Insights
What we're working out.
Notes on building AI software that holds up in a real business. No thought leadership, no predictions.
CLIP: barriers to unlocking AI's real potential in financial services
The blockers in large financial organisations are rarely technological. Four human and organisational barriers, and what to do about each.
Smart strategies for LLM memory expansion
Large Language Models have a limited context window. Three ways to give them a longer memory: retrieval, agentic memory, and a bigger window.
Recognise any of this in your own business?
Most of what we write comes out of a client problem we had to solve. If something here sounds like your situation, tell us about it and we will say honestly whether we can help.