Automating creative analysis to turn ad opinions into evidence
Benjamin Christopher Seer is Director of Blitz Growth, an agency that sells advertising decisions rather than activity. He built a pipeline that classifies every video and image asset against a fixed taxonomy and joins those labels to performance data, turning what used to be quarterly opinion into continuous evidence.
Prioritise measurement and classification before scaling AI creative production: automating analysis of assets and linking them to performance gives you actionable answers, while volume without measurement is wasted effort.
What’s the quick origin story of your business, and what makes what you do genuinely different from the other options your customers are looking at?
I built and recently sold Evercleaner, a just-add-water cleaning brand. Blitz Growth came out of that experience: I’d spent years on the client side of agency relationships and kept hitting the same wall – reporting that described what had happened but never told me what to do next.
What’s genuinely different is that I’ve been the person paying the invoices. When I tell a client a test isn’t worth running, it’s because I’ve already wasted my own money on it. Most agencies sell activity, because activity is easy to bill. We sell decisions.
The second difference is structural. Our analysis layer is AI-native: every client creative gets classified against a fixed taxonomy, hook, format, offer, talent, and joined to performance data. So we can answer "which hook actually drives ROAS on this account" instead of debating it in a meeting.
Where does most of your new business come from today, and where do you wish it came from?
Today: referrals, my own network, and outbound. A meaningful share traces back to mentoring at a VC, which put me in front of startups in a role where I wasn’t selling – trust arrives before the pitch, so by the time work comes up you’re not a vendor being evaluated.
Where I wish it came from: inbound. Organic search, and increasingly AI search. Referral is high-quality but it doesn’t compound and it isn’t a channel I control, I can’t turn it up when I want more of it. That’s exactly why we’re rebuilding our own site’s information architecture, schema and digital PR right now. We’re finally doing to ourselves what we do for clients, which is embarrassing to admit was overdue.
What’s one thing you’ve actually handed over to AI in the last year, and one thing you tried it on and went back to doing the human way?
Handed over: creative analysis. Classifying every ad creative against a taxonomy and joining it to performance data used to be a manual quarterly job. It now runs as a pipeline that reads the raw video and image assets directly. That single change moved us from having opinions about creative to having evidence about it.
Went back to the human way: client communication when something has gone wrong. We tried drafting it with AI and it was consistently well-written and slightly off. Anything containing an apology needs to be typed by the person responsible for the mistake, the customer can feel the difference, and being that fractionally wrong in that moment costs more than the time it saves.
Have you noticed a change in how customers find you or what they already know before they reach out? Is anyone arriving through AI assistants like ChatGPT yet?
Yes, though the change is in what people already know, not where they come from. Prospects now arrive having had a diagnostic conversation with an AI first. They come in with a hypothesis: "I think my CAC problem is creative fatigue", rather than a vague brief. Sometimes it’s right, sometimes it’s confidently wrong, but the first meeting starts further along than it used to.
On AI assistants as a traffic source: not meaningfully, not yet. I’d be sceptical of anyone in our industry claiming big numbers there. I suspect most of it is misattributed direct traffic. What I do think is real is the influence layer: people are being pre-informed by AI before they ever land on your site, and that shapes the conversation whether or not it shows up in analytics.
If you were advising someone in your industry on all of this for the next 12 months, what would you tell them to actually do, and what would you tell them to ignore?
Do: build the measurement layer before the generation layer. Everyone sprinted to AI creative production and now has ten times the output and no better idea which of it works. The real leverage is in classification and analysis, not generation. It’s less exciting and it’s where the money is.
Second: use AI as something that argues with you rather than something that agrees. On a market study recently it caught a seasonality artefact and a broken denominator in my own methodology, two errors that had put categories into a top-three list that didn’t belong there. A tool that only ever validates you is a liability.
Ignore: tool-hopping, and anyone selling "AI strategy" as something separate from strategy. Also ignore the pressure to publish more. AI made volume free, which means volume is now worthless. The scarce thing is a position worth holding and unique content.
Thank you to Benjamin Christopher Seer and the team at Blitz Growth for sharing what actually worked with Leaders Perception readers.
Want to share your playbook?
Leaders Perception publishes founder and operator interviews every week. A few questions, about five minutes, no calls. It is free and everyone we accept gets published.