Steny Solitude

Reasoning AI, Cognitive Engine & Network of Meanings

Steny Solitude. I build cognitive systems that turn data, content and knowledge into decisions people can act on, in the field.

Contact
FocusCognitive EnginesNetwork of MeaningsSpatial Computing
BuildingAllegorya

Abstract

I work on a new class of systems that (1) turn raw data into meaning that others can use, (2) let business reasoning steer AI rather than the other way around, (3) certify the authority, origin and recipient of every resource, and (4) bring real-time decision support to the people who act, where they act.

For decades, computing has served those who manage from a distance. Dashboards, data lakes and management systems describe the territory. They rarely help the people who have to cross it.

Large language models showed what AI can do. In a business they also showed its limit: they simulate expertise without embodying it, and they tolerate errors of context, data and interpretation that no decision-maker can sign off on.

Organisations need a Network of Meanings: knowledge that is structured, traceable and intelligible to anyone who needs it, without prior knowledge of where it came from. They need Cognitive Engines that combine AI with reasoning to deliver that knowledge in context, to the right operator, for the right task.

This is the work I do, from semantic technology for media and enterprise knowledge to real-time assistance in the operating room.

Decisions are not made on the map. They are made in the field.

Thesis

Cognitive intermediation is the next platform shift. Chatbots are only its first building blocks.

/0.1

Network of machines

The Internet links systems through IP.

→ a library
/0.2

Network of people

Web 2.0 adjusts the interface in real time.

→ an audience
/0.3

Network of Meanings

Cognitive Engines combine AI and reasoning.

→ expertise
Outlook · the future of AI

The future of AI is plural, and it runs on meaning.

A synthesis of my analysis “Why AI & LLMs are driving the rise of reasoning systems”. Large language models opened the market. They will not carry enterprise AI alone.

Formal · symbolic

Explain, infer, guarantee

  • OWL ontologiesShared, self-describing models of a business domain
  • Knowledge graphs · RDFData linked to its context, source and authority
  • Reasoners & rule enginesNew facts inferred from existing ones, step by step
  • Linked Open DataExternal knowledge connected without copying it

Fact → stable knowledge, described with a grammar

Orchestration

Cognitive Engine

Business reasoning in command. It calls each kind of intelligence when the task needs it, and keeps every result traceable.

Heuristic · statistical

Perceive, predict, generate

  • Generative LLMsA natural interface and a broad but dated general culture
  • Specific language modelsA profession’s discourse, trained from the knowledge graph
  • Machine & deep learningPatterns and predictions from large volumes of signals
  • Data mining & time seriesStatistical reporting that feeds the modelled knowledge

Observation → hypothesis → discovery

Two ways to build the foundation

Classic approach · data lakesSemantic approach · ontologies & knowledge graphs

Principles

Six convictions that run through my work and my writing. Each one links to the essay where I argue it.

Writing

Essays

On cognitive intermediation, AI governance, data strategy and spatial computing. Read in full here, in English or French.

Collaborate

I am most useful where deep technology has to meet demanding practitioners and a real market.

In their words

From people who built products, companies and research with me.

Correspondence

Open to conversations about partnerships, advisory work and leadership roles in product, data and AI.

Emailsteny@solitude.me
LinkedInlinkedin.com/in/stenysolitude ↗