“Document or die”: why knowledge management should start on day one

“Document or die”: why knowledge management should start on day one

When you're spending more time explaining what to do than actually doing the work, you've got a knowledge-management problem.

— Clément Soens, Senior Data Scientist

When you're spending more time explaining what to do than actually doing the work, you've got a knowledge-management problem.

— Clément Soens, Senior Data Scientist

When you're spending more time explaining what to do than actually doing the work, you've got a knowledge-management problem.

— Clément Soens, Senior Data Scientist

What knowledge management really means

When most people hear knowledge management, they picture dusty SharePoint folders or a 300-page “process bible.” Clément Soens, who spends his days turning raw data into enlightening answers, sees it differently:

It begins at the lowest level: every function gets a description. Then every script explains how it talks to the next. Stack those layers, and suddenly a stranger can run your whole project without panicking.

Think bottom-up scaffolding rather than top-down bureaucracy. A single undocumented function is a loose brick, enough loose bricks and the whole product wobbles.

At Agilytic, it’s two sides of the same coin

  1. Human-to-Human Sharing

    • Tech Talks: 30-minute lightning sessions where anyone walks the team through a new trick, a near-miss, or a hard-won insight.

    • Onboarding Playbook: a first-week checklist so complete “you could start while everyone else is on vacation.”

  2. Structured, Searchable Knowledge

    • GitHub for code (the universal CTRL + Z).

    • Centralized, searchable & collaborative document database for everything else: project retros, proposal templates, even the technical cheat-sheets.

      • The usual suspects here are Notion, Confluence, Coda or Microsoft’s Loop. But you must choose one, not five!

The magic happens when these two halves meet. A dazzling Tech Talk is great, but dropping the slide deck and transcript into Notion is what keeps the insight alive six months later.

I have a data scientist’s perspective on knowledge management. But the practice extends far beyond code. It's about capturing the skills and knowledge acquired during projects, whether that's reusable functions or business expertise. At Agilytic, we document all project retrospectives in a Notion database, which becomes incredibly valuable for future work.

— Clément

Why startups need it more than giants

Corporate folklore says knowledge management is for banks and big business. Clément disagrees:

In a startup nobody has 20 years of context. If you don’t write it down, you recreate the same wheel every sprint.

Early habits pay compound interest:

Without KM

With KM

“Ask Max how to set up VS Code.”

30-second search in the collaborative workspace.

Proposal reviews focus on format, not strategy or depth.

Review focuses on content and value, not line spacing.

Onboarding = shadowing the only expert.

Day-one resilience, even if the expert gets into an accident tomorrow.

Breaking down silos before they spring up

Silos aren’t just political; they’re technical. Clément illustrates this with error codes. Where in one IT department, error code 949 meant “missing value”, but in the marketing department it meant “unknown customer” :

Same number, different worldviews. Without shared documentation, you’ll never know whose reality is right.

The cure: a single source of truth that both sides can edit, and are expected to read before pinging a colleague.

How to tell you’re late to the knowledge management party

  • Signal #1: Leaders spend review meetings nit-picking form instead of debating content.

  • Signal #2: Seniors answer the same “Where’s that thing?” question twice in one week.

  • Signal #3: Rollbacks take longer than re-writing from scratch.

If any of these sound familiar, block two hours this Friday. Start a living doc:

  1. Write today’s biggest “tribal knowledge” risk.

  2. Add one-paragraph how-to.

  3. Link it somewhere obvious.

Repeat next Friday. Even if there’s a lot to do, momentum beats an endless to-do list.

Quick-start toolkit

Need

Tool

Why It Matters

Version control

GitHub, GitLab or any equivalent

Reverts, reviews, history: your code’s black box recorder.

Project retros

Collaborative workspace/database

Re-use estimates, pitfalls, and client tips instantly.

Team awareness

Monthly Tech Talk, with replays available in Notion.

Keeps quiet wins from dying in MS Teams.

Searchable assets

Organized databases, AI Summaries, tags (in that order)

Controlled vocabulary beats a tag free-for-all every time.

The payoff

  • Resilience: Projects survive turnover.

  • Efficiency: Answers in 30 seconds, not 5 minutes of playing the hot potato game with “where’s that document?”.

