Idmen vs. context files
One document describing how you work, what you do and which projects count, attached by someone to every question. That’s how most organizations do it today, and for the first few weeks it works.
Four things a file can’t do
The file is a snapshot of the day someone wrote it, and your work keeps moving after that.
- AI-briefing.docxupdated 4 months ago
- 9 projects completed
- 2 workflows revised
- 1 colleague left
Outdated before anyone opens it
The clock starts the moment someone saves it. New projects come in, you revise a workflow, a colleague leaves. The file knows nothing about it, and the answer doesn’t show it came from the old version.
- AI-briefing.docx
ChatGPTevery session
Claudeevery session
Geminievery session
Someone has to attach it every time
In every session, in every tool, by every colleague. Twelve people using three tools means thirty-six places where the same explanation has to go again whenever something changes. After the second time, nobody does it.
- What’s in it
What we do · our sectors · our writing style
What’s not in it- why that approach failed there
- what really went on in the meeting
- which version was final
The most valuable part can’t be written down
A file holds what one person remembered to type. Why an approach failed at that client, what really went on in the meeting, which version became final: that lives in nine years of projects and in the heads of the people who worked on them. Four pages can’t capture it.
- Examples of past work.docx
shared with: everyone
- Meerdink
- Bramerveld
- Van Duren
- client names, in the appendix
A single file has no permissions
As soon as it includes examples from earlier work, it includes client names. Whoever has the file has everything. There’s no version for people on that project and another for people who aren’t.
Everyone uses AI, each on their own
A project lead found a way in Claude to turn meeting notes into findings in twenty minutes. An employee keeps the best prompts in a document on their desktop. A team lead built a custom GPT with eight PDFs in it. All three work faster than six months ago, and the organization notices none of it.
Those twenty minutes stay with one person. The rest of the team never hears about it and reads the same kind of notes by hand. On every new project, everyone starts by hunting down the same files again.
Meanwhile, more tools keep arriving. The context sits in six places, and the work that matters is buried in chats nobody rereads. You’re not missing a tool or a file. You’re missing a layer underneath that all three can draw from.
- Sanneproject lead
- Workflow in Claude
- Notes to findings
- Marijkeemployee
- Prompts on the desktop
- 3 loose PDFs
- Joristeam lead
- Custom GPT
- 8 documents inside
Three setups. None of them knows about the other two.
The interface can change, the layer underneath stays
Your project lead works in a different app than your team lead, and another colleague in a third. They don’t need to match. They only need to draw from the same memory.
What stays fixed is the layer underneath: the memory, the permissions, the integrations and the documented workflows. The model and the interface sit on top and can change every year. If your organization switches to another model, it keeps its context, its rules and its way of working.
Anyone who joins doesn’t start from zero. They inherit the integrations, agreements and workflows that have already been tested, and with them what the organization learned over nine years.
- ChatGPT
- Claude
- Gemini
Teams
- Memory
- Permissions
- Integrations
- Workflows
It reads where the work already lives
SharePoint, Teams, the inbox. Not a file someone has to update, but a connection to where the work happens anyway.
New work flows in on its own
A finished report, meeting notes, a delivered project: they get processed and become part of memory. Without anyone having to think about it.
Contradictions surface
We find and resolve documents that contradict each other, instead of sending both along with equal confidence.
The workflow is documented, not stuck in someone’s head
Which sources count, in what order, what gets checked before anything goes out. We document it once and keep it up to date.
A context file, and a layer that keeps up
- IdmenA context file
- Stays currentYes, reads along where the work livesNo, outdated from the moment it’s saved
- MaintenanceContinuous, by PimManual, forever, by someone doing it on the side
- How it reaches the AIRetrieved through MCP, every sessionAttached again, per tool and per colleague
- CoverageAll of your tools and systems, cleaned up and structuredWhat one person remembered to type
- PermissionsPer client and per projectWhoever has the file has everything
- CitationsOn every line, back to the documentNone, you have to take it on faith
- ContradictionsFound and resolvedYou sort it out
- Grows with youEvery project makes it smarterFrozen from the moment it’s saved
A custom GPT or a Claude project with documents underneath is the same setup in different packaging: someone once picked those files, and nobody refreshes them. The problem isn’t the format. The problem is that nobody owns it.
When a file is enough
A context file is fine if…
your situation rarely changes, you work in one tool and it’s about your own work. A freelancer or a team of three gets far with it, and it costs nothing.
Idmen is for when…
it’s about more people’s knowledge than your own, and nobody updates the file anymore. When confidentiality and permissions apply, and someone still has to maintain it in month three.
Start with the file if you like. Switch once nobody updates it anymore.