In this episode, Andrew Foxwell and Brad Ploch sit down with Hannah Houg and Matthew Gattozzi to answer the question every team is quietly avoiding: how do you use AI for creative strategy without losing the judgment that makes ads work? The short answer is that you bring AI a hypothesis instead of asking it for one. You'll learn which four parts of the process to protect from AI entirely, why the briefing stage is where ad quality actually degrades, and how to give your strategists room to think without torching your output.
Hannah has built ads and campaigns for 300+ DTC brands. Matthew's studio has driven $100M+ in client revenue. Both use AI daily. Both have watched it quietly weaken the same muscle they get paid for.
We discuss:
- Why every creative brief in the space suddenly looks the same
- The difference between AI helping your research and AI writing your hypothesis
- Why AI is good at following rules and bad at breaking them, and why that matters for scale
- How Hannah lost the ability to write a script after a few months of AI-first drafting
- What a genuinely good AI input looks like (and why it's built before you open the tool)
- The one-sentence goal every brief needs before anyone writes a word
- Why building 7 personas on a $10k/month budget is a math problem, not a strategy
- What "deep context" actually means, and why most teams don't have it
- Why the efficiency-vs-quality fight is a leadership conversation, not a tools conversation
Join The Scalability School Newsletter: https://www.scalabilityschool.com/scalability-newsletter
This episode is sponsored by Proppel, the hiring agency that has placed more than 75 employees for Foxwell Founders members. Get 10% off your first hire when you mention the Scalability School Podcast: https://www.weareproppel.com/foxwell-founders-proppel
This episode of the Scalability School podcast is sponsored by NorthBeam and they just launched Northbeam Incrementality. Northbeam Incrementality gives you easy, automated, self-service incrementality tests, while protecting you from the major mistakes so many people make while running incrementality tests. Your MTA handles the daily tactics, your MMM guides the long-term planning, and Incrementality provides the causal truth. It's a closed loop that allows you to scale what works and cut what doesn't. Right now when you head over to www.northbeam.io/incrementality, they're offering Scalability School listeners 50% off unlimited tests for a year when you join. Just tell them we sent you!
To connect with Hannah Houg send a DM at: https://x.com/hannahhoug
To connect with Matthew Gattozzi send a DM at: https://x.com/MatthewGattozzi
To connect with Andrew Foxwell send an email [email protected]
To connect with Brad Ploch send him a DM at https://x.com/brad_ploch
To connect with Zach Stuck send him a DM at https://x.com/zachmstuck
Learn more about the Foxwell Founders Community at https://foxwellfounders.com
Learn more about the The Hive Haus Creators Community at http://HiveHausUGC.com.
Chapters 00:00 Intro: why AI creative strategy is producing the same ads everywhere
04:08 Why every creative brief suddenly looks the same
07:18 What makes a great ad, and why AI can't break rules
10:22 Is research or briefing where AI actually hurts ad quality?
12:26 Why AI can't come up with the big idea
14:33 The danger of letting AI write your hypothesis
16:39 How AI quietly degrades your copywriting muscle
19:45 How to find an insight before you open AI
25:55 Why hyper-specific creative is winning on Meta right now
29:57 Why your team needs space to think, not more output
30:58 Has any LLM actually cracked ad copywriting?
33:00 Why creator talent match beats a perfect script
37:07 What a good AI input actually looks like
40:10 The one-sentence goal every creative brief needs
44:14 The persona trap that sinks small ad budgets
47:21 What "deep context" actually means
50:24 Why strategists want to slow down while owners push efficiency
53:27 Why this is a leadership problem, not a tools problem
56:32 Break your margins today for performance later
57:32 Bonus: the Michigan method