Governance
What the EU AI Act actually asks of your model documentation
Most of the panic is misplaced. Here is the short version of what you need on file, who signs it, and when.
Lauren Fisk9 min read
Insights
Notes from delivery
Written by the people doing the work, not a content team. Mostly about modelling, cost control and what regulation means in practice.
Latest
Recent writing
Engineering
Your warehouse bill is a modelling problem, not a pricing problem
Before negotiating your contract, look at the twelve queries responsible for most of the spend.
Owen Mbeki7 min read
Machine learning
Why most pilots never reach production
It is almost never the model. Four organisational causes we see again and again.
Tasha Bramley6 min read
Analytics
One metric, one owner, one definition
How a semantic layer ends the argument about whose revenue number is right.
Devon Pruitt8 min read
Strategy
The data maturity ladder, and why most teams misplace themselves
Level two looks a lot like level four from the inside. Here is how to tell them apart.
Erin Castellanos10 min read
Governance
Writing a model card an auditor will accept
A practical template, with the three sections people usually leave out.
Lauren Fisk5 min read
Engineering
Testing data like you test code
Four tests that catch most pipeline failures before a business user does.
Devon Pruitt6 min read
Generative AI
Retrieval is a data problem wearing a model costume
Why most internal copilots fail on the corpus, not the prompt.
Tasha Bramley9 min read
Strategy
What a good data roadmap actually looks like
Three horizons, named owners, and nothing beyond twelve months.
Erin Castellanos7 min read
Tell us which number you cannot trust
Send one paragraph about the report that always starts an argument. A data lead, not a salesperson, replies within one business day.