Insights

Notes from the work.

We publish what we learn shipping AI integrations across industries. Patterns that survived a real engagement, sources for everything, and absolutely no generic AI takes.

Browse by topic

Featured

28 February 2026Department deep-dives11 min read

Where to put your AI lead: under the CTO, COO, or standalone?

CTO-line gets you delivery velocity but ships engineer-flavoured solutions. COO-line gets you process redesign but underinvests in eval infra. Standalone 'Head of AI' gets you executive air-cover but stalls on integration debt. We argue COO-line under 200 people, CTO-line above 500, with a 12-month checkpoint to detect a wrong placement before it calcifies.

Read the post

All posts

The full archive.

31 January 2026

Want a custom brief for your industry?

We write tailored notes for buyers who want the working version, not the published version. 45 minutes, no slide deck.

Request a brief

Notes from the work.

Where to put your AI lead: under the CTO, COO, or standalone?

The full archive.

Hire an AI engineer, or upskill an existing one? The honest maths at 50, 200, and 500 people.

Best LLM for sales call summaries: why most teams are buying the wrong tier.

From CRM to sales operating system: what changes when AI sits on top of your pipeline.

The real economics of AI customer support deflection: why most pilots lie about their numbers.

AP, AR, and close: where finance automation actually pays back in under 12 months.

Claude Cowork vs Claude Code vs Claude.ai: which should each team in your company be using?

Best LLM for spreadsheet QA: a field test across five real finance workbooks.

Want a custom brief for your industry?

Notes from the work.

Where to put your AI lead: under the CTO, COO, or standalone?

The full archive.

Hire an AI engineer, or upskill an existing one? The honest maths at 50, 200, and 500 people.

Best LLM for sales call summaries: why most teams are buying the wrong tier.

From CRM to sales operating system: what changes when AI sits on top of your pipeline.

The real economics of AI customer support deflection: why most pilots lie about their numbers.

AP, AR, and close: where finance automation actually pays back in under 12 months.

Claude Cowork vs Claude Code vs Claude.ai: which should each team in your company be using?

Best LLM for spreadsheet QA: a field test across five real finance workbooks.

Get the weekly note.

Want a custom brief for your industry?