A human-governed, AI-assisted operating system for decomposing vague AI ideas into intelligence architecture, proof requirements, value logic, governance controls, and build/buy/wait routes before teams spend time on demos, vendors, proofs of concept, or delivery planning.
Public portfolio prototype. Designed for ChatGPT Project use, executive review, and workflow demonstration. Not a SaaS product, autonomous approval engine, or substitute for accountable AI governance.
| Review question | Where to look |
|---|---|
| Status | Public portfolio prototype for AI opportunity review, ChatGPT Project use, and executive operating-governance demonstration. |
| Best evaluator | AI operations, PMO, portfolio, product, strategy, and transformation leaders deciding whether AI ideas deserve prototype, build, buy, wait, process-first, or stop routes. |
| Operating decision supported | What business problem, intelligence architecture, proof burden, value logic, workflow readiness, and governance controls must be visible before an AI idea receives delivery attention? |
| Concrete example | examples/sample-outputs/ shows synthetic AI opportunity outputs and route recommendations. |
| Before / after proof | Before: AI ideas arrive as demos, vendor claims, prompts, scripts, or broad automation requests. After: each idea has purpose, sensing, interpretation, decision, orchestration, learning, govern/assure controls, proof plan, and a human review route. |
| Boundary | This system reviews opportunity quality and routing. It does not approve AI projects, select vendors, accept risk, estimate ROI from invented numbers, or replace accountable governance owners. |
| Portfolio lane | Review AI opportunities. |
Start with the README to understand the operating model, then inspect:
chatgpt-project/for the flat ChatGPT runtime.examples/sample-data/for synthetic AI opportunity inputs.examples/sample-outputs/for generated architecture, routing, and governance outputs.quality-review/for cohesion and senior-operator critique.
Evaluate the repo on whether it separates AI enthusiasm from workflow value, proof burden, data readiness, controls, and human decision rights.
For local maintenance, compare value claims against
../roi-business-value-anti-patterns.md
so demo-as-proof, token-spend confusion, unsupported savings, and AI washing do
not creep into public examples.
Before: a team has a list of AI ideas, vendor claims, or prototype requests, but the work is bundled, the value evidence is uneven, and leaders do not know whether to build, buy, wait, process-map, or stop.
After: each idea is decomposed into purpose, sensing, interpretation, decision, orchestration, learning, and govern/assure controls, with a route recommendation, proof plan, missing evidence, and human portfolio review path.
AI makes it easier to propose projects than to understand them.
This project helps PMO, EPMO, product, operations, strategy, and AI transformation leaders turn rough AI concepts into clear operating designs. It asks what kind of intelligence the idea actually needs, what workflow it changes, what signals it observes, what judgments it makes, what it orchestrates, how it learns, and what controls keep humans accountable.
The goal is not to create another approval layer. The goal is to create an intelligence review layer.
Customer-language research shows that buyers recognize AI use-case intake, AI prioritization, ROI and feasibility review, duplicate-use-case filtering, proof planning, and human validation faster than they recognize the phrase "opportunity intelligence."
Use this module to answer the practical leadership question:
Which AI requests deserve delivery attention, and what must be proven before we spend scarce time on them?
The strongest external framing is:
- tame the avalanche of internal AI requests;
- separate workflow value from demo excitement;
- identify process-first, wait, build, buy, prototype, and stop routes;
- make ownership, proof burden, eval criteria, and controls visible before a pilot becomes a portfolio commitment.
The product is mostly structurally sound. The next capability watch item is a stronger portfolio view for comparing AI requests by value evidence, feasibility, risk, duplicate overlap, human review need, and readiness to test.
Organizations are about to get more AI ideas, cheaper prototypes, faster demos, and louder vendor claims. That creates a portfolio risk: wasting leadership attention, scarce change capacity, and engineering time on ideas that sound AI-native but do not improve a meaningful workflow.
The better question is not:
Can we build a proof of concept?
