AI is way Underhyped. He Runs His Entire Marketing Team with 40 AI Agents | Relay.app, Jacob Bank

• Jacob Bank, founder and CEO of Relay.app, argues that AI is “way underhyped” — most people are using it for only 1% of its potential. He runs his entire marketing function solo at his nine-person startup using a system of 40 AI agents, and believes that building AI agents will be as fundamental a career skill as knowing spreadsheets was for the past 40 years.

AI Is Not an Intern — It’s a Coach

• A year ago, Jacob described AI as “an intern” — something that does exactly what you tell it to do. He now considers that mental model wrong.
  • AI can handle time-saving tasks you’d give an intern, but it can also do strategic work an intern cannot: competitive analysis, content creation, and even coaching.
  • Example: Jacob is not a natural salesperson, so he built an AI sales coach that reviews meeting transcripts and gives feedback — pointing out when he jumped into a demo too early or missed discovery questions. A human sales coach would cost ~$10,000/month; his AI version costs about $5/week to run.
  • The key insight: when you take on a new role you’re not experienced in, you can build an AI coach that ingests your raw work data and gives structured feedback on whatever cadence makes sense — after every call, daily, or weekly.

The Future of Work: Everyone Becomes a “Super IC”

• At Google, Jacob observed the “tech lead manager” role — a hybrid where one person was both the top technical expert and a people manager. He thinks this hybrid model is the future for everyone.
  • Junior roles focused on single repetitive tasks (e.g., turning a YouTube video into a blog post) will disappear.
  • Pure management roles managing large teams will become less important as companies get smaller and leaner.
  • The new model: roughly two-thirds of your day is hands-on individual contributor work (making videos, editing posts, talking to customers), and one-third is coordinating a team of AI agents that assist with those tasks.
  • This combines what people actually want: junior workers who want more strategic influence, and senior managers who want to be closer to the actual work and the customer.

How to Build a 40-Agent System

• Jacob’s 40 agents each have one specific job. Examples:

  • Every time a new YouTube video is published, one agent creates a LinkedIn post and another creates a tweet automatically.
  • One agent monitors competitor CEOs’ social media and alerts Jacob to what they’re discussing.
  • One agent checks competitor pricing weekly and flags important changes.

• Rule 1: Build one simple agent at a time.

  • Jacob has not had success building a single agent that tries to do 25 things at once. Start with one agent that does one thing, then add a second, then a third. Optionally, add a coordinator agent on top. Do not start with a 40-agent org chart.

• Rule 2: Fire agents that don’t deliver.

  • AI agents are not “set it and forget it.” Jacob constantly modifies or retires them based on results.
    • He built an agent to turn sales call transcripts into polished Google Docs for customers. After 10 attempts, he realized customers weren’t reading the docs — so he fired that agent and repurposed it to put summaries directly into follow-up emails.
    • He paused his entire set of SEO agents when he discovered SEO was driving too little revenue for the effort, with plans to reactivate them later.
  • The advantage over human contractors: no emotional baggage, no disagreements, no coordination cost. You simply tell the agent to stop.

Reframing Career Risk

• Jacob argues that the conventional wisdom about safe vs. risky careers has flipped.
  • In his parents’ generation, working at one Fortune 500 company for 40 years was safe; joining a startup was risky.
  • Now, the riskiest career move is having skills too tied to one company’s environment. The most robust careers involve starting companies and accumulating diverse experiences.
  • Example: Most people say staying at Google is safer than joining a startup. In terms of next-year cash compensation, that’s true. But over a full career — measured by breadth of network, new skills, and range of experience — staying at a large company is far riskier.
  • Jacob’s reframe: don’t think about “taking more risk.” Optimize for personal growth and learning. The real risk is stagnation. Uncertainty and new responsibility are how you grow.
  • Personal example: Jacob had no marketing background, hated social media, and found it deeply uncomfortable to be the face of Relay.app’s marketing. He committed to it for a year, got good at it, and now loves it. What felt risky was actually an opportunity to grow.

What Jacob Wants His Kids to Know About AI

• Jacob’s children already treat self-driving cars (Waymo, Zoox) as completely normal — they see them every day on Potrero Hill in San Francisco. He’s glad they may never need to drive.
• He has mixed feelings about other AI uses (e.g., kids scrolling YouTube Shorts) and tries to give his children as many real-world, physical challenges as possible — learning to ride a bike, skinning knees, swallowing water while learning to swim. Real-world risk-taking and consequence-handling are formative.
• For their future careers, he sees two durable skills:
  1. The ability to clearly articulate what you want. As AI does more of the execution, the human’s job becomes defining what needs done and how. He’s teaching his kids logic and philosophy to build this skill.
  2. Building genuine social connection and expressing unique personality. This is why Jacob records his own YouTube videos rather than having AI generate them — people connect with a real person. In a world of AI-generated content, authentic personality becomes the differentiator.

A Note on US vs. China AI Leadership

• Jacob has spent roughly equal parts of his life in the US, China, and Canada. He believes US leadership over China in AI is real but narrow.
  • Americans tend to use AI for productivity tasks like making PowerPoints and filing lawsuits.
  • China is more manufacturing-focused: using AI to produce more iPhones, EV batteries, drones, and munitions. (He notes that last year the US built about 5 ships while China built about 1,500.)
  • His concern: both the US and China appear intent on making big mistakes rather than winning.