Podcast Summary

AI & I with Dan Shipper:

How We Built 'Claudie,' Our AI Project Manager (Full Walkthrough)

Date: February 4, 2026
Guests: Natalia (Head of Consulting, Every)

Episode Overview

This episode offers an actionable, in-depth walkthrough of how the team at Every built and implemented “Claudie”—their custom AI-powered project manager. Host Dan Shipper interviews Natalia, Every's Head of Consulting, about the process of creating Claudie, the essential patterns for company-wide AI adoption, lessons learned from deploying AI at leading organizations, and the broader transformation AI is driving in how teams work. The conversation includes a detailed, screen-shared demo of Claudie and practical insights for anyone interested in building agent-powered business workflows.

Key Discussion Points & Insights

1. AI Adoption Patterns from the Front Lines

• Top-Down Commitment:
  • Durable AI transformation requires leadership engagement, not just lip service.
  • “You’ll probably go as far in terms of AI adoption as your CEO has gone.” — Dan [04:56]
• Empowering AI Champions:
  • Early adopters within organizations should be elevated and their learning widely shared.
  • “Your job as an executive... is to identify those people and spread what they know and elevate their status so that they can kind of pave the path for everyone else.” — Dan [05:29]
• Coordinated Efforts vs. Isolated Initiatives:
  • Without an organized company-wide approach, AI implementation devolves into pockets of super-users and general inertia.
  • “For AI to be useful to a company, it needs to be a coordinated effort.” — Natalia [02:19, 00:00]

2. Real-World Case Studies: Unlocks & Roadblocks

• Private Equity Firm: Workflow Mapping to Drive High-Leverage AI Use
  • In-depth task mapping: Detailed documentation of every investor task across teams allowed precise pinpointing of high-leverage AI opportunities.
  • “We ended up with a map so detailed we could be very, very precise about looking for solutions that would not just give the team bandwidth, but really high leverage.” — Natalia [08:00]
  • Result:
    • AI-generated investment memo drafts in 30 minutes, instead of 2–3 weeks.
    • “Now you get a really solid draft in like 30 minutes.” — Natalia [13:04]
• Tech Company Engineering Org: Missing the Planning Phase
  • Engineers were great at delegating and assessing tasks with AI, but omitted upfront planning—limiting overall impact.
  • Simple enablement around planning unleashed significant acceleration.
  • “You can only really compound as much as you plan.” — Natalia [17:16]
  • Result:
    • “We’re frequently seeing engineers generate two weeks of work effectively in an afternoon.” — Natalia [17:59]

3. How Claudie Was Built: Framework, Iteration, & Tactics

• The “Vibe Code Addict” Ethos:
  • Progress accelerated by carving out creative, protected time (“vibe coding” outside regular hours) to experiment with agent-based workflows.
  • “We decided to start our day three hours early. So we would meet at 6 am and basically just vibe code from 6 to 9 am.” — Natalia [19:36]
  • Multiple full restarts and iterative learning were necessary: “We got 85% of the way three times and then had to scrap it.” [19:36]
• Pairing Domain Experts with Applied AI Engineers:
  • Successes came from combining intimate knowledge of “what good looks like” (project management nuance) with leading-edge agent building skills.
  • “It wasn’t until we realized it’s a mix of knowhow...and all this Claude code infrastructure that it really came together and worked.” — Natalia [28:51]

4. Claudie Demo: Architecture & Live Walkthrough

[30:06+]

• System Architecture:
  • All instructions and context are stored in a Claude MD file (“the job description”), read every time Claudie acts.
  • Commands structure Claudie’s activities; tasks manage dependencies and trigger sub-agents for QC (quality checks).
  • Data sources include direct connections to Gmail, Calendar, Google Drive, and meeting transcripts.
  • “The most important part is the data sources: we enabled mcps that connect to Gmail, to Calendar, to Google Drive, and then to the meeting transcripts for the work that we do.” — Natalia [31:28]
• Job Description File & Principles:
  • Always identifies core context: who the agent is, who it reports to, where core data lives, and company conventions.
  • Important best practices built in (data accuracy, proactiveness, formulas over manual entry, escalation).
  • “Claude is really smart…But these boundaries, conventions, and sort of sharp edges have really allowed Claudie to do really good work for us.” — Natalia [34:00]
• Client Onboarding Example:
  • When onboarding a new client, Claudie launches multiple sub-agents to collect all necessary context from email, calendar, drive, and transcripts, then auto-populates a project dashboard.
  • “You just launched four sub agents to, like, look through your Gmail, look through your calendar, look through your drive, look through your meetings to, like, get context on the project.” — Dan [38:35]
  • “The only thing that’s crazier is that the alternative to Claudie doing this is me doing this.” — Natalia [39:01]

5. Impact: Real Productivity Gains & Cultural Shifts

• Time Savings:
  • Project management work dropped from 10–15 hours/week to 1 hour/week.
  • “On any given week I spend at least 10 to 15 hours on just project management. Now with Claudie, I’m collecting information for an hour a week.” — Natalia [44:36]
• Changing the Nature of Work:
  • Agents free up human time for interpersonal, creative, and higher-leverage activities.
  • “Any hour that I am not spending tabulating information, I am spending with the people that I get to work with.” — Natalia [40:30]
• Lessons for Organization Design:
  • Play and experimentation must be structurally supported—true adoption comes from risk-free, unstructured time to “play with the car” rather than race the horse-drawn buggy.
  • “You can’t learn to drive a car until you take some time out of your horse-and-buggy race to be like, what is this car thing?” — Dan [24:34]

Notable Quotes

• “For AI to be useful to a company, it needs to be a coordinated effort.” — Natalia [00:00]
• “You’ll probably go as far in terms of AI adoption as your CEO has gone.” — Dan [04:56]
• “The only thing that’s crazier is that the alternative to Claudie doing this is me doing this.” — Natalia [39:01]
• “We’re frequently seeing engineers generate two weeks of work effectively in an afternoon.” — Natalia [17:59]
• “Any hour that I am not spending tabulating information, I am spending with the people that I get to work with.” — Natalia [40:30]
• “On any given week I spend at least 10 to 15 hours on just project management. Now with Claudie, I’m collecting information for an hour a week.” — Natalia [44:36]

Timestamps for Major Segments

• 00:00 – 02:19: Patterns of successful AI adoption in companies
• 08:00 – 13:05: Case study: Private equity firm’s AI transformation
• 16:10 – 17:59: Case study: Engineering organization and the planning gap
• 19:36 – 28:51: Building Claudie: vibe coding, iteration, pairing with engineers
• 30:06 – 39:09: Claudie system deep dive & live onboarding demo
• 39:20 – 44:36: Impact: role shift, resistance, time savings, dashboard automation

Conclusion & How to Connect

This episode is a must-listen for anyone leading or supporting AI-powered business transformation. For more on Every’s consulting or to connect with Natalia, visit every.to/consulting. Finance leads can reach out to Brooker Belcourt, and tech companies can contact Every’s tech team via the same link.