EO StudioAI is way Underhyped. He Runs His Entire Marketing Team with 40 AI Agents | Relay.app, Jacob Bank
CHAPTERS
- 0:00 – 1:30
Jacob’s 40-agent marketing “team” and why AI is underhyped
Jacob Bank explains how he built a system of ~40 AI marketing agents that effectively functions as his marketing team at Relay.app. He argues most professionals are using AI at only a fraction of its potential and points to dramatic leverage (viral posts, low tool cost) as evidence.
- •Built ~40 AI marketing agents while being the only human marketer at Relay.app
- •Uses an “AI coach” to analyze what content styles/themes are working
- •Example results: LinkedIn post about his 40 agents reached ~1.5M impressions
- •Cost comparison: ~$500/month AI tools vs tens of thousands for contractors
- •Claim: building AI agents will be a defining career skill for decades
- 1:30 – 2:02
Relay.app background and the productivity multiplier of AI inside a small team
Jacob gives context on Relay.app and his own career path (research, founder, Google product lead). He frames AI as the reason a small team can perform like a much larger org.
- •Relay.app builds AI agents/workflows that do work on your behalf
- •Current team composition and size (9 people) vs estimated need without AI (~15)
- •Jacob’s vantage point from AI research, Google, and startup building since 2021
- •AI described as a force multiplier for execution and output
- 2:02 – 3:09
Lesson 1: Why AI is not an intern—use it as a coach and strategist
Responding to the ‘AI is an intern’ framing, Jacob explains why that mental model breaks down. He shares how AI can provide high-level coaching and analysis—not just task execution.
- •Intern framing covers only a subset of AI value (time-saving tasks)
- •AI can do strategy, competitive analysis, and content creation beyond intern capability
- •Example: AI sales-call coach reviewing transcripts and giving specific feedback
- •AI coaching can replace expensive human coaching for a fraction of the cost
- •Key idea: build coaching loops around your real work outputs (transcripts, artifacts)
- 3:09 – 5:40
Lesson 2: The rise of the “Super IC” and the decline of pure management roles
Jacob predicts many roles will blend hands-on execution with coordination of AI agents. Pure task-only junior roles and large-scale people management become less central as companies get leaner.
- •Certain junior repurposing tasks (e.g., turning videos into blog posts) will disappear
- •Large-company management at massive scale becomes less important as orgs shrink
- •Future role split: mostly IC work + a portion coordinating AI agents
- •People want more strategic influence and closer connection to real work
- •Super ICs must define what good looks like and still be hands-on in final output
- 5:40 – 6:10
Inside the 40-agent system: practical examples of what the agents do
Jacob walks through how the agent org works in practice: many small, single-purpose agents rather than one complex generalist. He lists concrete workflows that automate distribution and competitive monitoring.
- •Agents are simple individually but powerful in aggregate
- •Auto-generate LinkedIn posts and tweets from each new YouTube video
- •Competitor monitoring: alerts when competitor CEOs post on social channels
- •Weekly competitor pricing checks and change detection
- •System built incrementally, one agent at a time
- 6:10 – 6:42
Lesson 3: Two rules for building an agent organization—start simple and compose
Jacob outlines his first rule: avoid building a single agent that does everything. Instead, create narrowly scoped agents and optionally add an orchestrator agent later.
- •Avoid ‘one agent that does 25 things’—it becomes unreliable and hard to steer
- •Begin with one agent doing one job; expand gradually
- •Add additional specialized agents as needs become clear
- •Optionally add a higher-level agent to invoke others when necessary
- •Don’t start with a 40-agent org chart—earn complexity over time
- 6:42 – 8:13
Lesson 3 (rule two): Maintain, iterate, and fire agents that don’t deliver value
Jacob’s second rule is that agents require ongoing management—prompts, formats, and outputs must evolve with reality. He gives examples of ‘firing’ or repurposing agents when a workflow isn’t effective anymore.
