AI Product Intelligence Suite 1.04

Portfolio case study for AI product workflows in regulated enterprise software.

Live demo · Release v1.04

This project is a portfolio case study, not a commercial product.

In 60 seconds, this app shows how a regulated product team can convert ambiguous compliance input into reviewed requirements, QA coverage and safer release communication.

The project demonstrates AI-assisted product workflows for regulated enterprise software:

1. domain understanding;
2. AI workflow design;
3. human-in-the-loop review;
4. risk-aware communication;
5. evaluation discipline;
6. enterprise readiness judgment;
7. product strategy.

Positioning: AI-assisted workflow design for turning ambiguous compliance inputs into auditable product decisions.

10-second summary

The suite shows how a compliance-sensitive product team could move from regulatory source material to reviewed requirements, QA coverage, safer release communication and audit-ready evidence.

It is intentionally human-in-the-loop. The goal is not to automate compliance decisions; the goal is to make product decisions more traceable, reviewable and measurable.

What this is

• A working Streamlit portfolio prototype.
• A regulated-enterprise AI product workflow case study.
• A demonstration of product judgment, release risk thinking and evidence design.
• A local prototype with tests, services, docs, usage metrics and exports.

What this is not

• Not a production SaaS platform.
• Not legal, tax, financial or regulatory advice.
• Not a replacement for compliance, legal, QA or product approval.
• Not a claim of enterprise deployment completeness.

Visual walkthrough

Start here / Hero Demo	Hero workflow output	QA, release and audit

These screenshots show the intended first-click path: start with the hero demo, inspect the compliance-to-product output, then review QA, release-risk and audit evidence.

5-minute review path

1. Open the live demo.
2. Keep Demo Mode ON.
3. Review the hero workflow below.
4. Read HERO_CASE_STUDY.md.
5. Read WHAT_THIS_DEMONSTRATES.md.
6. Open docs/SYSTEM_OVERVIEW.md for architecture and real-vs-simulated scope.

System flow

flowchart TD

A[Regulatory source material] --> B[INTERPRET<br>Obligations<br>Requirements<br>Traceability]

B --> C[DISCOVER<br>Gaps<br>Risks<br>Inconsistencies]

C --> D[COMMUNICATE<br>Safer release messaging<br>Grounding<br>Risk scoring]

D --> E[INVESTIGATE<br>Decision timelines<br>Evidence<br>Citations]

E --> F[Governance and Auditability<br>Logs<br>Staleness<br>Human review flags]

Loading

Hero workflow

The core case is SAF-T PT / e-invoicing compliance-to-product traceability:

source document
→ extracted obligations
→ reviewer corrections
→ before/after requirement
→ Jira-style ticket
→ QA case
→ negative test coverage
→ risky release note
→ safer release note
→ incident if missed
→ final audit report

This is intentionally deeper than a typical AI demo output. It shows how one compliance-sensitive input can propagate through product, QA, release communication and incident learning.

Key product capabilities

Interpret — Compliance-to-Product Studio

Turns source material into obligations, source-linked evidence, reviewer decisions, requirement candidates, QA coverage and audit-aware exports.

Discover — Product Discovery Studio

Converts a product idea into structured product artefacts such as assumptions, trade-offs, Jira-style tickets, Gherkin acceptance criteria, QA matrix and PRD completeness checks.

Communicate — Release Readiness Copilot

Reviews release communication for risky claims and suggests safer wording with caveats, scope and approval boundaries.

Investigate — Decision Timeline Builder

Builds incident and decision timelines with owners, severity, contradictions, risk register, postmortem actions and customer-escalation context.

Quality and risk controls

Implemented as local portfolio controls:

• claim hygiene scanner;
• citation-support heuristics;
• mandatory negative test coverage;
• reviewer mode;
• approval workflow simulation;
• document hashes and versioning;
• run history and usage metrics;
• connector outbox payloads;
• real vs simulated capability table.

These are product judgment demonstrations, not claims of production SaaS readiness.

Useful review documents

• PORTFOLIO_REVIEW_GUIDE.md — 5-minute, 15-minute and 45-minute review paths.
• WHAT_THIS_DEMONSTRATES.md — skills-to-evidence map.
• PRODUCT_STRATEGY.md — ICP, personas, wedge, roadmap and metrics.
• VALIDATION_LIMITATIONS.md — what is validated, synthetic, local or not production-ready.
• docs/SYSTEM_OVERVIEW.md — architecture, workflow and real-vs-simulated explanation.
• docs/UI_UX_REVIEW.md — public UI/UX review criteria.
• docs/CLAIM_HYGIENE_SCANNER.md — release-claim risk and safer wording control.

Companion domains

The main hero case is SAF-T PT / e-invoicing.

Additional companion domains show generalization thinking:

• Swiss QR-Bill / invoice payment compliance;
• SEPA / ISO 20022 structured-addresses companion playbook for a separate MIT agentic AI course project.

The payments playbook is intentionally kept outside the main Streamlit workflow to preserve scope discipline.

Runtime and reproducibility

The project is tested with Python 3.12.

Dependency versions are pinned with compatible-release constraints in requirements.txt to keep the public demo reproducible while still allowing safe patch-level updates.

Run locally

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt
python -m pytest -q
python -m streamlit run app.py

Start with Demo Mode ON to avoid API usage.

Responsible AI note

This project does not provide legal, compliance, tax, financial or regulatory advice. Human review is required before using any output in operational decisions.

Release note

This repository was published as a sanitized consolidated public portfolio release. Earlier internal iterations were developed offline and are kept outside the public repository.