AI Insights

Governance, evaluation, reproducibility.

Analysis and perspectives on AI governance, compliance with reference frameworks, and what it concretely means to make AI governed and accountable.

Latest analysis.

Featured · August 24, 2026 · 12 min read

Bell AI Fabric and Cohere: an evaluation note for regulated organizations

The Merritt stack is a real improvement in operational and contractual control, and it is not a legal firewall. What is live as of August 24, 2026, what the sovereignty claim withstands, and why the decision belongs to a use-case tiering exercise rather than a vendor comparison.

Three tiers, one platform each

01 Assistive, human review
02 Regulated data
03 Direct consequence

June 23, 2026

Method · 7 min read

Move 38: the advantage is in the move after AlphaGo

In 2016, AlphaGo played Move 37, the move no human would have played. The lasting lesson is in the move that follows: Move 38, the one that belongs to the human, who decides, frames and answers. Why it holds for a small business and a multinational alike.

June 17, 2026

Trends · 8 min read

AI and data trends 2026: governance is the advantage

The 2026 trend maps converge on agentic AI. Read closely, the list rewards the organization that defined, measured and governed its AI. An agent amplifies the governance of its context, not the speed of its launch.

June 14, 2026

Security · 9 min read

Securing the generative-AI workflow: six steps, from dataset to feedback loop

A generative AI's security is distributed across six steps, from data acquisition to the feedback loop. The whole system is only as strong as its least-governed link. Here are the six steps to govern, and the single inventory that ties them to the four frameworks.

June 10, 2026

Regulation · 8 min read

AI Act 2026: what enforcement concretely changes for Canadian businesses

The European AI regulation has come into force. Even in Canada, organizations that process data from European residents or deploy high-risk models are directly affected. Here is what this means for your governance program.

June 05, 2026

Evaluation · 7 min read

Why a strong AI Index isn't enough to earn the designation

A high overall Index looks promising. But if a single property falls below the floor, the designation is refused, because the ten properties are not interchangeable. Here's why the floor per property exists and what it means for your program.

June 01, 2026

Positioning · 7 min read

Generic is the default. Identifiable is the advantage.

Access to AI has become a commodity. Everyone can buy a model and prompt it. Competitive advantage is therefore no longer about access; it lies in AI the organization can name, explain, and own. Here is the case we make.

Published: June 2026 · Updated: June 23, 2026

The topics we cover.

Reference frameworks

ISO/IEC 42001, NIST AI RMF, AI Act, Law 25. How to read them together, how to apply them without conflating them, and what they concretely require from your organization.

Evaluation and measurement

The Responsible AI Practice Program, the iDIA framework, the ten properties, the designation floor. Everything related to reproducible measurement of AI maturity.

Practical governance

AI officer, acceptable use policies, review committees, system registries. Governance as it actually takes shape in a real organization.

Reproducibility and auditability

A result that cannot be reproduced is not a result. How to document inputs, trace outputs, and make an AI system accountable before a regulator or a board.

Don't wait for the audit to set the framework.

The most useful first step is the diagnostic. Twelve questions, ten properties, an express Index from 0 to 100 in five minutes.

Get your AI Index Talk to identifiable