Personal project
Agentic Product Stack Mapper
A worksheet for the decisions around an AI agent: where it belongs in a workflow, how much authority it should have, and what would make it worth shipping.
- Format
- Browser-based worksheet
- Output
- A draft product brief
- Access
- No account or email required
Why I built it
Make the open decisions visible.
The mapper is a companion to my essay, The Agentic AI Product Gap. I wanted a way to work through its questions in the context of a specific product, with an artifact a team could discuss.
Take a support-refund agent. Reading an order, drafting a reply, and issuing a refund carry different consequences. The mapper gives each action its own boundary and brings those decisions together with evaluation, recovery, and value.
Using it
From an idea to a first brief.
Describe the work
Start with the user, the task, and the outcome you want to change. A support-refund example is included if you want to explore first.
Set the boundaries
Decide what the system may read or change, which actions require approval, and when it should escalate.
Work through the evidence
Specify evaluation gates, human review, monitoring, recovery, and a measure of business value.
Review and export
Read the assembled brief, revise the assumptions, and export it as Markdown or JSON, or print it to PDF.
What to expect
A starting point for review.
The brief is assembled from your answers and built-in guidance. The mapper does not run an AI agent or validate a product in the real world. Suggested answers need checking against the users, systems, and constraints of your own project.
You can work through the sections in any order and return to the assumptions that need more evidence. The useful result is a clearer discussion about what to build and what still needs to be learned.