# How a Remote Company Builds AI Fluency — Em Shreve, Automattic

## Executive summary

Automattic, a fully distributed company, emphasizes that AI fluency requires dedicated, real-time learning rather than simply providing tools. They run intensive, two-week in-person programs for engineers, cross-functional teams, and non-technical staff. The curriculum progresses from foundational skills and responsible AI use to advanced topics like creating custom MCPs, utilizing the hook system of Codecs and Claude, and developing autonomous agents. The program is designed to make every graduate a 'multiplier' through structured knowledge sharing, local meetups, and building practical, company-adopted tools.

## Key takeaways

- AI Enablement Requires Dedicated Learning Time: The core philosophy is that simply handing out AI tools is insufficient; people need real time to learn and experiment to change their way of working. (0:55)
- The Two-Week Immersive Program Model: The program is structured as 50% facilitated learning (workshops) and 50% real project work, allowing participants to build tools or create new skills/MCPs relevant to their daily tasks. (1:15)
- Scaling Knowledge Through Multipliers: To ensure reach across the distributed company, graduates are encouraged to become 'multipliers' through a 'guides program' (champions), local AI gettogethers, and AI ride-alongs. (6:19, 7:09)

## Technical details

- AI Infrastructure & Development: The curriculum covers foundational skills, spec-driven development, and advanced concepts like creating MCPs (Micro-Capability Platforms) for new functionality. Advanced topics include utilizing the hook system of Codecs and Claude, and building autonomous agents/routines. (3:25)
- Internal Tooling and Governance: Automattic uses an internal MCP server called `context automatic` to rapidly add new tooling sets, allowing for governance and observability layers on custom MCPs. (5:24)
- Project Examples and Adoption: Successful projects built by participants include an agentic analytics system for WooCommerce, a Zendesk connector (adopted company-wide), and a shared skills directory that supports multiple tools (Codecs, Claude, etc.). (12:29)
- Measurement and Improvement: Effectiveness is measured using surveys at 72 hours, 30 days, and 90 days, tracking metrics like 'intentionality' (thoughtful AI application) and 'flow' (integration into workflows). (11:09)

## Practical implications

- Implement structured, hands-on learning programs rather than simply distributing tools.
- Design curricula that progress from foundational skills to advanced, agentic workflows.
- Establish internal platforms (like an MCP server) to govern and accelerate the adoption of new AI capabilities.
- Formalize knowledge transfer by designating 'champions' or 'multipliers' within the organization.

## Topics

AI Enablement, Build Engineering, Remote Work, Micro-Capability Platforms (MCPs), Continuous Learning, Automattic, Codecs, Claude, context automatic

Source: https://www.youtube.com/watch?v=3XI90A3ENSg