- •Agents aren’t ‘set it and forget it’; they need continuous modification
- •Example: post-call Google Doc deliverable failed—customers didn’t read it
- •Agent was ‘fired’ and repurposed to embed the summary directly in follow-up email
- •Stop or pause entire functions when priorities shift (e.g., SEO no longer a focus)
- •Benefit: no emotional baggage or coordination overhead—rapid experimentation
- 8:13 – 8:43
AI as augmentation, not replacement: reclaiming time for high-value work
Jacob frames agents as giving him “superpowers” to do more of the marketing work he actually enjoys and that drives results. The emphasis is on ambition and leverage rather than headcount reduction.
- •He resists the ‘replacing people’ framing; prefers ‘increasing capability’
- •Agents free time to pursue more marketing activities and experimentation
- •Enables doing work that’s valuable but previously time-constrained
- •Leaner teams can still execute across many channels with agent support
- 8:43 – 10:44
Lesson 4: Why staying at Google (or any single environment) can be the riskiest path
Jacob redefines career risk: tying your identity and skills too tightly to one company environment can be fragile in a changing world. He argues that optimizing for learning, network, and breadth is more robust long-term than optimizing for near-term salary stability.
- •Historical definition of ‘safe’ vs ‘risky’ careers has flipped
- •Cash comp may be safer at big tech short-term, but career robustness is different
- •Long-term advantage comes from broad networks, new skills, and diverse experiences
- •Choose to engage with the AI shift rather than ignore it
- •Reframe: the biggest risk is stagnation; progress comes from growth environments
- 10:44 – 11:44
Personal growth example: learning marketing from scratch via consistent effort
Jacob shares his discomfort with public marketing and social media, and how he pushed through because it mattered for the company. He frames uncertainty as a chance to grow rather than something to avoid.
- •No marketing background; initially avoided social platforms and self-promotion
- •Organic LinkedIn became Relay.app’s best channel despite discomfort
- •A year of deliberate learning turned into competence and enjoyment
- •Using uncertainty as a lever for skill-building rather than fear
- 11:44 – 13:17
Lesson 5: Parenting in an AI world—real-world resilience and healthy tech boundaries
Jacob describes how his kids are growing up with futuristic technologies (e.g., self-driving cars) as normal. He emphasizes giving children real-world challenges while being thoughtful about addictive algorithmic content.
- •Kids normalize autonomous vehicles; Jacob sees safety and convenience upside
- •Mixed feelings about algorithmic feeds like YouTube Shorts
- •Parenting focus: real-world experiences and challenges build resilience
- •‘Skinned knees are good’—learning requires manageable real-world risk
- •Accepting discomfort in learning (e.g., swallowing water while learning to swim)
- 13:17 – 14:43
Future-proof skills: instructing AI clearly and building value from personality
Jacob distills what he wants his kids (and future workers) to learn: clear articulation and social connection. As AI handles more execution, humans differentiate through precise direction-setting and authentic relationship-building.
- •Skill #1: clearly articulate what matters, what to do, and how to do it
- •Interest in logic/philosophy to improve clarity of thinking and instruction
- •Jobs shift toward efficiently instructing AI rather than doing everything manually
- •Skill #2: find where your unique personality/abilities create value AI can’t
- •Creator-led content (e.g., his YouTube videos) builds social connection and trust
- 14:43 – 15:44
Host’s geopolitical coda: AI leadership, manufacturing capacity, and US–China stakes
The host closes with a broader critique: the US may lead in some AI areas but risks misallocating it toward low-leverage uses. He contrasts that with China’s manufacturing orientation and raises concerns about strategic industrial capacity.
- •Claim: US AI lead over China is narrow and fragile
- •US incentives skew toward PowerPoints/litigation vs manufacturing applications
- •China applying AI to iPhones, batteries, drones, and munitions (host’s view)
- •Shipbuilding comparison used to illustrate industrial capacity gap
- •Conclusion: both countries may be making major strategic mistakes
