The Autonomous Business OS - growing list of AI agents

3 founders. 20+ AI agents. $10M ARR per employee target. For the first time, I'm sharing our complete AI agent stack that's helping us get to $30M without a single hire.

Most startups think about scale in terms of org charts and hiring plans. Their solution to every growth challenge? Add more bodies to the equation.

We've flipped that model on its head. Instead of building departments, we're building an intelligence network that lets each founder operate at enterprise scale.

In the last 30 days alone, we've:

- Generated $1M+ in qualified pipeline

- Onboarded 25 customers (white glove)

- Shipped 20+ major features

- Resolved 500+ support tickets

Let me pull back the curtain on each founder's personal AI taskforce.

The GTM Swarm

🕵️‍♂️ The Observer

Purpose: Surfaces high-intent ICP leads by analyzing engagement on LinkedIn posts (yours + competitors)

Stack: Make + Unipile + Anthropic

Workflow Summary:

• Pulls all comments and reactions from your recent LinkedIn posts

• Flags repeat engagers and first-timers

• Assesses engagement velocity and sentiment

• Cross-checks each profile against your ICP definition

• Enriches with LinkedIn bio and role data

• Shares interesting interactions in Slack with context

• Logs all interactions in a database for pattern tracking

Output/Outcome:

Slack feed of warm leads showing intent, ready for conversation.

Adoption Tips:

• Be mindful of LinkedIn rate limits

• Regularly refine your “interesting signal” criteria

Sneak-peek:

🤝 The Connector

Purpose: Filters and responds to inbound LinkedIn connection requests with ICP-aware personalization

Stack: Make + Unipile + Anthropic

Workflow Summary:

• Pulls last 100+ incoming LinkedIn requests

• Scores each profile based on your custom persona rules

• Accepts all to grow reach

• Sends personalized DMs to top prospects

• Staggers actions to avoid LinkedIn automation detection

Output/Outcome:

Clean inbox, growing network, warm conversations started.

Adoption Tips:

• Personalize your DMs — don’t spray

• Watch your pacing to protect your LinkedIn account

Sneak-peek:

🧙‍♂️ The Hunter

Purpose: Turns anonymous website visitors into qualified outreach targets — autonomously, in real time

Stack: Swan AI

Workflow Summary:

• Assigns an AI agent to each non-converting visitor

• Deanonymizes visitor at person + company level

• Enriches, qualifies, and segments by ICP type

• Identifies buying committee and finds contact info

• Sends personalized LinkedIn messages from your team’s profiles

• Pushes data to HubSpot, Slack, LinkedIn automation, webhook

• Controlled entirely from Slack

Output/Outcome:

Qualified, enriched leads dropped in Slack — already messaged, already in your CRM, already in motion.

Adoption Tips:

• Refine your ICP and outreach flows by segment

• Treat Slack as your GTM command center

Sneak-peek:

🛡 The Gatekeeper

Purpose: Filters inbound demo/trial requests and routes only best-fit leads to product or sales

Stack: Tally + Make + Anthropic + Generect

Workflow Summary:

• Collects essential data via Tally form

• Enriches leads using external sources

• Qualifies based on fit, traffic, industry, etc.

• Routes leads to Demo / Trial / Waitlist paths

• Assigns nurturing cadences to waitlisted leads

Output/Outcome:

A high-signal inbound funnel — cleanly routed and conversion-optimized.

Adoption Tips:

• Keep the form short — let enrichment do the work

• Make sure it’s easy to adjust filters as GTM goals shift

Sneak-peek:

✍🏻 Shakespear

Purpose:

Helps turn unstructured thoughts into high-converting LinkedIn posts using a structured co-writing workflow

Output/Outcome:

Narratives that reflect your voice and expertise — without sounding like every other LinkedIn growth hacker.

Adoption Tips:

• Iterate section-by-section — don’t rush to full drafts

• Save all your previous posts in a ChatGPT/Claude project and work from there

Prompt:

Personality

You are a LinkedIn Content Stylist and Revision Specialist with deep expertise in tone matching, narrative consistency, and persuasive business writing.

Task

Your task is to provide thoughtful revision suggestions for the user’s LinkedIn posts, ensuring alignment with the user’s line of thinking and maintaining the style, tone, and voice demonstrated in previous posts provided by the user.

Guidelines

Don't over rely on previous's posts content, just the styling!

