Documentation Culture: Lessons Learned from Asynchronous Management

by Sylvain Artois on Mar 29, 2026

  • #Management

Yellow Hickory Leaves with Daisy, 1928 - Georgia O’Keeffe - www.nga.gov
Yellow Hickory Leaves with Daisy, 1928 - Georgia O’Keeffe - www.nga.gov
AI-generated translation

In 2021, at LemonLearning, I hired a lot. I ended up leading a technical team partly working fully remote, including a developer based in Mexico. I made a clear choice: give people real freedom over their working hours. When a colleague is seven hours behind you, enforcing a 9-to-6 Paris schedule makes no sense.

At the same time, we were preparing for ISO 27001 certification. For those unfamiliar with it, this standard requires documenting and maintaining all company procedures related to information security: policies, risk management processes, access control procedures, incident response plans… The list of mandatory documents is long and each one must be kept up to date.

It was in this context that I started talking about document-based asynchronous management, and then, more simply, about documentation culture.

The painful reality

Making this documentation culture work in 2021 was extremely difficult. The problems were many:

  • Our meeting notes were consistently late, when they arrived at all. Their quality was poor.
  • Technical and organisational procedures were rarely updated. We would discover outdated procedures at the worst possible moment.
  • Many meetings started without a brief, without a structured agenda. The first thirty minutes were spent getting everyone up to speed — a collective waste, especially in a team spread across multiple time zones.

The problem was not people’s willingness. The problem was tooling and friction. Documenting took time, was not valued, and the available tools made nothing easier.

Those who understood it before me

Jeff Bezos and the six-page memo

In 2004, Jeff Bezos sent an email to his leadership team to announce the end of PowerPoint presentations at Amazon. Instead: a narrative memo of six pages maximum, written in full sentences, with verbs, structured paragraphs and a well-built argument.

The ritual is now famous: each meeting starts with 30 minutes of silent reading. Everyone reads the same document. Only then does the discussion begin.

The reason writing a good 4-page memo is harder than “writing” a 20-page PowerPoint is because the narrative structure of a good memo forces better thought and better understanding of what’s more important than what.

Bezos calls this decision « the smartest thing we ever did ».

GitLab, the radical example

GitLab pushes the concept even further with its handbook-first approach. Their principle: if it is not in the handbook, it does not exist. Every decision, every process, every policy must first be documented in the handbook before being communicated.

The result: more than 2,700 pages publicly accessible, maintained by over 2,100 employees across more than 60 countries, without a single physical office.

The numbers speak for themselves: companies that invest in asynchronous communication and documentation reduce time spent on tasks by 58.8%. Atlassian’s State of Teams 2024 report shows that employees working asynchronously gain 29% productivity and 53% focus capacity. A Zapier case study reports 23% additional throughput and 42% reduction in response times.

Why it matters more than ever

In the startup ecosystem, where companies often have little to offer besides uncertainty and mediocre salaries, full remote has become a major recruiting argument. Asynchronous management is no longer an edge case. It is a reality for a growing number of teams.

And then AI changed everything.

LLMs work primarily with text. A well-structured document, in markdown, with clean metadata, is exactly the format that language models handle best. Investing in your documentation culture is investing in your ability to leverage AI. A well-maintained knowledge base becomes a strategic asset that you can query, summarise, and automatically enrich.

And the good news: keeping this culture alive has never been easier. We now have plenty of tools that drastically reduce the friction I was struggling with in 2021.

1. Obsidian and Claude Cowork: the backbone

Obsidian

Obsidian is a note editor based on raw Markdown files stored locally on your disk. No proprietary format, no opaque database, no vendor lock-in. Your notes are .md files that any editor can read.

What makes Obsidian powerful for a team:

  • Bidirectional links: each note can reference other notes, creating a navigable knowledge graph.
  • Tags and YAML metadata: structure your notes with frontmatter, making filtering and searching easy.
  • The plugin ecosystem: hundreds of community plugins to adapt the tool to your needs.

For team collaboration, Obsidian Sync allows sharing a vault between collaborators (up to 20 users). Important note: for AI tool integration to work, you should not encrypt the vault. Markdown files must remain readable on disk.

Claude Cowork

Claude Cowork, launched by Anthropic in January 2026, is an agentic mode integrated into the Claude desktop application. Where the standard Claude chat answers one question at a time, Cowork handles multi-step tasks and executes them directly on your machine.

Give it access to your Obsidian vault and it becomes a remarkably efficient documentation assistant:

  • Structuring: point Cowork at a folder of raw notes and ask for a structured document — a project brief, meeting notes, a weekly summary. It reads everything, identifies the relevant elements and produces a structured first draft.
  • Enrichment: it can add tags, create links between notes, complete YAML metadata.
  • Creation: it can generate new notes from existing sources, maintain an index, or update procedures.

The combination of Obsidian + Claude Cowork turns a vault of raw notes into a living knowledge management system. This is the backbone of my documentation culture today.

2. Meeting transcriptions

Online meetings

All major video conferencing tools now include transcription features:

  • Google Meet offers automatic transcription with speaker identification.
  • Microsoft Teams provides transcripts with attribution, AI summaries and action item detection via Copilot.

