8 min read
The persistent memory advantage
Why useful organisational intelligence depends on preserving context, provenance and decisions over time.
Ideas for intelligent organisations
3 min read
The agent that can act is also the agent that needs governance
How bounded permissions, risk-based approval and audit trails let agents act responsibly.
4 min read
Your AI agent does not need a bigger brain. It needs a better memory.
Why durable, temporal and selective memory matters more than ever-longer agent histories.
3 min read
Agentic RAG: why enterprise search is becoming a conversation with the evidence
Why enterprise retrieval must inspect, compare and explain evidence instead of returning a single document.
3 min read
AI FinOps is coming: every agent will need an economic model
Why agent costs, latency and quality need to be measured at the level of the work they perform.
3 min read
The AI ROI gap is becoming a management problem, not a model problem
Why realised AI value depends on workflow design, adoption and measurement—not model choice alone.
3 min read
The real enterprise AI opportunity is not chat. It is workflow.
How bounded workflows turn capable models into repeatable enterprise work.
4 min read
The most expensive token is the one your agent didn’t need
Why selective context, compaction and retrieval make useful work per token the enterprise metric.
3 min read
Multi-agent systems will matter — but probably not for the reason you think
Why specialised context, careful orchestration and shared memory matter more than agent count.
3 min read
Rent the LLM. Own the intelligence.
Why durable enterprise value lives in memory, knowledge, tools and workflows around interchangeable models.
3 min read
AI agents are turning software from something you buy into something you can make
How agentic coding shifts the build-versus-buy boundary toward small, organisation-specific software.
7 min read
Enterprise search is becoming an evidence system
The best AI search does more than retrieve documents: it resolves scope, conflict and provenance.
7 min read
Evaluate AI in the work, not just on the benchmark
Enterprise evaluation should measure complete tasks, evidence quality and operational consequences.