Insight

Evidence before assertion

Field notes, cases and frameworks from real deployments, each marked with how sure we are.

Field note 04

Deployment starts with the workflow, not the model

The model is the easiest part to change and the least likely reason a programme fails. We map the work first, and the choice of model usually becomes obvious.

Read the note

Library

Notes, cases and frameworks

Everything we publish says what it rests on: one deployment, several, or a controlled comparison.

Field note

The baseline nobody measured

Why the first week of any deployment is spent timing work that already happens.

Read

Case

A council that halved its backlog

Relay in housing repairs, and the approvals we deliberately kept manual.

Read

Framework

Five questions before a pilot

Ownership, data, risk, measure and exit - answered on one page.

Read

Field note

When a model should say no

Designing refusals that route work to a person rather than to a dead end.

Read

Research

Adoption after twelve months

What still runs a year later across 38 deployments, and what quietly stopped.

Read

Case

Citations that survive an audit

Casebook in a regulated legal team, and the review trail the regulator asked for.

Read

Topics

What we write about

Every piece carries an evidence grade, from a single observed case to a controlled comparison, so you know how much weight it will bear.

Operating design
Governance and risk
Adoption and change
Measurement
Model selection
Cases

The monthly note

One useful note, once a month

No announcements and no round-ups. Unsubscribe from any issue.

AURELIA

Consultancy for putting AI to work inside real operations: mapped, built, adopted and measured.

The firm

Home Products Insight About Contact

Offices

London14 Charterhouse Square, EC1M 6AX
Edinburgh3 St Andrew Square, EH2 2BD