Skip to content
Aakash GuptaAakash Gupta

AI Agents • Live Demo: The Game-Changer for 2025

2% of PMs use AI agents for productivity. That's too low. You can 10x your productivity with AI agents: So I had on former Director of Product at Gmail, and founder of AI agent startup Relay.app Jacob Bank to break it all down. ---- ⏰ Timestamps: 00:00 Intro 01:54 Meet Jacob: The One-Man Marketing Army 02:14 Live Demo: 12-Agent Executive Assistant 08:27 Meeting Briefing Generator Walkthrough 15:05 Ads 18:52 Building Follow-Up Drafter Agent 25:19 Live Build: Reddit Brand Tracker 30:57 Ads 35:49 Managing AI Agent Notification Overload 37:57 Current AI Agent Limitations Explained 41:00 Relay's $2M ARR with 10-Person Team 49:11 Why PMs Are Losing to Sales/Marketing Teams 54:52 Choosing Between Agent Platforms 1:00:40 55-Agent Marketing Team Breakdown 1:06:40 Building AI Into Your Product Strategy 1:11:41 Why Jacob Left Google Director Role 1:15:36 Brutal Truth About PM-to-Founder Transition 1:19:09 Outro ---- 🏆 Thanks to our sponsors: 1. Miro: The innovation workspace is your team's new canvas: https://miro.com/innovation-workspace/?irclickid=yIg1Kj2P2xycUXeyopwbUQf0UkpwPezrCXtgyg0&irgwc=1 2. Jira Product Discovery: Plan with purpose, ship with confidence: https://www.atlassian.com/software/jira/product-discovery 3. Mobbin: Discover real-world design inspiration: http://mobbin.com/aakash 4. Product Faculty: The #1 AI PM certificate (Get $550 off): https://maven.com/product-faculty/ai-product-management-certification?promoCode=AAKASH550C7 ---- Transcript: https://www.news.aakashg.com/p/jacob-bank-podcast ---- 10 steps to 10x your productivity with agents: 1. Catch Up on Adoption Only 2% of PMs have built an AI agent, yet Jacob Bank runs his entire marketing operation with 55 agents. Go-to-market teams are already leveraging AI while PMs lag behind despite considering ourselves the tech innovators. 2. Map Your Repetitive Work Audit your calendar for recurring activities like weekly bug reports, monthly competitor analysis, or quarterly updates. These repetitive tasks with common patterns are perfect for agent automation. 3. Select Your Platform For less technical PMs: Zapier, Relay.app, or Lindy ($20-100/month). Technical teams: N8N or Make. Jacob recommends experimenting with multiple tools since pricing is usage-based, not per-seat. 4. Begin with Templates Don't build from scratch. Most platforms offer proven templates for meeting follow-ups, competitive research, and metrics reporting. Copy and customize these to your specific processes. 5. Create Your Executive Assistant Build 8-12 agents for calendar management, email filtering, meeting prep, task reminders, and newsletter summaries. Jacob built this system to handle all operational overhead automatically. 6. Automate Competitive Intelligence Deploy agents to track competitor pricing, feature launches, social posts, and job postings. Configure scheduled delivery that aligns with your planning cycles, like Friday competitive updates. 7. Systematize Customer Feedback Create agents that synthesize user interviews, support tickets, social mentions, and reviews. Generate weekly digests highlighting themes, pain points, and feature requests across all touchpoints. 8. Streamline Product Operations Build agents for release notes, bug reports, metrics dashboards, and stakeholder updates. Establish automated "rhythm of business" reports that generate and deliver themselves when needed. 9. Test and Refine Start with simple prompts and 1-3 examples rather than complex instructions. Test thoroughly before removing human oversight. Maintain approval loops for high-stakes communications until you build confidence. 10. Scale Through Workflows Combine agents into sophisticated workflows. Jacob's webinar process: one calendar event triggers social posts, email campaigns, reminder sequences, and follow-up reports. Think systems, not individual tasks. ---- 👨‍💻 Where to find Jacob: LinkedIn: https://www.linkedin.com/in/jacobdlevy/ Relay: https://relay.app YouTube: @relayapp ---- 👨‍💻 Where to find Aakash: Twitter: twitter.com/aakashg0 LinkedIn: linkedin.com/in/aagupta/ Newsletter: news.aakashg.com #aiagents #productmanagement 🧠 About Product Growth: The world's largest podcast focused solely on product + growth, with over 187K listeners. Hosted by Aakash Gupta, who spent 16 years in PM, rising to VP of product, this 2x/week show covers product and growth topics in depth. 🔔 Subscribe and turn on notifications to get more vidoes like this.

