What is on the bench

Status labels are honest. Most of these will not become products, and we would rather say so than call an experiment a launch.

Experiments

In use internally

Invoice field extractor

Pulls line items out of supplier invoices into our accounting export. Runs on our own bookkeeping every month, which is how we found the edge cases.
Prototype

Schema drift detector

Watches a database for structural changes that break assumptions elsewhere in the codebase. Born from an incident we would rather not repeat.
Exploring

Estimate calibrator

Compares what we quoted against what projects actually took, to find where our estimates are systematically wrong. Uncomfortable reading so far.

Why we do this

Client work rewards using what you already know. That is the right instinct for delivery and a bad one over five years, so we keep a small amount of time for building things without a business case.

Some of it ends up in client systems — the invoice extractor started here before it became a feature in two document platforms. Most of it does not, and we learn where a technique stops working, which is more useful in an estimate than a demo ever is.

Working on something similar?

We are happy to compare notes, no pitch attached. Labs exists because talking to people who have hit the same wall is faster than hitting it again.