  • Autonomy: Fewer interruptions, deeper focus.

  • Valuation Bonus: A tidy knowledge base is diligence gold when you’re fundraising or selling.

What knowledge management really means

When most people hear knowledge management, they picture dusty SharePoint folders or a 300-page “process bible.” Clément Soens, who spends his days turning raw data into enlightening answers, sees it differently:

It begins at the lowest level: every function gets a description. Then every script explains how it talks to the next. Stack those layers, and suddenly a stranger can run your whole project without panicking.

Think bottom-up scaffolding rather than top-down bureaucracy. A single undocumented function is a loose brick, enough loose bricks and the whole product wobbles.

At Agilytic, it’s two sides of the same coin

  1. Human-to-Human Sharing

    • Tech Talks: 30-minute lightning sessions where anyone walks the team through a new trick, a near-miss, or a hard-won insight.

    • Onboarding Playbook: a first-week checklist so complete “you could start while everyone else is on vacation.”

  2. Structured, Searchable Knowledge

    • GitHub for code (the universal CTRL + Z).

    • Centralized, searchable & collaborative document database for everything else: project retros, proposal templates, even the technical cheat-sheets.

      • The usual suspects here are Notion, Confluence, Coda or Microsoft’s Loop. But you must choose one, not five!

The magic happens when these two halves meet. A dazzling Tech Talk is great, but dropping the slide deck and transcript into Notion is what keeps the insight alive six months later.

I have a data scientist’s perspective on knowledge management. But the practice extends far beyond code. It's about capturing the skills and knowledge acquired during projects, whether that's reusable functions or business expertise. At Agilytic, we document all project retrospectives in a Notion database, which becomes incredibly valuable for future work.

— Clément

Why startups need it more than giants

Corporate folklore says knowledge management is for banks and big business. Clément disagrees:

In a startup nobody has 20 years of context. If you don’t write it down, you recreate the same wheel every sprint.

Early habits pay compound interest:

Without KM

With KM

“Ask Max how to set up VS Code.”

30-second search in the collaborative workspace.

Proposal reviews focus on format, not strategy or depth.

Review focuses on content and value, not line spacing.

Onboarding = shadowing the only expert.

Day-one resilience, even if the expert gets into an accident tomorrow.

Breaking down silos before they spring up

Silos aren’t just political; they’re technical. Clément illustrates this with error codes. Where in one IT department, error code 949 meant “missing value”, but in the marketing department it meant “unknown customer” :

Same number, different worldviews. Without shared documentation, you’ll never know whose reality is right.

The cure: a single source of truth that both sides can edit, and are expected to read before pinging a colleague.

How to tell you’re late to the knowledge management party

  • Signal #1: Leaders spend review meetings nit-picking form instead of debating content.

  • Signal #2: Seniors answer the same “Where’s that thing?” question twice in one week.

  • Signal #3: Rollbacks take longer than re-writing from scratch.

If any of these sound familiar, block two hours this Friday. Start a living doc:

  1. Write today’s biggest “tribal knowledge” risk.

  2. Add one-paragraph how-to.

  3. Link it somewhere obvious.

Repeat next Friday. Even if there’s a lot to do, momentum beats an endless to-do list.

Quick-start toolkit

Need

Tool

Why It Matters

Version control

GitHub, GitLab or any equivalent

Reverts, reviews, history: your code’s black box recorder.

Project retros

Collaborative workspace/database

Re-use estimates, pitfalls, and client tips instantly.

Team awareness

Monthly Tech Talk, with replays available in Notion.

Keeps quiet wins from dying in MS Teams.

Searchable assets

Organized databases, AI Summaries, tags (in that order)

Controlled vocabulary beats a tag free-for-all every time.

The payoff

  • Resilience: Projects survive turnover.

  • Efficiency: Answers in 30 seconds, not 5 minutes of playing the hot potato game with “where’s that document?”.

  • Autonomy: Fewer interruptions, deeper focus.

  • Valuation Bonus: A tidy knowledge base is diligence gold when you’re fundraising or selling.

What knowledge management really means

When most people hear knowledge management, they picture dusty SharePoint folders or a 300-page “process bible.” Clément Soens, who spends his days turning raw data into enlightening answers, sees it differently:

It begins at the lowest level: every function gets a description. Then every script explains how it talks to the next. Stack those layers, and suddenly a stranger can run your whole project without panicking.

