Practical notes on data, Adobe and applied AI
Architecture decisions, implementation lessons and operating guidance grounded in the work enterprise teams actually have to deliver.
Latest insights
Recent perspectives on enterprise content, reliable data, measurement and applied AI.
Applied AI
AI adoption without engineering discipline increases the blast radius
AI can accelerate useful work and also scale insecure access, untested decisions and invisible dependencies. The answer is a safe paved road, not blanket prohibition.
Read moreApplied AI
The dangerous journey from AI proof of concept to production
A successful AI demo proves possibility under controlled conditions. Production requires evidence about risk, reliability, operations, cost and real user outcomes.
Read moreApplied AI
Prompt management is not enough
A prompt only has meaning beside the model, tools, retrieval settings, evaluations and production traces that determined its behaviour.
Read moreApplied AI
Why company knowledge assistants hallucinate—and what to test
A wrong knowledge-assistant answer may begin in ingestion, retrieval, context, generation or citation. Testing only the final response hides the cause.
Read moreApplied AI
Enterprise RAG security starts with retrieval permissions
A knowledge assistant must never retrieve evidence the user could not access at source. That boundary belongs inside retrieval, not after generation.
Read moreApplied AI
Enterprise RAG fails at ingestion before it fails at the model
A knowledge assistant cannot recover structure, context, permissions and freshness that its ingestion pipeline discarded.
Read more