For the complete documentation index, see llms.txt. This page is also available as Markdown.
🗺️Co-founder Responsibility Map
A free tool for co-founders and solo founders looking for one. Map who should own what, expose overlaps and gaps, and build a more efficient team with fewer avoidable conflicts.
Start with responsibilities
Most new founding teams start with two questions: who should lead, and how should responsibilities be divided? Solo founders looking for a co-founder face the same questions.
It makes sense to agree early. But even founders with relevant backgrounds, strong motivation, or experience working together have limited evidence of how they will operate this specific startup.
Early stage roles are broad. Each founder covers several functions, and responsibilities move as the company changes. Titles do not tell you enough about the work.
This map changes the order. Start with responsibilities. Decide who wants each area and who can handle it today. Surface overlaps, gaps, learning bets, and burnout risks. Then assign ownership and define when each decision should be reviewed.
A three month trial is a practical first review point. Review the map again after one year, or when a funding round, major hire, pivot, or another company event changes the work.
A starting list of 49 responsibilities common to early stage tech startups, grouped into seven functional areas. An eighth open group lets you extend the map for your business model. These are not job descriptions or titles. They are the work that needs an owner.
Each founder answers two independent questions per row, alone:
Want: do you want to own this?
Can: can you do it today, with evidence of having done it before?
Those two axes stay separate on purpose. Wanting work you cannot do is a learning bet. Being able to do work you do not want is a burnout risk. One score would hide both.
You then merge the sheets into one map. It shows where you agree, where you overlap, and what none of you covers.
Before you start
Allow 10 to 30 minutes for each founder to work alone.
Allow 60 to 90 minutes to work together, on a call or in a room.
Use it with two, three, or more founders. A solo founder can use it to define what a future co-founder needs to cover.
Answer Want and Can for every row. If Can equals Yes, write what you have actually done in the Evidence column. Leave a cell blank if you do not know. Do not tidy your answers later.
3
Do not compare answers yet
No screenshots or summaries over lunch. Once one founder sees another founder's sheet, both start negotiating instead of answering. The overlaps this tool should surface can disappear.
4
Export each sheet to CSV
Select File, Download, then Comma Separated Values.
Upload them to an AI model that can create spreadsheets (Claude, ChatGPT, Gemini, etc), together with the merge prompt below. Do this live, with all founders present.
6
Open the spreadsheet it returns
It contains a short guide, the merged map, and one protected tab per founder with answers exactly as submitted.
That conversation is the point. Everything before it was preparation.
What the merge tells you
Every row gets one of six classes. The class comes only from your answers. It decides nothing.
Class
What it means
What to do
Clear
One founder wants it and can do it.
Confirm it and move on.
Overlap
More than one founder wants it and can do it.
Discuss the trade. Someone must give up work they want and can do.
Stretch
Someone wants it, but nobody can do it yet.
Treat it as a learning bet. Decide who supports it and set a deadline.
Reluctant
Nobody wants it, but someone can do it.
Assign it if needed, but treat it as debt and set an end date.
Gap
Nobody wants it and nobody can do it.
Buy it, hire for it, or accept the risk on purpose.
Missing
Someone left an answer blank.
Answer it together before anything else. It may change the class.
Two patterns are worth finding by eye. A Stretch row where more than one founder says Want only can hide a conflict over strategic leadership. You both want work that neither of you can do yet. A Reluctant row where more than one founder says Can only is a trading chip. Nobody wants it, either founder could take it, and it may help settle an Overlap row.
What you decide, and the model does not
The merged file has three empty columns. The model never fills them. Neither should it.
Owner: who does the work. Use a founder's name, New co-founder, Hire, Outsource, or Nobody. Nobody is a valid answer. It means you accept the risk on purpose. New co-founder is not a hire. If the map shows you need another co-founder, that is a useful finding.
Status: Agreed, Parked, or Needs data. Parked matters. Without it, founders make rushed decisions just to fill the column.
Condition: what makes this ownership expire. Assume every owner is temporary. This prevents ownership from becoming permanent by default.
Rules that make it work
Fill independently. Use separate files, not hidden tabs or filtered views.
Blank is not No. An unanswered cell and a refusal are different answers. One blank makes the whole row Missing, whatever anyone else answered.
Can requires evidence. A capability claim with no track record is a hypothesis. The merged file flags Can equals Yes with no evidence in red. It never judges the evidence or compares founders.
Nothing is scored. No totals, percentages, or counts per founder. Totals distract from the decisions you need to make row by row.
Overlap is a finding, not an error. Discuss contested rows. Never allocate them automatically.
The output starts the argument. It is not a verdict.
Group 8, and why divergence is useful
The first seven groups cover responsibilities common to most tech startups. Group 8 is open. Extend it for your model, whether that is hardware, a marketplace, regulated data, field operations, or something else.
When you merge, group 8 rows are matched by text, not number. If founders added different rows, that is a finding. Founders describing different companies often write different group 8s.
After the map
In the same session: fill Owner, Status, and Condition together.
Optional follow up: upload the filled file again for help resolving contested rows and covering gaps.
After three months: review the map based on real work. Review it again after one year, or when a funding round, major hire, pivot, or another company event changes the work.
The merge prompt, v1.0
Paste everything below into your model, then attach both CSV files. Tested on Claude Sonnet 5 and Gemini 3.7 Flash.
Version and reuse
Map v1.0, merge prompt v1.0. Free to use and adapt with attribution. If you run it with your own founding team or inside a program, I would like to hear what broke.