Twenty years of
knowing better

I spent twelve years as a producer and director of corporate film, and more than twenty as a motion graphics designer. Four hundred and twenty-two numbered jobs, mostly for agencies and production partners, for the kind of brands whose conferences have to go right.

That work taught me something the current wave of AI content hasn’t caught up with yet: an audience decides whether to believe you in about four seconds, and they decide it on craft signals they can’t articulate. Grade continuity. Whether a performance has intent behind it. Whether the edit has a spine. Get those wrong and it doesn’t matter how impressive the individual frames are — the message doesn’t land.

I use AI heavily now, and it has genuinely changed what’s possible. In 2026 I delivered a full film for a water company’s internal conference in eight days, on a brief that had been scoped as a stock-footage assembly with a voice-over. It was not producible by conventional means at that budget or that speed. That’s the upside, and it’s real.

But the same tools in untrained hands produce content that audiences quietly discount. Not because it’s AI — because it’s bad, in ten specific, nameable ways. Most teams generating their own content can’t see which ones they’ve hit, because seeing them is a trained skill that takes years to build and doesn’t come bundled with the software.

So that’s what Shapeshifter does now. I make AI content that doesn’t read as AI content, and I tell other people why theirs does.