Path to Production.
The heart of "Decide"

Match the size to the need.

The same idea can take four different forms. Most of the value in this whole framework comes from picking the right one up front, because that single choice decides how much building, securing, and supporting it will ever need. The mistake that buries a lean team is turning something into a full app when it should have stayed a prompt.

Size 1 · ●○○○

Prompt

Use it when it's a one-off, or something only you do.
Delivery: just words, reused in the AI tool Cost to run: none Owner: the person using it
Examples
  • Summarizing a single earnings call or a long PDF you're reading today
  • Drafting the first pass of a memo, then editing it yourself
  • Reformatting a messy spreadsheet or cleaning up your own meeting notes
  • A one-time "pull the key terms out of this contract" ask
Size 2 · ●●○○

Shared tool

Use it when a team keeps doing the same thing the same way.
Delivery: a saved, versioned, shared recipe Cost to run: low, no hosting Owner: the team that uses it
Examples
  • A standard "summarize a diligence document our way" recipe the whole team reuses, so every summary comes out consistent
  • A saved competitor-scan the research team runs every week
  • A house-style helper that turns rough notes into the firm's memo format
  • A shared checklist prompt that reviews a draft before it goes out
Size 3 · ●●●○

Lightweight app

Use it when it needs a screen or a memory, but the data isn't sensitive.
Delivery: a small hosted app, basic sign-on Cost to run: medium, watched Owner: named owner + light support
Examples
  • An internal lookup over public filings or press, with a simple search box
  • A small dashboard the ops team checks each morning
  • A form-driven helper that several people use but that only touches public or low-risk information
  • An intake form that routes requests and remembers what came in
Size 4 · ●●●●

Production app

Use it when it touches sensitive data, or a lot of people rely on it.
Delivery: full path, all the controls Cost to run: full, budgeted & supported Owner: owner, supporter, sign-off, failure plan
Examples
  • A question-and-answer tool over confidential financials, where people must only see what they're entitled to
  • Anything that touches investor, client, or deal information
  • A tool the investment or finance teams depend on to do their actual work
  • Anything external-facing, or anything a regulator or auditor might ask about
The tell: if you can't clearly say who would support it and what happens if it's wrong, it is not a Size 4 yet. Keep it smaller until the need is real. It is always cheaper to grow an idea up a size than to unwind a full app nobody needed.
Ideas graduate

Start small. Grow only when it earns it.

Most things should enter as a prompt and stay there. The few that prove themselves move up a size at a time, and each step up is a deliberate decision, not an accident.

Prompt
Someone finds it useful.
→
Shared tool
The team wants it consistent.
→
Lightweight app
It needs a screen or memory.
→
Production app
It touches real data or real users.
← Back
The three moves