Think bottom-up scaffolding rather than top-down bureaucracy. A single undocumented function is a loose brick, enough loose bricks and the whole product wobbles.

At Agilytic, it’s two sides of the same coin

  1. Human-to-Human Sharing

    • Tech Talks: 30-minute lightning sessions where anyone walks the team through a new trick, a near-miss, or a hard-won insight.

    • Onboarding Playbook: a first-week checklist so complete “you could start while everyone else is on vacation.”

  2. Structured, Searchable Knowledge

    • GitHub for code (the universal CTRL + Z).

    • Centralized, searchable & collaborative document database for everything else: project retros, proposal templates, even the technical cheat-sheets.

      • The usual suspects here are Notion, Confluence, Coda or Microsoft’s Loop. But you must choose one, not five!

The magic happens when these two halves meet. A dazzling Tech Talk is great, but dropping the slide deck and transcript into Notion is what keeps the insight alive six months later.

I have a data scientist’s perspective on knowledge management. But the practice extends far beyond code. It's about capturing the skills and knowledge acquired during projects, whether that's reusable functions or business expertise. At Agilytic, we document all project retrospectives in a Notion database, which becomes incredibly valuable for future work.

— Clément

Why startups need it more than giants

Corporate folklore says knowledge management is for banks and big business. Clément disagrees:

In a startup nobody has 20 years of context. If you don’t write it down, you recreate the same wheel every sprint.

Early habits pay compound interest:

Without KM

With KM

“Ask Max how to set up VS Code.”

30-second search in the collaborative workspace.

Proposal reviews focus on format, not strategy or depth.

Review focuses on content and value, not line spacing.

Onboarding = shadowing the only expert.

Day-one resilience, even if the expert gets into an accident tomorrow.

Breaking down silos before they spring up

Silos aren’t just political; they’re technical. Clément illustrates this with error codes. Where in one IT department, error code 949 meant “missing value”, but in the marketing department it meant “unknown customer” :

Same number, different worldviews. Without shared documentation, you’ll never know whose reality is right.

The cure: a single source of truth that both sides can edit, and are expected to read before pinging a colleague.

How to tell you’re late to the knowledge management party

  • Signal #1: Leaders spend review meetings nit-picking form instead of debating content.

  • Signal #2: Seniors answer the same “Where’s that thing?” question twice in one week.

  • Signal #3: Rollbacks take longer than re-writing from scratch.

If any of these sound familiar, block two hours this Friday. Start a living doc:

  1. Write today’s biggest “tribal knowledge” risk.

  2. Add one-paragraph how-to.

  3. Link it somewhere obvious.

Repeat next Friday. Even if there’s a lot to do, momentum beats an endless to-do list.

Quick-start toolkit

Need

Tool

Why It Matters

Version control

GitHub, GitLab or any equivalent

Reverts, reviews, history: your code’s black box recorder.

Project retros

Collaborative workspace/database

Re-use estimates, pitfalls, and client tips instantly.

Team awareness

Monthly Tech Talk, with replays available in Notion.

Keeps quiet wins from dying in MS Teams.

Searchable assets

Organized databases, AI Summaries, tags (in that order)

Controlled vocabulary beats a tag free-for-all every time.

The payoff

  • Resilience: Projects survive turnover.

  • Efficiency: Answers in 30 seconds, not 5 minutes of playing the hot potato game with “where’s that document?”.

  • Autonomy: Fewer interruptions, deeper focus.

  • Valuation Bonus: A tidy knowledge base is diligence gold when you’re fundraising or selling.

Ready to reach your goals with data?

If you want to reach your goals through the smarter use of data and A.I., you're in the right place.

Ready to reach your goals with data?

If you want to reach your goals through the smarter use of data and A.I., you're in the right place.

Ready to reach your goals with data?

If you want to reach your goals through the smarter use of data and A.I., you're in the right place.

Ready to reach your goals with data?

If you want to reach your goals through the smarter use of data and A.I., you're in the right place.

© 2025 Agilytic

© 2025 Agilytic

© 2025 Agilytic