Generative AI

From SaaS to intelligence native: the feedback loop.

Intelligence-native systems need agent access to decision artefacts and feedback loops. Why context, not models, is the differentiator — and how MCP, traditional ML, and versioned artefacts fit together.

Cyril Noirot

May 4, 2026

Share on

10 min read

Enterprise software was built to digitise work. Forms, records, dashboards, workflows.

That layer is largely solved. Every function in a modern company has a SaaS tool, and most of them have several.

The next layer is different. It is not about recording what happened. It is about helping the organisation decide what to do next. That is the layer "intelligence-native" actually refers to — and the part most demos still get wrong.

The demo question is "can the agent answer questions over our data?" The production question is "can the agent reach the artefacts a human would reach to decide, and does it learn from whether its recommendation was followed?"

About the author

Cyril Noirot

Lead Data Scientist

Freelance data scientist. I design and ship decision systems — forecasting, pricing, marketing measurement, optimization.

Who controls what your AI recommends?

Seasons, campaigns, and weekly events — retail runs on overlapping cycles, and the AI recommender has to keep up with all of them. Notes on the business-rules control surface that lets merchandising teams steer a conversational recommender without editing prompts, filing tickets, or waiting for a deploy.

Generative AI11 min

Ce que le commerce agentique exige vraiment en production

Macy's annonce que les clients utilisant son assistant IA dépensent 4,75x plus. Sephora vient de lancer une app dans ChatGPT. Zalando déploie son assistant dans 25 marchés. La question pour tous les autres retailers n'est plus 'faut-il le faire ?' mais 'comment l'architecturer pour que ça tienne en production ?'

Generative AI11 min

Building a recommendation engine that doesn't trust the LLM

This is the engineering companion to the production architecture piece. Instead of re-arguing why open-ended agents are risky in commerce, it walks through the actual implementation choices in `ai-florist`: FastAPI boundaries, LangGraph orchestration, pgvector retrieval, learned scoring weights, deterministic fallbacks, and runtime observability.

Newsletter

Technical writing on forecasting, pricing, and decision systems. No fixed schedule, no spam.

Enter your email