The useful cost of AI
by the AI for leaders team
FinOps, model selection, agents and sovereignty: practical methods for moving from experiments to measured usage.
AI seats or tokens: calculate total cost without a false comparison
A reproducible way to compare a per-seat plan with usage billing, without confusing list price with total cost.
By Alexandrine Marjollet · November 5, 2026
Internal support assistant: trusted answers, controlled costs
Design an internal support assistant around authorized sources, human escalation, and cost per resolved request.
By Alexandrine Marjollet · October 17, 2026
Shipping an enterprise agent: the LLMOps practices to clear in testing
A demo that holds on a curated sample is not a production go. The test phase is where four practices (business evals, tool-call traces, drift and cost, rollback) decide whether traffic can widen.
By Florian Gabon · October 13, 2026
Central AI layer, embedded suite AI, or build: a decision matrix for the CIO
The PoC proved the use case. The next decision is ownership: build it, take AI already inside business suites, or hold a vendor-independent central layer the enterprise controls.
By Alexandrine Marjollet · October 9, 2026
Sovereign AI infrastructure vs Copilot and Claude: what an enterprise IT organization should actually compare
SaaS assistants make employees more effective. For a multi-BU enterprise information system, the real question is a controlled AI layer: data residency, governance, audit, FinOps, and exit from vendor lock-in.
By Florian Gabon · October 5, 2026
Mutualizing enterprise AI cost: budget governance and a vendor-agnostic central layer
SaaS seats, API tokens, business clouds, and shadow AI: without a central layer, AI spend stays a pile of contracts. AI for leaders recenters quotas, chargeback, and audit.
By Alexandrine Marjollet · September 22, 2026
Why an AI Gateway becomes the control plane of your AI stack
As models and tools multiply, without a single choke point Shadow AI moves to the LLM layer. An AI Gateway recenters auth, quotas, routing, and audit.
By Florian Gabon · September 21, 2026
Enterprise MCP: wiring agents into the IS without Shadow AI
A standard tool contract, real user identity, a governed catalogue: how MCP lets you connect agents to the IS without opening the door to Shadow AI.
By Florian Gabon · September 16, 2026
AI for leaders as a trusted AI layer inside the IS
Shadow AI, prompt leakage, unsanctioned tools: the real risk is not the model, it is the missing controlled layer in the information system.
By Florian Gabon · September 11, 2026