contrarian AI skill live

32:31 – HyperAgent vs Perplexity Computer, Manus, OpenClaw, Codex

34:52 – Reviewing writing skill

36:55 – The arbitrage of persistence

41:31 – Confidence milestones: first dollar, $10K/month

35:27 – Reviewing contrarian tweet drafts live

45:05 – Giving the agent feedback and building rubrics

50:15 – Connectors, OAuth, and building custom API skills

53:03 – How to get started with HyperAgent

01:01:54 – Credit giveaway for listeners

01:03:31 – Closing Thoughts

Key Points

  • Frontier agents have crossed a threshold in the last 4–5 months where they function as true autonomous coworkers, not just chat assistants.
  • Reframe agent cost by value delivered: a $150 token spend for a board memo beats hours of human time,
The Startup Ideas Podcast

More from this podcast

The Startup Ideas Podcast →