Peter Handley, AI Marketing Engineer

Now

Day to day I'm building AI content workflows at Sage, and a lot of my time goes on testing them and reading run records.

They're research, writing, visual and QA agents on a vectorised knowledge source, so our own data and points of view make it into the content. I build with Claude, Codex and Gemini (and I'm experimenting with local models), alongside software engineers on the system we actually use.

How I work

I'm a helper at heart. I'll argue my corner when it's my area, then get behind wherever the team lands.

A lot of my career has been translating between marketing and the people who build and run the website - why schema matters, why a page needs words on it, how a request fits into a sprint. Years of that built trust with our developers, and I think it's a big part of why we've been able to start using AI safely on the site.

FAQ

What is an AI Marketing Engineer?

Probably GTM engineering with AI super powers. Right now that's automating go-to-market work, from finding the right prospects in target account lists to the knowledge that tailors our content to their pain points. It reaches across social, search and marketing automation too.

How did you start doing this?

Mostly curiosity. I started in earnest in June 2025, copying code snippets out of chats and asking more than one AI, until I was working with AI in code. What I learn gets written down (first a CLAUDE.md, now a wiki), and anything I've done about three times becomes a skill. It never clicked in one moment, more a gradual "oh, this is better for that".

How do you know if AI content works?

The content is only a leading indicator. What I'm really after is pipeline at a cost that beats what we pay for it now - a better LTV:CAC and a short payback period.

Why all the hobby projects?

Most of what I build at work can't be shown, so my hobbies have become a test bed for trying a technique properly before it goes near the day job. None of it is commercial (yet?).

Built

GM Assistant: the iPad app I run my tabletop campaign from.
Woodcut image model: trained on public-domain woodcuts, so the GM Assistant's art leans less on other people's work.
Weyland-Yutani data slate: a spare iPad turned into an AI-powered prop for our Alien RPG game.
3D village: a campaign location as a walkable diorama (parked, but I learnt a lot).
3D printed props: AI-designed, printed for the table.
Back in 2012-13: one of the first hreflang sitemap tools.