A human-governed, AI-assisted portfolio decision-support system for evaluating approved projects and programs through transparent weighted scoring, portfolio metadata, strategic themes, constraints, risks, dependencies, ownership, and executive review artifacts.
This is not a funding engine, autonomous prioritization system, or replacement for a portfolio board. It helps humans structure the conversation, inspect tradeoffs, and make better decisions.
Public portfolio prototype. Designed for ChatGPT Project use, executive review, and workflow demonstration. Not a SaaS product, optimization engine, autonomous funding tool, or replacement for portfolio governance.
Open these first:
chatgpt-project/for the flat ChatGPT runtime.examples/sample-data/for synthetic portfolio inputs.examples/sample-outputs/for scored matrices, executive briefs, logs, HTML, and DOCX outputs.quality-review/for senior portfolio-manager critique.
Evaluate the repo on whether it makes scoring transparent, separates mandatory and discretionary work, surfaces weak metadata, preserves human decision rights, and turns prioritization into an auditable conversation.
For local maintenance, compare value scoring against
../roi-business-value-anti-patterns.md
so high-value narratives do not outrun evidence confidence or measurement
readiness.
Before: leaders have a portfolio of approved or proposed initiatives, but priorities are defended through narrative, urgency, politics, or incomplete data rather than visible criteria and tradeoffs.
After: initiatives are normalized, scored through explicit criteria and weights, separated by mandatory/discretionary treatment, reviewed for missing owners and risks, and prepared for a human portfolio forum.
Customer-language research confirms the core pain: "everything is urgent," the loudest request wins, pet projects get protected, and executive overrides become invisible. The module should be framed as a way to make tradeoffs explicit, not as a ranking machine.
Lead with:
- transparent scoring criteria and weights;
- mandatory versus discretionary separation;
- missing-owner and weak-evidence flags;
- sensitivity views for weight changes;
- visible rationale when leaders override the model;
- handoff to capacity sequencing when a high score does not mean the work can actually start.
The module is structurally strong. The improvement target is clearer public copy around human-governed scoring, override rationale, and capacity-aware follow-up.
Organizations often have many approved or proposed initiatives but no consistent way to compare their strategic value, financial contribution, risk, effort, mandatory status, dependencies, and operational impact. The result is noisy prioritization, unclear ownership, overloaded teams, and decisions that are hard to defend later.
This project helps portfolio leaders convert initiative metadata into a visible scoring model and decision-support package.
- PMO, EPMO, and portfolio leaders
- Program and project leaders
- Business owners and sponsors
- Finance, technology, operations, and product stakeholders
- Knowledge workers who need a structured prioritization model without pretending the AI owns the decision
- Defines portfolio categories, criteria, weights, gates, and review cadence
- Normalizes initiative metadata from business cases, charters, spreadsheets, or notes
- Applies transparent weighted scoring
- Separates mandatory and discretionary work
- Surfaces missing owners, unclear decision rights, weak assumptions, risks, and dependencies
- Produces portfolio summaries, scoring matrices, decision briefs, logs, and quality reviews
- It does not make final funding or sequencing decisions
- It does not replace sponsor, finance, technology, or executive governance review
- It does not create full business cases or project charters
- It does not perform opaque optimization or machine-learning ranking
- It does not require real company data; all examples are synthetic
| Boundary question | Answer |
|---|---|
| This module starts when | Multiple initiatives have enough metadata, sponsorship, value logic, and governance context to compare through a transparent scoring model. |
| This module ends when | A scored portfolio view, tradeoff summary, decision-support brief, quality review, and downstream handoff note are ready for human review. |
| This module produces | Scoring model, normalized initiative metadata, weighted scores, confidence flags, mandatory/discretionary separation, decision brief, decision log seed, and handoff note. |
| This module hands off to | Portfolio Capacity Sequencing Planner, PMO Governance Operations Log, Executive Portfolio Review Pack Builder, Value Realization Governance Ledger, Portfolio Intake Readiness Triage System, Portfolio Signal Quality Auditor, or Controls Exposure Governance Toolkit. |
| This module does not | Approve funding, cancel initiatives, sequence work as an official decision, create full business cases or charters, run recurring governance, or accept risk. |
| Need | Better destination |
|---|---|
| Raw demand is not ready to score | Portfolio Intake Readiness Triage System |
| Portfolio data is stale, duplicated, missing owners, or KPI-weak | Portfolio Signal Quality Auditor |
| Ranking creates capacity, dependency, or fixed-window tradeoffs | Portfolio Capacity Sequencing Planner |
| Decisions, actions, risks, or follow-ups need tracking | PMO Governance Operations Log |
| Sponsor or executive review needs a broader pack | Executive Portfolio Review Pack Builder |
| Benefit claims need measurement after approval | Value Realization Governance Ledger |
Upload only the files inside chatgpt-project/ when creating a ChatGPT Project. Do not upload the full repository. The other folders provide examples, sample data, templates, generated outputs, workflow diagrams, quality review, and local tooling for GitHub or Codex use.
The runtime folder is flat, self-contained, and designed to stay under the ChatGPT Project file-count limit.
Recommended first prompt after upload:
"Use the runtime instructions and trigger map to help me set up a portfolio prioritization scoring model. Interview me for strategy, initiative categories, scoring criteria, weights, governance cadence, budget/capacity constraints, and decision-rights assumptions. Keep final decisions human-owned."
flowchart TD
A[Approved Initiative Metadata
Business case and charter summaries, costs, benefits, owners, risks] --> B[Runtime Intake
Normalize fields, classify initiative type, identify missing data]
B --> C{Governance Basics Present?}
C -->|No| D[Clarification Queue
Sponsor, owner, decision rights, assumptions, constraints]
D --> B
C -->|Yes| E[Scoring Model Setup
Criteria, weights, gates, mandatory vs discretionary treatment]
E --> F[Weighted Scoring
Strategic alignment, value, risk, impact, effort, dependencies]
F --> G[Portfolio Review
Rankings, tradeoffs, constraints, risk/dependency view, KPI summary]
G --> H{Decision Support Needed?}
H -->|Yes| I[Executive Decision Brief
Options, tradeoffs, risks, assumptions, open questions]
H -->|No| J[Portfolio Summary View
Scored matrix, watch items, recommended follow-up]
I --> K[Human Portfolio Forum
Leaders decide funding, sequencing, acceptance, or deferral]
J --> K
K --> L[Decision Log and Follow-up
Record rationale, owners, actions, next review]
L --> M[Handoff note
Sequencing, governance log, executive review, value ledger, or cleanup]
chatgpt-project/- flat runtime folder for ChatGPT uploadexamples/- synthetic data, sample prompts, sample outputs, and source artifactstemplates/- reusable templates for review outside the runtime foldertools/- local Python scoring utilityworkflow/- Mermaid workflow diagramquality-review/- senior portfolio manager self-critique
The system may recommend scoring approaches, identify gaps, calculate weighted scores, summarize tradeoffs, and prepare decision briefs. It must not approve, cancel, fund, sequence, or accept risk on behalf of leaders.
Source code and scripts are licensed under MIT. Documentation, prompts, templates, examples, and other non-code materials are licensed under CC BY 4.0 with attribution to Marco Policani. See LICENSE.md.