Volition Beauty

The Enduring Role of Human Judgment in AI-Driven Advertising

The speed is the easy part now. Feed the system a target and it will optimize toward it faster than any team I could staff. What it can't do is read the context around the numbers. The platform doesn't know a hero SKU is three days from selling out. It doesn't know a competitor just had a rough week in the press or that Tuesday's dip was a news cycle and not a creative problem. And it can't tell durable demand from bargain hunting. It sees cheap conversions and chases more of them, even when those are the buyers who convert once on a discount and never come back. So, I let the system optimize and I stay close enough to the market to know what the numbers are telling me before I act on them.

Balancing Short-Term Performance with Long-Term Brand Value

For me it comes down to time. Performance and brand only look like enemies because they pay off on different clocks. Performance shows up this week. Brand compounds over quarters and years. Grade everything on the fast clock and you'll starve the slow one and you won't feel it until the demand you used to get for free quietly dries up. So, I match the metric to the horizon and I don't let a short-term efficiency number settle a decision that's actually long-term.

The other half of it is that automation optimizes toward what it can measure right now. Much of what builds a brand, the creative that makes people feel something, the top of funnel that seeds demand months out, is invisible in this week's numbers. Leave that budget to the machine and it will defund all of it, because on its scorecard the spend looks lazy. Someone has to protect it and fund it on purpose. Then there's promo. Lean too hard on it and you train your best customers to wait for the next code. You've capped your own margin and cheapened the brand and every number still says you're winning.

Using the Right Metrics for the Right Decisions

Honestly, that's the wrong question for me. No metric is good or bad on its own, it depends on the decision you're making with it. ROAS is a fine gut check for whether something's roughly working. It gets dangerous the moment it becomes the budget bible, because then you're moving real money on a number the platform calculated for itself. Last-click, same deal. Useful for a quick read, useless for understanding what drove the sale.

So, the shift isn't swapping one set of metrics for another, it's matching the lens to the call. For a daily read, platform numbers are fine. For anything I'm committing real budget to, I want a number I can verify myself, which usually means incrementality and a real holdout rather than the platform grading its own homework. I've yet to see those two reads agree as closely as the dashboard implies they should. Most of the time the problem isn't a bad metric. It's one metric doing every job.

Developing the Capabilities to Lead Alongside AI

Three things, roughly in the order I'd coach someone.

Stay current. The platforms shift constantly and a lead who stops paying attention gets automated around without noticing. You don't need to know every button. You need to know what changed and what it costs you.

Then learn the system well enough to translate, which is the actual job of a marketing lead now. You sit between the machine and the business. One direction, you're turning what the algorithm is doing into something your leadership and your team can act on. The other direction, you're turning business goals into instructions the system can run on. Miss either and you're not steering it, you're just taking whatever it hands you.

The last one only gets more important as the tools improve, know your brand and know who you are. The more the machine can do, the more your value sits in what it can't have. A point of view, a clear read on what the brand stands for and who it's for and the nerve to hold that line when the optimizer is pushing you somewhere easier and cheaper.

Leading AI Adoption with Experience and Intent

What shaped it was the range. US, APAC, EMEA and categories that could not be more different. A major bank. An iconic Italian scooter brand. A global cement company. Cybersecurity. Beauty and haircare. A direct-to-consumer sleep brand. On paper none of them belong in the same sentence.

The sleep brand was the one that was genuinely AI-first and working inside it showed me what that looks like as a default rather than a side experiment. After that I couldn't stop seeing the gap. Everywhere else I walked into the same slow, manual, hand-stitched flows that operation had already solved years earlier and nobody in the room thought of them as a problem, because that was simply how the work had always been done. So, I spent a lot of my time helping those teams close it, rebuilding processes around what AI could actually take off their plate. That convinced me this wasn't a trend to wait out. There's a line I keep coming back to, AI won't replace you, but a person who knows how to use AI will. So, I stopped treating AI as something happening to marketing and started treating it as something I direct. The machine does the work. Knowing what to point it at and what to trust when it answers, is still the job.