Your input must be collaborative, helping refine the post for clarity, engagement, and impact while staying true to the original essence and objectives of the user's strategy.

Don't jump straight into copy suggestions, start with more high level strategic advice, and then work your way down collaboratively with the user to the micro level.

Unless explicitly instructed by the user, always suggest 2 options for the user to decide.

When brainstorming about the hook, assume that mobile users can't see after the first break line.

Always assume that we're addressing a new audience, that some of my readers viewed my previous posts but most of them won't.

Execution Order & Workflow

1. Start High-Level: Suggest multiple strategic arcs before diving into copy edits.

2. Refine in Layers: First optimize structure, then tone, then wording.

3. Collaborative Iteration: Engage the user in decision-making rather than dictating changes.

4. Final Pass: Ensure clarity, engagement, and stylistic consistency before submission.

How to Write Great Hooks: A Step-by-Step Guide

A strong hook captures attention immediately and compels the reader to continue. The most effective hooks break traditional best practices by being longer, information-dense, and story-driven.

This guide outlines a proven two-line hook structure that maximizes engagement - The 2-Line Hook Formula:

Line 1: Social Proof + Personal Storytelling

The first line establishes credibility and introduces a compelling narrative.

• Social proof – Demonstrates expertise or a noteworthy achievement. Example:"I grew my company to $80K MRR using only LinkedIn."

• Personal storytelling – Adds uniqueness and authenticity. Example:"In just 40 weeks, I went from $0 to $4M ARR."

Why it works: Readers trust content backed by real results, and personal stories make the post more engaging.

Line 2: Expectation Setting

The second line clearly states what the reader will gain from the post while sparking curiosity.

• Creates intrigue – Offers a clear preview without revealing everything. Example:"If I had to start from 0 again, here are the exact 6 steps I’d follow."

Why it works: Readers understand the value of the post and are motivated to keep reading to learn more.

Key Elements of a Great Hook

A compelling hook is built on three principles:

1. Emotion – Engages the reader on a personal level.

2. Clarity – Communicates the message without fluff.

3. Curiosity – Encourages readers to continue for more details.

USER INPUT

This is what I want to write about today: [free form text]

These are my previous posts: [previous posts]

Workflow Summary:

• Starts from raw user thoughts and 2+ example posts as style references

• Suggests strategic narrative arcs before touching the copy

• Collaboratively edits structure, tone, and final wording

• Follows a defined formula for hook writing:

  • Line 1: social proof + storytelling

  • Line 2: curiosity + expectation

• Handles tone-matching, persuasive structure, and pacing

• Produces final post options in under 20 minutes

The Product Swarm

🛎 The Concierge

Purpose: Handles 70%+ of customer conversations in Slack — onboarding, support, feedback

Stack: Swan AI + Slack + PyCon

Workflow Summary:

• Lives in Slack, answers common questions instantly

• Escalates to team only when needed

• Learns from escalations and expands knowledge base

• Performs product actions (e.g. start trial, push docs)

• Generates new FAQs from escalations automatically

Output/Outcome:

Self-scaling customer support that gets smarter with every ticket.

Adoption Tips:

• Let it take real action — not just respond

• Treat every escalation as a learning opportunity

Sneak-peek:

📊 The Analyst

Purpose: Synthesizes 100+ customer conversations per month into product insights and feature requests

Stack: Make + Circleback + OpenAI

Workflow Summary:

• Auto-records and transcribes calls

• Extracts questions, challenges, requests, and “wow” moments

• Populates Notion with:

  • Common objections

  • Most requested features

  • Excitement triggers

• Informs roadmap and GTM messaging

Output/Outcome:

Clear visibility into what customers want — without watching hours of video.

Adoption Tips:

• Don’t over-engineer tagging at first

• Review insights weekly to guide decisions

👂 The Listener

Purpose: Automates CRM updates, task creation, and follow-ups after every sales call

Stack: Circleback + Make + OpenAI + Attio

Workflow Summary:

• Classifies call type (demo, lost, etc.)

• Updates pipeline stage and contact info

• Extracts action items with owners + due dates

• Opens tasks in Attio

• Drafts follow-up emails using call context

• Attaches notes + draft email to CRM record

Output/Outcome:

Clean CRM, clear pipeline, and personalized follow-ups — all on autopilot.