Aakash GuptahostJacob Bankguest
Sep 19, 20251h 19mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 2:08

    Why AI agents feel like the future (and what you'll see today)

    Aakash tees up a live look at AI agents with Relay’s founder, promising practical demos rather than hype. Jacob previews three agents PMs can build and use immediately, setting the tone around leverage and real productivity gains.

    • AI agents as a step-change vs traditional productivity tools
    • Episode focus: real, working agents (not theory)
    • Preview of an executive assistant, follow-up drafter, and brand tracking
    • Framing: savings vs hiring and time reclaimed
  2. 2:08 – 3:15

    12-agent executive assistant overview: replacing an EA with workflows

    Jacob introduces his “executive assistant” composed of 12 coordinated agents across calendar, email, and task management. They discuss the cost comparison versus paid human/virtual assistants and why this is a customizable skill PMs should learn.

    • 12 agents grouped into calendar/email/task management
    • Cost comparison: $20–$40/month vs thousands for an assistant
    • Agents are built incrementally as needs emerge
    • Customization is the core advantage (timing, channels, data sources)
  3. 3:15 – 7:21

    Meeting Briefing Generator: research + context delivered before calls

    They walk through Jacob’s Meeting Briefing Generator, which prepares a dossier for each attendee and sends a combined briefing to Slack 30 minutes before meetings. Jacob shows an example briefing for this very podcast recording and explains personalization rules for first-time vs repeat contacts.

    • Automated guest research: emails, past meetings/notes, LinkedIn
    • Context-aware dossier logic (new contact vs familiar contact)
    • Slack delivery in a dedicated channel; timing is configurable
    • Example output includes links, prep instructions, and prior email context
  4. 7:21 – 9:31

    Model selection strategy: cost/quality tradeoffs and task-fit

    Jacob explains why he mixes models (OpenAI, Claude, Gemini) within one workflow and how those choices evolve over time. The emphasis is on building intuition, running quick tests, and swapping models as capabilities shift.

    • Different models for different tasks: writing, extraction, analysis
    • Cost-to-quality tradeoff as a first-order decision
    • Avoid rigid advice: models change quickly—run lightweight evals
    • Easy model swapping inside an agent workflow
  5. 9:31 – 10:48

    Tooling deep dive: auto-finding LinkedIn profiles via email + search

    Jacob explains a sub-workflow that discovers a likely LinkedIn URL using Google queries derived from an email address, then fetches profile data. This illustrates how small reusable building blocks power richer “assistant” behaviors.

    • Generate targeted Google queries from work email + ‘LinkedIn’
    • Programmatically execute searches and pick best-match URL
    • Fetch structured LinkedIn profile data for the dossier
    • Reusable sub-workflows as composable agent components
  6. 10:48 – 13:09

    Follow-Up Drafter: turning meeting transcripts into draft emails (with review)

    Jacob demos an agent that triggers when Fireflies creates a transcript, decides whether a follow-up is appropriate, gathers attendee emails from Google Calendar, and drafts a concise follow-up in Gmail. Both emphasize keeping a human-in-the-loop for high-stakes external communication.

    • Trigger: Fireflies transcript creation
    • AI gate: determine whether follow-up is appropriate (no-show/internal vs customer/prospect)
    • Enrich with attendee emails via Google Calendar event lookup
    • Output as Gmail draft to preserve human final review
  7. 13:09 – 18:58

    Prompting philosophy for follow-ups + adding extra context (recent emails)

    They highlight that simpler prompts with a few good examples often outperform long, complex instructions. Jacob then shows how to extend the workflow to pull in recent emails between attendees after the meeting, so the draft reflects any already-sent follow-ups.

