Decision intelligence for transformation — which processes are actually worth automating with AI.
Cursor for AI value discovery: point it at a company's processes, get a ranked, ROI-scored automation backlog.
Every company wants to "use AI." Few know where. DelValue ingests a company's processes, scores each one on feasibility, business value, risk, ROI, and payback, and returns a prioritized backlog — so transformation leads and CTOs spend their budget on the workflows that pay, not the ones that demo well.
The bottleneck in enterprise AI isn't models — it's deciding what to automate. Teams pick projects by hype, burn quarters on low-value pilots, and can't defend the roadmap to finance. There's no rigorous, repeatable way to go from "here are our processes" to "here's the value-ranked plan."
DelValue is that scoring layer.
| Step | Example |
|---|---|
| 01 — Ingest | Drop in process docs (PDF / DOCX / TXT) or describe workflows. |
| 02 — Discover | The Discovery agent extracts structured processes from unstructured input. |
| 03 — Score | Each process is scored on feasibility, value, risk, ROI, and payback. |
| 04 — Decide | Get a ranked backlog + recommendations; track predicted vs. actual over time. |
- ✅ You're a CTO, transformation lead, or AI consultant deciding where to start
- ✅ You have a long list of "AI ideas" and no rigorous way to prioritize them
- ✅ You need an ROI/payback case finance will accept
- ✅ You want the roadmap to learn from what actually shipped
🔎 Value Discovery Agent — Parses PDF / DOCX / TXT via Claude and turns unstructured processes into structured, typed models.
🧮 Multi-Factor Decision Engine — Scores each process on feasibility, business value, risk, ROI, and payback period.
🤖 Four-Agent Pipeline — Analysis, Discovery, Monitoring, and Recommendation agents plus a judgment layer.
📈 Ranked Backlog — A prioritized, defensible list of what to automate first.
🔁 Learning Loop — Stores predictions and compares them to actual outcomes to improve future scoring.
🌐 API + UI — Streamlit app for exploration and an api_server.py for programmatic access.
🧪 Tested — 17 passing pytest cases lock the scoring logic.
┌───────────────────────────────────────────────────────────┐
│ DELVALUE AI │
│ │
│ Discovery Agent ──► structured processes (Pydantic) │
│ │ │
│ ▼ │
│ Decision Engine ──► feasibility · value · risk │
│ │ ROI · payback │
│ ▼ │
│ Recommendation ──► ranked automation backlog │
│ │ │
│ ▼ │
│ Monitoring ◄──────► predicted vs. actual (learning loop)│
│ │
│ Persistence: SQLite UI: Streamlit + Plotly │
└───────────────────────────────────────────────────────────┘
Discovery Agent — document ingestion (PDF/DOCX/TXT) → Claude extraction → validated Pydantic models.
Decision Engine — multi-factor scoring that combines feasibility, value, risk into an opportunity score, with ROI and payback.
Recommendation + Judgment — turns scores into a ranked, explained backlog.
Monitoring — records predictions, ingests outcomes, and feeds the learning loop.
- Language: Python
- AI: Anthropic Claude SDK
- Modeling: Pydantic
- Storage: SQLite
- API: FastAPI-style
api_server.py - UI: Streamlit + Plotly
- Testing: pytest (17 cases)
git clone https://github.com/westfellow25/delvalue-ai.git
cd delvalue-ai
python -m venv .venv && .venv\Scripts\activate # Windows
pip install -r requirements.txt
cp .env.example .env # add ANTHROPIC_API_KEY
# UI
streamlit run app.py
# or API
python api_server.pyRun the tests:
pytest -v- Not a generic AI ideas list. It scores your processes with ROI and risk.
- Not a one-shot report. It tracks predictions against outcomes and learns.
- Not an automation tool. It tells you what's worth automating — you build it (or use PRAXIS to deploy it).
- Four-agent pipeline (Analysis / Discovery / Monitoring / Recommendation)
- Multi-factor decision engine (feasibility / value / risk / ROI / payback)
- Document ingestion (PDF/DOCX/TXT) → structured models
- SQLite persistence + predicted-vs-actual learning loop
- API server
- Connectors (Jira / Linear / process-mining imports)
- Portfolio view across departments
- Hand-off to PRAXIS for deployment
Built by @westfellow25.