Skip to content

About

Guy HUI-BON-HOA

AI Product Owner & Agile Project Manager
— at the intersection of business, technology and delivery.

I am Guy HUI-BON-HOA, an AI Product Owner and Agile Project Manager with 20 years of experience in IT consulting — spanning product ownership, Agile project management and business-side (MOA) project management. Most of that work was delivered through French IT consultancies (ESNs); more recently I work independently through an umbrella company.

I work where business, engineering and UX meet. As a Product Owner, I turn business needs into clear requirements, requirements into a prioritised backlog, and that backlog into a digital product that ships. Prioritisation is where the client's value is decided: what earns its place in a release, and what does not.

Today I apply that same product and delivery discipline to agentic AI — designing agents, declarative workflows, and the quality gates that decide whether AI-generated work is fit to keep. I also work as a Business Analyst (MOA) and Web Integrator, which is why the specification and the front-end integration often arrive in the same pair of hands.

In one sentence: I build AI systems that stop and ask the business the right questions instead of guessing — and every run is recorded, dated and open to inspection.

What I do

I design and build AI systems that can be used in real business processes — not just demos. The work combines product management, business analysis and agentic AI: defining the business need, structuring the process, orchestrating specialised AI capabilities, and putting controls around the output. The projects on this site are the working examples — a versioned catalog of agents and skills, a headless runtime that executes them under blocking gates, and RAWLY, a nutrition product where every scientific claim is re-resolved against its source before it can be published.

The method matters more than the stack. An agent that produces a plausible answer is worth nothing if nothing checks it, so the interesting engineering sits in the gates, the traces, and the discipline of never promoting an offline pass to a live proof.

Where I've worked

20 years of IT consulting across CAC 40 groups, large enterprises and SMEs, in French and in English, frequently in international and offshore contexts.

Client environments: Orange · Chanel · Accor · Renault · Air France · SNCF · EDF · MBDA · MGEN · Travelex · Aginode

Consulting / ESN environments: Sogeti · Astek · Gfi (Inetum) · Viveris · Kapito Consulting · Ethic Technology

Recent work includes a carve-out programme (PIM / CMS / MDM / SAP migration onto an open-source Drupal solution), functional evolutions on a B2B portal for an investment bank, and a brand-wide digital asset migration.

The detailed history is on my LinkedIn profile, which LinkedIn shows in full only to signed-in visitors.

Focus areas

  • Agentic AI — multi-agent orchestration, declarative workflows, eval-gates.
  • Generative AI / LLM — prompt engineering, RAG, AI product ownership.
  • Product ownership (Scrum / SAFe) — backlog, user stories, DoR / DoD, Agile delivery at scale.
  • Business analysis (MOA) — functional specifications (SFG / SFD), requirements, V-model deliverables, business-side UAT.
  • Web integration — semantic HTML / CSS, responsive front-end integration, turning designs into production-ready pages.
  • Quality & licence governance — ISO-aligned audits, open-source compliance.

How I work

  • Start with the business outcome, then define what must be delivered.
  • Lead with a clear recommendation, then validate step by step.
  • Prove it live — never present an offline pass as production evidence.
  • Quality is a delivery issue, not one to postpone — reduce scope, never rigour; anomalies fixed as part of delivery, not deferred.
  • Prefer controlled scope over uncontrolled assumptions.

Deciding that a system should stop and ask rather than produce something plausible is a product decision, not a technical one. It costs more to build, it makes the demo less impressive, and it is the difference between a pilot that survives contact with a client and one that does not. Twenty years of running delivery is what makes that trade-off obvious rather than clever. It is visible in the dispatcher runs.

Certifications

Professional certifications

  • Professional Scrum Product Owner I (PSPO-1) — Scrum.org, 2023.
  • SAFe 6 Product Owner / Product Manager — Scaled Agile, 2023 (lapsed 2024).

Badges

  • Claude Code 101, Claude Code in Action, Claude 101 — Anthropic, 2026.
  • Data Landscape of GenAI for Project Managers — Project Management Institute, 2026.
  • Generative AI Overview for Project Managers — Project Management Institute, 2025.

Not to be confused with the agent personas

These are the author's own certifications. The certifications named inside the catalog's agent identity blocks (PMP, CBAP, ISTQB, TOGAF, …) describe a simulated persona's knowledge frame and are not held by the author.

Education

  • EPITA — Master's degree in Management and Information Technology, 2004.
  • ISEG — Management and International Business, 1999.

Elsewhere

  • My LinkedIn profile — professional background, detailed work history and recommendations.

  • My GitHub profile — public repositories:

    • claude-agents — the agent catalog.
    • claude-agentic-runtime — the headless runtime.
    • claude-projects — companion projects.

Both profiles open in a new tab.

Contact

Have a delivery process you want to make executable, auditable, and repeatable?

I take on a limited number of engagements. Tell me what you are working on — the request, the constraints, the governance you answer to — and I will tell you plainly whether this method fits, and how it would apply to your work.

Email Guy →