    • Simple prompt + 1–3 examples to match voice
    • Long prompts can degrade output quality
    • Add a ‘find email’ step with participant-based filters
    • Use time bounds (after event time / last X hours) to avoid noise
  8. 18:58 – 19:59

    Relay setup choices: built-in AI credits vs bringing your own API keys

    Aakash asks how model usage works in Relay. Jacob explains Relay’s default “credits” approach for non-technical users and the advanced option to connect your own provider credentials for fine-tunes, agreements, or custom data policies.

    • Default: Relay manages model accounts; users consume Relay credits
    • Supports major providers (OpenAI, Anthropic, Gemini, etc.)
    • Advanced: connect your own API key for special needs
    • Positioning: accessible to non-technical builders
  9. 19:59 – 24:37

    Competitor pricing tracker: monthly scraping + change detection alerts

    Jacob introduces a competitive intelligence agent that scrapes pricing pages on a schedule, summarizes plans into a spreadsheet, and flags material changes in Slack. The key value is not just time savings, but enabling research you usually never get around to doing.

    • Runs on a schedule (monthly by default; can be weekly/daily)
    • Scrapes competitor pricing pages and summarizes into Google Sheets
    • Compares last month vs this month to detect material changes
    • Slack alerting for noteworthy updates (e.g., new low-tier plan)
  10. 24:37 – 29:30

    Live build: Reddit brand tracker agent (search → summarize → email report)

    They build a brand-mentions tracker from scratch that searches Reddit weekly, summarizes sentiment and themes with an LLM, and emails a formatted report with links. Jacob argues Reddit is a crucial input to LLM “brand perception,” so monitoring it affects how products show up in ChatGPT-like tools.

    • Reddit as an underappreciated source that influences LLM outputs
    • Scheduled trigger + Reddit search step (query, sort, time window)
    • AI write step: sentiment, quotes, use cases, notable posts
    • Email delivery with rich-text formatting and drill-down links
  11. 29:30 – 35:48

    One-shot prompts in workflows + Jacob’s pragmatic prompt framework

    Jacob explains why prompting inside automations differs from ChatGPT: you only get one shot, and you must manage inputs/outputs carefully. He shares his lightweight framework—role/context, task guidance, and examples—then reviews a surprisingly strong report generated from a quick prompt.

    • Workflow prompting is ‘one-shot’—no iterative back-and-forth
    • Design inputs/outputs explicitly between steps
    • Framework: 1 sentence role/context + 2–6 task sentences + 1–2 examples
    • Example output: sentiment split, pain points, use cases, actionable quotes
  12. 35:48 – 37:57

    Managing notification overload: cadence, batching, and digest agents

    Aakash raises the concern that agents create more pings and review work. Jacob explains how he schedules reports to match weekly rhythms and uses additional agents to aggregate information (e.g., daily newsletter digests) so output stays manageable.

    • Put scheduled reports on predictable days aligned to team cadence
    • Ad-hoc agents fit naturally into the flow after a triggering action
    • Build meta-agents that aggregate/summarize other incoming info
    • Example: daily newsletter digest delivered at a set time
  13. 37:57 – 40:49

    Limitations of agents: workflows vs autonomy, and when to keep humans in the loop

    Jacob clarifies that today’s agents struggle with complex autonomous work and that most successful real use is still workflow-based. He offers a practical decision framework using two axes—AI reliability and task stakes—to decide when to automate fully vs require approval.

    • Workflow (predefined steps) vs agent (goal + tools + autonomy) spectrum
    • Most users succeed faster with workflows than fully agentic setups
    • Human-in-loop for high stakes (customer emails, public posts)
    • Two-axis framework: AI quality × stakes to set autonomy level
  14. 40:49 – 49:11

    Relay’s traction and the ‘smaller teams’ thesis (10 people, thousands of customers)

    Jacob shares Relay’s customer profile, funding, and team size, then argues the future belongs to smaller, high-leverage teams. He explains why he doesn’t feel capital-constrained and how big-company coordination costs, not raw work, are often the real bottleneck.