Adoption Tips:

• Use deal-stage rules to avoid over-triggering automation

• Tweak tone if needed, but use the draft structure as-is

🧪 The Prototyper

Purpose: Flips the traditional PRD process — starts with a working prototype to test ideas fast

Stack: v0 by Vercel

Workflow Summary:

• Ingests product ideas from internal or customer feedback

• Builds functional prototype directly in Vercel v0

• Shares with team + customers for emotional feedback

• Iterates until it feels “right”

• Backfills a PRD from the final prototype

• Ships faster with higher confidence

Output/Outcome:

Validated concepts — built emotionally, not speculatively.

Adoption Tips:

• Start with the user flow, not the feature list

• Capture feedback while it’s still messy — that’s where the signal lives

🧠 Dorothy the PM

Purpose: Acts as your product thinking partner — refining ideas and turning them into prompts or specs

Stack: Custom GPT (built-in RAG on past PRDs + product docs)

Workflow Summary:

• Uses Retrieval-Augmented Generation (not training) to reference past PRDs + product context

• Mirrors product ideas back to the PM (Ido), challenging logic and surfacing blind spots

• Suggests relevant examples or concerns based on past decisions

• Helps translate rough ideas into prototype prompts or PRD outlines

• Keeps product thinking sharp — even as a solo PM

Output/Outcome:

Sharpened product direction without needing another person in the room.

Adoption Tips:

• Speak naturally — treat it like a real teammate

• Refresh its context base regularly with new wins, launches, and learnings

🛡 Security & Privacy Expert

Purpose:

Answers security, privacy, and compliance questions instantly — without needing legal or security reviews

Stack:

Custom GPT (using built-in RAG)

Workflow Summary:

• Includes a centralized knowledge base:

  • Past vendor security questionnaires

  • Filled security/compliance docs from real deals

  • GDPR, CCPA, and other relevant privacy policies

• Lives as a dedicated GPT project, always available via chat

• First stop for customer security questions before involving legal/security team

• Instantly surfaces previous answers, definitions, and frameworks used in similar scenarios

• Ensures consistent, fast, and informed responses across the team

Output/Outcome:

Responds to 90% of common security or compliance questions. Reduces dependency on legal and security teams and accelerates sales cycles.

Adoption Tips:

• Update regularly with newly filled questionnaires and policy updates

• Can be repurposed for different compliance standards by layering in additional docs

🤖 The Creator

Purpose:

Generates complete prompts for AI agents — including role, tone, inputs, outputs, and execution steps — in structured XML

Prompt:

<AgentDescriptor> <Personality> You are a Persona-Oriented AI Roleplay Specialist and Behavioral Modeling Writer with expertise in framing intelligent systems through human-style behavioral roles. You write clear, structured, and instructional character briefs that help AI agents understand how to behave, communicate, and make decisions in service of a specific user task. </Personality>

<Task> Your task is to create XML-based instructional personas for AI agents that define their tone, domain of operation, communication patterns, and expected behaviors. You are writing directly to the AI as if briefing a human assistant on how to perform a role. These personas should clearly instruct the AI on how to think, speak, and behave—without ever referencing itself as an AI. </Task>

<Guidelines> <Guideline>Always address the AI agent directly, as if you are giving it role instructions.</Guideline> <Guideline>Do not mention that the AI is artificial or a model—write as if it is a persona being instructed for a role.</Guideline> <Guideline>Use XML to structure your output, but keep it shallow and readable for both machines and humans.</Guideline> <Guideline>Include guidance on expected input and expected output from the AI agent.</Guideline> <Guideline>Provide one archetypal example at the end to show how a persona can be structured in context.</Guideline> </Guidelines>

<ExecutionOrder> <Step>Start with role-level instructions: what the AI should become.</Step> <Step>Define tone, voice, and communication behaviors.</Step> <Step>List the expected types of inputs the agent will receive.</Step> <Step>Define the expected structure, style, and format of outputs.</Step> <Step>Wrap in XML blocks using flat, human-readable elements.</Step> <Step>Provide one example implementation at the end.</Step> </ExecutionOrder>

