Guides

How to put AI into systems you already run.

Written for the stage before anyone signs anything: what to automate first, how to reach a system with no API, and why most pilots never reach production. No gated PDFs, no email wall.

Where should we start with AI?

How to choose your first AI automation

The first project decides whether there is a second one. Pick it on arithmetic, not enthusiasm.

7 min read

Our software is ancient. Can we still automate?

How to connect AI to systems that have no API

Legacy systems are the normal case, not the exception. There are four ways in, and only one of them involves replacing anything.

8 min read

Why did our AI pilot never go live?

Why AI pilots stall before production

The model is almost never the problem. Five failure modes account for most of the gap, and all are visible before you start.

8 min read

Do we need to fix our data before we start?

How ready does your data need to be?

Not perfect. Ready enough for one workflow, which is a much smaller job than the data programme someone will try to sell you.

6 min read

Should we build this ourselves?

Build it in-house, buy a tool, or hire a studio

Three routes, three honest failure modes. The right answer depends on how specific your process is and whether you can hire.

7 min read

How do we stop it doing something stupid?

Guardrails: shipping AI without shipping risk

Six controls that separate a system you can put in front of customers from a demo you cannot.

6 min read

Not sure where you stand

Score your readiness in two minutes.

Eight questions about one process you want automated. You get a score, the failure mode most likely to bite you, and what to do first. Nothing is emailed to you.

Take the readiness check