    • Customer base: low thousands; mix of tech non-technical roles + SMBs
    • Funding: $8.8M seed (Coastal Ventures + a16z)
    • Team: ~10 people; heavy use of internal agents for leverage
    • Thesis: scale to far more customers without linear headcount growth
  15. 49:11 – 54:53

    Who’s adopting agents fastest—and why PMs are behind

    They discuss market adoption, with Jacob arguing go-to-market functions (support, sales, marketing) lead AI agent usage. He challenges PMs to catch up and offers a mental model: everyone will use chatbots, copilots, and agents—choose the right modality per job.

    • Fastest adoption: support, sales, marketing (clear workflows + metrics)
    • PM adoption lags due to fewer repetitive patterns and less experimentation
    • Three modalities: chatbot vs copilot vs agent depending on the task
    • Heuristic: recurring weekly tasks are prime targets for agents
  16. 54:53 – 1:00:35

    Choosing agent platforms: technical vs non-technical, and how Relay compares

    Aakash asks how product leaders should pick tools across Relay, Lindy, Make, n8n, and Zapier. Jacob recommends letting teams experiment since pricing is usage-based, then outlines segmentation by technical comfort and contrasts Relay’s workflow-first simplicity with Lindy’s broader ‘agentic’ feature set.

    • Recommendation: let people try multiple tools; avoid forced standardization early
    • Primary axis: technical (n8n/Make) vs non-technical (Zapier/Relay/Lindy)
    • Zapier as incumbent with broad integrations but older UX constraints
    • Relay: simpler workflow building/testing; Lindy: more built-in agentic features
  17. 1:00:35 – 1:04:53

    55-agent ‘one-man marketing team’: how agents run webinars and content ops

    Jacob breaks down his viral “AI agent org chart,” explaining how ~55 agents support channel-by-channel marketing. He walks through a webinar workflow where agents handle promotion, signup, reminders, attendance tracking, recording follow-ups, and repurposing content—leaving him to pick a topic and show up.

    • Marketing agents organized by channel (LinkedIn, X, Reddit, YouTube, email, webinars, partners)
    • Webinar flow: calendar event → landing page → promo posts/emails → reminders → recording → repurposed content
    • Automation reduces coordination overhead vs traditional marketing teams
    • System built incrementally over time as pain points appeared
  18. 1:04:53 – 1:11:41

    Building AI into products + MCP: becoming a ‘tool’ for other agents

    They pivot from personal productivity to product strategy: AI experiences will become table stakes, but the right modality depends on the product (chatbot, copilot, or agent). Jacob explains MCP as a way to expose product actions as tools for agents, encourages systems-of-record to invest in APIs/MCP, and notes MCP is still early for broad real-world success.

    • Re-think products with AI: chatbot for search, copilot for canvases, agents for automations
    • MCP concept: expose product actions as agent-usable tools
    • Strategic advice: systems of record should strengthen APIs and MCP servers
    • Reality check: MCP adoption is still mostly technical/early-stage
  19. 1:11:41 – 1:18:49

    Jacob’s founder story: leaving Google, and the brutal PM-to-founder truth

    Jacob explains why he left a Director PM role at Google Workspace: personal growth and belief that cross-product AI workflows couldn’t be built inside incumbents. He closes with candid advice for aspiring PM founders: expect rapid growth paired with frequent demoralization, and recognize that big-tech PM work can be poor preparation for starting from zero.

    • Motivations: growth beyond big-company leadership + startup-only vision execution
    • Incumbent bias toward improving existing UIs vs cross-tool workflows
    • Founder reality: loss of status, rebuilding from nothing, constant uncertainty
    • Hot take: large-company PM is weak founder prep—learn from small businesses
  20. 1:18:49 – 1:19:50

    Wrap-up and call to subscribe

    Aakash closes the episode, thanking Jacob and encouraging listeners to subscribe, follow, and leave reviews to support the show. The ending reinforces the practical value of the agent demos and the broader shift toward agent-driven work.

    • Final thanks and episode recap energy
    • Subscribe/follow on major platforms
    • Ratings/reviews help distribution and production quality
    • Teaser for future episodes

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.