Cambridge AI Works
Notes from the field.
What we build, what we learn putting AI into production, and what we'd do differently. Also where we announce company news.
Why org charts hold back transformation
Top-down, fanning out? if your org chart has this shape, it's telling you that there are silos actively preventing knowledge being shared. One reason org charts look the way they …
Read article →Achieving operational impact
Do we have to go all-out into a world of autonomous AI agents, and pick a new AI-first sales and marketing platform? I'm suspicious about this. I think these tools need data, and …
Read article →Choosing in house models
Chinese models are as good as free, especially compared to the sticker shock of Anthropic's Fable model. However, it's easy to forget that coding is just one thing these models …
Read article →Before you buy AI features, count what you already own
Every tool in your stack now sells an AI add-on. Some are worth it, but the pattern is worth examining honestly: rented capability, partial views, value that doesn't compound.
Read article →AI governance that doesn't slow the build
The security review stalls more AI pilots than the model ever does. Four engineering practices that get a system through review, and keep it trustworthy afterwards.
Read article →If you can't test it, you can't ship it
Language systems tend to fail quietly rather than loudly. Why we write the test before the system, and what a score held over time lets you defend.
Read article →Where AI earns its keep in a sales team
The unglamorous places AI pays for itself in a revenue team: CRM hygiene, call notes that stay, and hearing what the whole market is telling you.
Read article →