<Example> <Agent> <Role>You are a Call Summary Analyst with a focus on sales effectiveness.</Role> <PersonaInstruction> You are a calm, perceptive, and structured listener. You capture what matters, ignore fluff, and synthesize conversations into outcomes and actions. Speak like a consultant preparing a debrief for a sales team, not like a stenographer. </PersonaInstruction> <Tone>Professional, concise, and insightful. Avoid exaggeration or filler language. Use plain, structured English.</Tone> <Input> You will receive the full transcript of a sales call between a seller and a prospect, the required summarization framework, and an example follow up email. </Input> <Output> <Deliverable>1. A short paragraph summary of the call</Deliverable> <Deliverable>2. A bullet list of the main tasks mentioned during the call, including who is responsible for each (seller or prospect).</Deliverable> <Deliverable>3. A follow-up email draft the seller should send to the prospect to confirm the discussion and next steps.</Deliverable> <Formatting>Keep all outputs in plain text. Separate each section with a clear title.</Formatting> </Output> </Agent> </Example> </AgentDescriptor>

Workflow Summary:

• You define the purpose of a new AI agent

• The Creator returns a full XML-based instruction brief written as if it’s briefing a human assistant

• Each prompt includes:

  • Role description

  • Behavioral/tone guidance

  • Expected inputs + outputs

  • Execution steps

  • XML formatting for consistency

• Ends with a sample persona for clarity and reuse

Output/Outcome:

Clean, readable, machine-and-human friendly prompts that define how each AI agent should behave and perform — repeatable across functions.

Adoption Tips:

• Use this as your standard when spinning up new agents — it guarantees clarity and consistency

• Don’t overcomplicate the XML — clarity beats depth

🧾 Invoice Classifier

Purpose:

Automatically detects invoices in your inbox and routes them to bookkeeping

Stack:

Relevance AI + Google Drive

Workflow Summary:

• Monitors incoming emails with attachments

• Opens PDFs and classifies file contents using text analysis

• If document = invoice → saves it to shared Google Drive folder for accounting

• Runs on autopilot daily, with no manual tagging required

Output/Outcome:

Invoices never go missing. Bookkeeping gets done without founders doing grunt work.

Adoption Tips:

• Consider tagging the folder by vendor name or month to simplify reconciliation later

• Run audit checks monthly to ensure classification accuracy

⏱️ Meeting Prep Agent

Purpose:

Builds fully researched prospect summaries for every booked meeting — auto-enriched in CRM

Stack:

Cal.com + Make + Generect + Anthropic + Attio

Workflow Summary:

• Triggered automatically when someone books a call via cal.com

• Gathers prospect and company data:

  • Role + seniority

  • Buying power

  • Company ICP fit

  • GTM tech stack

  • Website traffic

  • Sales team size

• Enriches prospect record in Attio

• Pulls in past interactions and call summaries (if any)

• Posts full meeting brief to Slack or calendar invite for prep

Output/Outcome:

You join every call with full context — no research, no scramble, just signal.

Adoption Tips:

• Tailor research depth by meeting type (e.g. deep for demos, light for intros)

• Limit the scope of the research to what’s necessary and nothing more

🏗 The Architect

Purpose:

Turns product specs into dev-ready architecture and boilerplate code — instantly

Stack:

Cursor + Devin AI

Workflow Summary:

• Takes a finished PRD or approved prototype

• Suggests the ideal architecture for implementation

• Highlights decisions: data model, infra, integrations, auth, etc.

• Generates boilerplate code scaffolding for each component

• Developer just plugs into the logic, skipping setup and structure decisions

• Evolves with your tech stack — can integrate existing services and libraries

Output/Outcome:

Weeks of dev planning reduced to hours. Cleaner builds, faster iteration.

Adoption Tips:

• Keep specs tight — the clearer the PRD, the better the output

• Pair with Dorothy or The Prototyper to create an end-to-end build loop

🔁 The Repurposer

Purpose:

Transforms content from format X to format Y while preserving intent and tone

Stack:

Custom GPT

Workflow Summary:

• Starts with a source content piece (e.g. blog post, tweet, podcast transcript)

• User defines desired format (e.g. LinkedIn post, thread, blog)

• Proposes strategic direction before rewriting

• Offers 1–2 tone/style variations if unspecified

• Never translates literally — adapts for tone, platform, and attention span

Output/Outcome:

Platform-native, high-converting content that still sounds like you — not a robotic summary.

Adoption Tips:

• Use repurposing for both expansion (blog → thread) and condensation (webinar → hook)

• Bring strong inputs (even if messy) — it thrives on raw thought, not polish