Guest · Ep 016
Phil Maggs on Fast Hours.
On Fast Hours
Episode 16 is a guest interview with Phil Maggs, Director of Generative AI at Superside, on scaling generative AI from a five-person skunkworks team to 500 trained designers.
The 30-second version
- 01
Superside grew its AI team from about 5 people to 500 trained designers in under two years via quiet 'shadow projects' on live client briefs.
- 02
In blind internal tests, the exact same image rated as 'colder' the moment reviewers were told it was AI-generated — the label changed perception, not the pixels.
- 03
Roughly 20% of any team are early adopters needing no push, 60% need structured coaching, and 20% will resist — forcing that last group is the wrong goal.
- 04
Superside has never shipped a raw Midjourney output across hundreds of thousands of images; every asset still goes through human post-production first.
The episode
Midjourney at Scale: The Superside Blueprint for Teams
ft. Phil Maggs
Episode 16 is a guest interview with Phil Maggs, Director of Generative AI at Superside, on scaling generative AI from a five-person skunkworks team to 500 trained designers. Phil describes blind internal tests where labeling an image "AI" made viewers rate the identical picture as colder, and explains why Superside has never shipped a raw Midjourney output without human post-production. He closes with tool picks and advice against over-regulating AI adoption before you understand it.
Open the episodePhil's stack, as they told it
- Flux ▲ Rates it
Phil calls it a 'step change' for brand-specific fine-tuning, replacing Superside's reliance on Stable Diffusion
- Figma ▲ Rates it
Superside's central hub for collaboration and automations, and the target of its brief-to-editable-design pipeline
- Stable Diffusion ◆ Mixed
Superside's prior fine-tuning backbone; Phil says Flux's arrival relieved his worry about depending on it
Phil, on the record
- 01 Blind test finding that the 'AI' label alone changes perception
“turns out if you just say, if it's AI, people then just have that feeling on the whatever image that you've said is AI. Would you look at that? Who knew?”
- 02 Superside's Midjourney seat spend, addressed directly to Midjourney
“$60 x 250, that's $15,000 a month, times 18 months on that. So mid-January we spent quarter of a million dollars with you, why don't you answer my emails please?”
Phil Maggs Ep. 16
- 03 Why every Midjourney output still gets human post-production
“We've never shipped anything direct from Mid Journey. I don't think once ever. Hundreds of thousands of images, not one time that we would just go, yep, good enough.”
- 04 Why Superside judges AI tools on volume of trials, not a single test
“But don't say it doesn't work on one. You don't know. You just don't know. Come back to me after the hundred one.”
Phil Maggs Ep. 16
- 05 His top advice for agencies just starting to adopt AI
“If you're a mid-size agency, don't overregulate is my number one advice. Do not put like the absolute iron curtain down before you even understand what it means to your business.”
