# Runpoint AI Project Underwriting Skill

Use this skill before approving an AI project. The goal is to produce a short underwriting artifact that a CEO, CFO, operator, or PE partner can use to decide whether the project is ready to build, needs discovery, or should wait.

## How to run the skill

Paste the prompt below into your AI workspace. Replace the bracketed context with your company and project details.

```text
You are helping us underwrite an AI project before we build it.

First ask which output artifact we want:
1. Executive memo
2. CFO ROI worksheet
3. Data access map
4. Implementation brief
5. Board-style decision page

Then walk us through these sections one at a time. Ask follow-up questions when the answer is vague.

Project context:
[Describe the workflow, team, systems, current pain, and why this matters now.]

Sections to complete:

1. Business constraint
- What workflow are we trying to change?
- Who owns it?
- What slows down, breaks, or gets expensive today?
- What happens if nothing changes for 6 months?

2. Current-state baseline
- How often does the work happen?
- How many people touch it?
- How long does it take?
- Where does work wait?
- What error, rework, or delay shows up?
- What evidence exists: time studies, queues, reports, logs, payroll data, screenshots, or exports?

3. Economic case
- Which KPI should move?
- Who tracks it?
- What is the current baseline?
- What would a meaningful improvement be worth?
- Which assumptions carry the ROI case?
- Which assumptions would kill it?

4. Data map
- Which systems create the work?
- Which systems store the data?
- Which system is the source of truth?
- What data goes in?
- What output comes out?
- Are there PDFs, emails, spreadsheets, portals, screenshots, forms, or scanned documents?
- Are there examples of past inputs and correct outputs?

5. Access reality
- Do the systems have APIs?
- Who has admin access?
- Can data be exported?
- Are credentials available?
- Is vendor approval needed?
- Is security or compliance review required?
- Can access be granted before engagement?

6. Expected output
- What should the system produce?
- Who consumes it?
- Where should it appear?
- What format should it take?
- How accurate does it need to be?
- What should happen when confidence is low?
- What needs human review?

7. Operating owner
- Who owns the workflow?
- Who can approve process changes?
- Who can unblock access?
- Who will test early outputs?
- Who is accountable for adoption?

8. Shovel readiness
Rate each 1 to 5:
- ROI clarity
- Data availability
- Access confidence
- Workflow stability
- Owner engagement
- Implementation urgency
- Testability
- Adoption likelihood

Classify the project as one of:
- Ready to build
- Discovery first
- ROI case incomplete
- Bad first project

Final output:
Create the selected artifact. Keep it concise. Include:
- Decision recommendation
- ROI hypothesis
- Required data and access
- First milestone
- Main risks
- Questions to answer before build
```

## Rubric

Ready to build: economics are clear, access is likely, and the workflow owner is engaged.

Discovery first: the opportunity is real, but the data or access picture needs inspection.

ROI case incomplete: the pain is real, but the business case needs sharper math.

Bad first project: the workflow is blocked, political, hard to measure, or dependent on a future state the company does not control.