The better question is:
What intelligence system is this idea actually asking for, and what is the cheapest trustworthy way to prove or disprove it?
- PMO, EPMO, and portfolio leaders
- AI transformation and operations leaders
- Product, strategy, and business operations teams
- Chiefs of staff and executive operators
- Governance teams reviewing AI demand
- Leaders deciding whether to automate, build, buy, hire, wait, prototype, process-map, or stop
- Converts rough AI ideas into structured concept records.
- Decomposes each idea into an intelligence stack: purpose, sensing, interpretation, decision, orchestration, learning, and governance.
- Separates business problems from AI enthusiasm.
- Identifies whether the idea is one workflow, several smaller workflows, a data problem, a process problem, a vendor fit, or a capability gap.
- Reviews value clarity, workflow readiness, data readiness, proof burden, governance controls, and strategic fit.
- Routes opportunities to prototype, build, buy, automate with existing tools, hire/upskill, wait, process-first, decompose further, or stop.
- Produces an AI opportunity architecture brief, experiment plan, governance/control review, and portfolio triage summary.
- It does not approve AI projects.
- It does not replace executive judgment, finance review, legal review, security review, architecture review, procurement review, or product ownership.
- It does not estimate ROI from invented numbers.
- It does not connect to live systems or send data to vendors.
- It does not claim that a proof of concept is always cheaper than deciding not to build.
- It does not treat every AI idea as a build opportunity.
This package is based on lessons synthesized from five public YouTube videos supplied as research inputs. The public project contains derived lessons and workflow design, not copied transcript text.
Key synthesized lessons:
- AI investment is a work-shape question before it is a model or vendor question.
- Repeatable AI work needs scaffolding: instructions, tools, scripts, connectors, skills, logs, evals, and human review.
- Automation often creates higher-order human work: monitoring, exception handling, problem solving, design, and assurance.
- Agentic systems need an intelligence loop: purpose, sensing, interpretation, decision, orchestration, learning.
- That loop needs a govern-and-assure wrapper: evals, searchable logs, rollback, permissions, and human review queues.
- Claims need proof. AI-native positioning without evidence is fragile.
See research/source_video_lessons.md for the source lesson synthesis.
The operating framework is AI Opportunity Intelligence Review.
| Stage | Purpose | Output |
|---|---|---|
| Concept Intake | Capture the AI idea, sponsor, workflow, user, pain, expected outcome, and proposed AI role. | Structured concept record |
| Problem Test | Separate the business problem from tool enthusiasm. | Problem clarity rating |
| Intelligence Decomposition | Break the idea into purpose, sensing, interpretation, decision, orchestration, learning, and govern/assure controls. | Intelligence stack map |
| Work Shape Review | Classify the work as bounded task, synthesis, exception handling, decision prep, interaction, data extraction, coordination, or unclear work. | Workflow-shape rating |
| Value And Proof Review | Define measurable value, baseline, evidence, proof burden, and stop conditions. | Proof plan |
| Build/Buy/Hire/Wait Routing | Recommend the next route based on architecture, value, risk, and maturity. | Route recommendation |
| Governance And Assurance | Define human review, evals, logs, rollback, approvals, and risk controls. | Control model |
| Decision Brief | Summarize route, rationale, missing evidence, next proof, owner, and human decision needed. | AI opportunity brief |
This system sits upstream of the current product set:
- AI Opportunity Intelligence Review System - decomposes and evaluates AI ideas before investment.
- Business Case System - develops decision-ready business cases for opportunities that require investment justification.
- Project Charter Initiation Agent - converts approved intent into sponsor-ready project charters.
- Portfolio Prioritization Scoring Agent - compares approved work through transparent portfolio scoring.
- PMO Governance Operations Log - helps operate the recurring governance rhythm once work is underway.
This repository has two shapes:
- GitHub repository layer: this full folder, used for discovery, examples, sample data, templates, workflow diagrams, local tooling, and quality review.
