Notes from the workshop
Plain spoken writing about custom AI solutions for business: when to build, when to buy, and how to make automation survive contact with real work. No hype, and no pretending AI is always the answer.
-
10 September 2026
What custom AI workflow automation looks like in practice
Not an agent with a personality. A quiet piece of software that does a chunk of your team's daily work. The five stages of a real build, from intake to audit trail, and why the review queue is the product.
Read the note -
10 September 2026
Enterprise AI solutions, translated out of the brochure
What the phrase means once the transformation language is removed: identity, data boundaries, audit, failure behaviour and change control. Why pilots stall before production, and the honest answer on compliance.
Read the note -
10 September 2026
Synthflow, Vapi, Retell, Bland: what the voice AI platforms get right, and when to build instead
We ran a production voice product long enough to learn where every platform's demo stops. The three layers people are really comparing, the four walls that end most pilots, and what building on your own number looks like now.
Read the note -
10 September 2026
Does your business actually need a custom AI solution?
Most of the people who ask us for one do not need one. How to tell the difference between a process worth automating and a problem that needs a better spreadsheet, and what affordable actually looks like when a build is justified.
Read the note -
10 September 2026
Custom AI integration: the case for putting AI inside the software you already use
The most common failure in business AI is a good model in the wrong place. Why integration beats another dashboard, where the person belongs in the loop, and the questions to ask before commissioning a workflow.
Read the note