Applied AI lab
We're starting with the work that quietly eats enterprise time. Reading and re-keying documents, chasing approvals, and hunting for answers that already exist somewhere. Our first product, Proba, is in development.
Products
Software we build in the lab and share with the community.
Problem areas
Beyond Proba, these are the problems we're looking at next, and how far along each one is.
Contracts, policies, forms and reports that people read, check and re-key by hand.
Our angleAI that extracts, cross-checks and summarises, with a person approving anything that matters.
Requests that sit in inboxes and spreadsheets while someone chases the next sign-off.
Our angleRouting, reminders and draft decisions that fit the approval rules a team already follows.
Answers buried across wikis, shared drives and old tickets that take hours to find.
Our angleSearch that cites its sources and respects who is allowed to see what.
Problem areas may change as we learn. The Discord hears about it first.
Why "small seed"
Every product here starts as a small, testable idea. Most won't make it past the prototype. The ones that prove useful get the time and engineering to grow into software an enterprise can rely on.
Our approach
A deliberate, four-stage process decides which ideas earn more investment. Community feedback is part of every stage.
Pick a specific, recurring piece of busywork and agree how we'll measure whether it got better.
Build a small working version against realistic documents and workflows to test quality and cost.
Put it in front of community members, collect honest feedback, and stop if it isn't useful.
Harden what works with security, integration and monitoring, then keep improving it with users.
Principles
Commitments we hold ourselves to from the first prototype onwards.
We design for least-privilege access and audit trails. Customer data will never be used to train models without explicit opt-in.
For decisions that matter, AI drafts and a person decides. We'll publish how we evaluate quality and failure cases.
We design for single sign-on, standard APIs and existing data stores. No new silo.
We agree the measure of value before writing code, and report against it honestly, including when something didn't work.
Community
Join to see build notes as they happen, try prototypes early and tell us where they fall short.
Main community
Our main home for the lab. Engineers, operations people and anyone tired of enterprise busywork are welcome.
Also on