- ChatGPT Project runtime layer: the flat
chatgpt-project/folder, used as the actual operating system inside ChatGPT.
Upload only the files inside chatgpt-project/ when creating a ChatGPT Project. Do not upload the full repository. The other folders provide examples, templates, generated outputs, workflow diagrams, local tooling, and public portfolio context.
Start with:
Use this project to decompose the attached AI opportunity ideas into intelligence stacks, value/proof reviews, route recommendations, and governance controls.
Use the full repository when you want local sample runs, CSV validation, generated Markdown/HTML outputs, or project modifications.
python tools/decompose_ai_opportunities.py \
--input examples/sample-data/synthetic_ai_opportunities.csv \
--output-dir examples/sample-outputsai-opportunity-intelligence-review-system/
README.md
AGENTS.md
LICENSE.md
.gitignore
chatgpt-project/
examples/
sample-data/
sample-prompts/
sample-outputs/
templates/
tools/
workflow/
quality-review/
research/
flowchart TD
A[Raw AI Idea] --> B[Concept Intake]
B --> C[Problem Clarity Test]
C --> D[Intelligence Stack Decomposition]
D --> D1[Purpose]
D --> D2[Sensing]
D --> D3[Interpretation]
D --> D4[Decision]
D --> D5[Orchestration]
D --> D6[Learning]
D --> D7[Govern And Assure]
D1 --> E[Work Shape Review]
D2 --> E
D3 --> E
D4 --> E
D5 --> E
D6 --> E
D7 --> E
E --> F[Value And Proof Review]
F --> G[Govern And Assure Review]
G --> H{Route}
H --> I[Prototype]
H --> J[Build]
H --> K[Buy]
H --> L[Automate With Existing Tools]
H --> M[Hire Or Upskill]
H --> N[Wait]
H --> O[Process First]
H --> P[Decompose Further]
H --> Q[Stop]
I --> R[AI Opportunity Architecture Brief]
J --> R
K --> R
L --> R
M --> R
N --> R
O --> R
P --> R
Q --> R
R --> S[Human Portfolio Review]
S --> T[Decision Log And Carry Forward]
flowchart LR
P[Purpose] --> S[Sensing]
S --> I[Interpretation]
I --> D[Decision]
D --> O[Orchestration]
O --> L[Learning]
L --> S
G[Govern And Assure] -. evals .-> P
G -. logs .-> S
G -. permissions .-> I
G -. human review .-> D
G -. rollback .-> O
G -. outcome review .-> L
flowchart TD
A[Decomposed Opportunity] --> B{Workflow clear?}
B -- No --> C[Decompose Further]
B -- Broken process --> D[Process First]
B -- Yes --> E{Value evidenced?}
E -- No --> F[Stop Or Manual Proof]
E -- Directional --> G{Data and controls ready?}
E -- Strong --> G
G -- Blocked --> H[Wait]
G -- High risk --> I[Control Review Before Proof]
G -- Ready --> J{Differentiated workflow?}
J -- Common --> K[Buy Or Existing Tools]
J -- Strategic --> L[Prototype Or Build]
J -- Capability gap --> M[Hire Or Upskill]
The sample tool generates:
- AI opportunity portfolio summary
- Intelligence stack decomposition table
- Route recommendation table
- Opportunity architecture briefs
- Experiment plan starters
- Governance and assurance control review
- Findings logs in CSV and JSONL
- Standalone HTML review pack
The system may recommend that an idea appears ready for a prototype, should wait, needs process cleanup first, requires further decomposition, or lacks credible value evidence. It must not approve funding, commit resources, accept risk, select vendors, authorize hiring, or change portfolio sequencing. Those remain human leadership decisions.
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.
AI portfolio management, AI governance, AI opportunity triage, AI business value, AI use case evaluation, AI transformation, intelligence stack, AI operating model, AI decomposition, PMO, EPMO, portfolio governance, build buy wait, AI proof of concept, human-in-the-loop, AI decision support, ChatGPT Project, Codex.