One of the least glamorous problems in building our AI tools has been configuration drift. A service works on Monday, then a package update

Foundry began with a practical problem: how can a small consultancy reuse its own reports, spreadsheets and project experience without starting every assignment from a blank page?

Consider a fictional five-person environmental consultancy responding to a tender for a carbon baseline. The response is due on Friday.

Slop is the word for it now: that grey, bloated, everyone-has-read-this-a-thousand-times prose AI produces by default.

When an AI-assisted report contains a wrong number, the professional signing it cannot simply blame the model.

A client may be perfectly happy for you to use AI to format a report and deeply unhappy to discover that interview transcripts were uploaded to an external service.

A consultant asks how much electricity a local AI machine uses. Rather than compare it vaguely with a data centre, calculate it.

AI uses electricity, water and manufactured hardware. The footprint depends on what the system does, where it runs and how long it runs.

We track AI developments through use cases and lessons learned, not just model announcements. That gives professionals a way to decide whether a new tool deserves attention.

A five-person environmental consultancy does not need to buy a room full of equipment to begin using AI. It needs to decide what work is worth improving.

Imagine a five-person environmental consultancy monitoring regulatory notices across six countries. Each week an analyst visits ministry websites, checks publication dates, copies links into a spreadsheet and sends a short update.

A useful prompt gives the model a job, evidence and limits