These transcriptions are raw material. They are not directly usable as they are. The essential work is to process and integrate them into your information system: extract decisions made, actions to take, open questions, and inject them into the relevant Obsidian notes. This is exactly the kind of task that Claude Cowork handles very well.

And in the real world?

For in-person meetings, things are more demanding. Claude recommended several solutions, and I am currently testing Notta because of French language support. The application natively supports 58 languages, offers a dedicated mode where you simply place your phone in the middle of the table, and claims 98.86% accuracy. The free plan is very limited (3 minutes per recording), but the premium tier (1,800 minutes/month) is reasonable for professional use.

One limiting factor will be the microphone. Whatever the application, your phone’s microphone quality is the bottleneck. For regular meetings, invest in a small USB-C omnidirectional microphone placed in the centre of the table. The difference in transcription quality is significant.

That said, let us be honest: in a noisy café, no application will work miracles. A moderately quiet environment remains necessary to get usable results.

3. Technical watch

Technical watch is a permanent flow of information that feeds the documentation culture. Without a capture system, articles read, videos watched and technical discoveries get lost and forgotten.

Raindrop.io: the single entry point

I use Raindrop.io to centralise all my technical watch. It is a modern bookmark manager that does much more than save URLs:

  • The Android integration is excellent: I can save a web page, an Instagram post, a YouTube video, directly from the system share button.
  • Full-text search indexes the complete content of every saved page and PDF.
  • Permanent cache preserves a copy of pages even if the original site disappears.
  • Tagging and collections allow organising by topic, project, or priority.

Raindrop is my capture net. Everything that catches my attention goes through it, then I sort and process later.

YouTube: an underused source of technical knowledge

I use YouTube a lot for my technical watch. Conferences, tutorials, experience reports: the content is rich, but in video form, it is hard to use in a text-based documentation system.

The workflow I have set up:

  1. Get the transcript of the YouTube video (most videos have automatic subtitles).
  2. Give the transcript to Claude Cowork, which summarises the content, extracts key points and generates a structured Obsidian note with the appropriate metadata (source, date, topic, links).
  3. The note joins the vault and becomes searchable, linked to other notes on the same topic.

This workflow lends itself perfectly to automation. n8n offers ready-to-use templates that automatically chain: YouTube transcript retrieval → AI summary → Markdown note creation with frontmatter → deposit in a folder synced with the Obsidian vault.

4. Interoperability and exports

A documentation culture does not live in isolation. You need to generate diagrams, visuals, and interoperate with the team’s project management tools. This is where the MCP protocol (Model Context Protocol) changes the game.

MCP allows AI assistants to connect directly to your tools. Some concrete examples:

  • Notion MCP: your AI assistant can read and write in your Notion databases, synchronise information between your Obsidian vault and your shared Notion spaces.
  • Asana MCP: task creation, project updates, status extraction directly from your documentation environment.
  • Excalidraw MCP: generation of diagrams, architecture schemas, flowcharts from text descriptions. The diagram lives in your vault like any other note.

The idea is not to connect everything for the sake of it. It is to reduce friction between documenting and acting. If updating an architecture diagram or creating an Asana task from a meeting note takes 30 seconds instead of 5 minutes, people actually do it. That is the whole difference between a documentation culture that lives and one that dies.

The CTO’s role

Documentation culture does not install itself. It is the CTO’s role — or the technical manager’s — to keep it alive. Not by imposing it by decree, but by practising it visibly:

  • Write your own briefs before each meeting.
  • Document your technical decisions in the shared vault.
  • Use the tools in front of the team, show that it works.
  • Value documentation contributions as much as technical contributions.

The quality of thinking follows the quality of writing. A team that writes well thinks well. A team that documents well collaborates well, even with an ocean and seven hours of time difference between its members.

In 2021, I learned the pain of a missing documentation culture. In 2026, with the tools we have, not investing in this culture is a choice that is hard to justify.

Further reading

1. Amazon, The Six-Page Memo

Great memos are written and rewritten, shared with colleagues who are asked to improve the work, set aside for a couple of days, and then edited again with a fresh mind.

The reference CNBC article on Bezos’s 2004 decision. The Lex Fridman interview provides additional insight into Amazon’s narrative philosophy.

2. GitLab, How to embrace asynchronous communication for remote work

If it’s not in the handbook, it doesn’t exist.

GitLab’s complete guide on asynchronous work. More than 2,700 pages of public handbook that serve as a worldwide reference on documentation-first culture.

3. Atlassian, State of Teams 2024

A report establishing that teams working asynchronously gain 29% productivity and 53% focus capacity. The report also documents a reduction in burnout and decision-making bottlenecks in distributed teams.

4. Asynchronous Project Management: The Remote Team Playbook 2025

Documentation is the fuel for asynchronous communication. If it’s not written down, decisions don’t exist in a distributed team.

A practical playbook that includes the Zapier case study (23% additional throughput, 42% reduction in response times) and McKinsey 2025 recommendations on workflow automation.