7 min read
Every AI agent needs an identity of its own
Agent identity, delegated authority and short-lived credentials are becoming foundations of enterprise AI security.
Ideas for intelligent organisations
8 min read
Govern AI as a management system, not a policy document
Responsible enterprise AI requires continuous ownership, measurement and improvement across the operating lifecycle.
7 min read
Human oversight without approval fatigue
Good oversight concentrates human attention on consequential decisions instead of asking people to approve every step.
8 min read
Memory is becoming the operating layer for enterprise AI
Why enterprise agents need selective, evidence-linked memory rather than ever-longer conversation histories.
7 min read
Model portfolios beat model monocultures
Enterprise AI should preserve the freedom to route work across models as capability, cost and risk change.
7 min read
Start with workflows. Earn the right to deploy agents
A practical progression from reliable AI assistance to bounded autonomy in enterprise operations.
7 min read
The enterprise AI gap is now an operating design problem
Access to capable models is widespread. The advantage comes from redesigning work, ownership and feedback around them.
8 min read
The productivity evidence is real—and uneven
AI can create substantial gains, but the outcome depends on the worker, task and system around it.
2 min watch
Inside Iridium: how intelligence compounds
A studio preview of the video story format, pairing an immersive player with a concise written companion.
6 min read
Model independence in practice
A practical way to preserve organisational context while teams choose the models and interfaces that suit each task.
7 min read
Designing governed agent collaboration
How shared context, explicit boundaries and attributable actions help multiple agents work together safely.
5 min read
Evidence-linked answers at scale
Why an answer should preserve the sources, boundaries and confidence that made it useful.