By Stevie Antonioni, Managing Director UK at Adnami
I have a confession: for the last couple of years, I’ve artfully dodged most conversations about AI. Not because I thought it was a fad, or because I didn’t care, but because I didn’t feel equipped to weigh in. And let’s face it: a lot of people in our industry are doing the same. We toss around terms like “LLM” or “automation layer,” nod knowingly, and hope no one asks a follow-up.
But avoiding the topic is becoming impossible. Last week at Web Summit, AI wasn’t just the hot theme; it was the air. Every stage, every panel, every startup stand had the same gravitational pull: agentic AI. Not chatbots, not copilots, not “AI-enhanced” dashboards. Actual systems that observe, decide, and act without waiting for someone to click a button.
You could feel a mix of excitement and discomfort across the room. Excitement because the potential is real. Discomfort because this is where the narrative stops being friendly.
For years, AI in marketing behaved like a diligent intern: eager, efficient, and ultimately dependent on human instruction. But agentic AI is different. It doesn’t wait for tasks – it identifies them. It doesn’t just optimise – it decides. And that shift forces a question we’ve been dancing around: what happens when your tools start acting like teammates with initiative?
Breaking boundaries
Most of us still cling to the idea that as long as humans “set the goals,” we’re in control. That’s generous. Let’s be honest – half the time our goals are vague, contradictory, or copied from last quarter’s deck because no one had the energy to rethink them. Handing that to an AI that can interpret, infer, and execute isn’t empowerment; it’s delegation with unclear boundaries.
And boundaries matter. The marketing ecosystem already struggles with transparency, accountability, and consistency. Programmatic taught us that automation, for all its brilliance, inevitably introduces opacity. Attention measurement taught us that even the most promising new metric can turn into a battleground of definitions. And now, we’re adding systems capable of taking action with even less visibility into how they’re deciding.
But here’s the twist: humans aren’t exactly predictable either. I’ve seen more questionable decisions made throughout the course of my career than any machine could dream up. So, the fear of AI behaving unpredictably is a bit rich when we’re still optimising multimillion-pound campaigns around outdated KPIs and gut feeling.
The real issue isn’t that agentic AI might make mistakes; it’s that we won’t understand the logic behind them. When a human f*cks up, you can ask, “What were you thinking?” When an AI does, you get a graph, a probability score, and an apology from someone in product.
What struck me at Web Summit was the split in the room. Some people were leaning in – imagining faster workflows, automated planning, creative iteration in seconds, decisioning across thousands of variables no human could reasonably process. Others were visibly tense, realising that once AI acts independently, it stops being a harmless assistant and becomes an operational force.
The future isn’t AI doing what we tell it. It’s AI interpreting what we meant to tell it. And that should make everyone pause.
Discipline and accountability
So where does this leave marketers, agencies, publishers, and the rest of us who’ve successfully faked AI fluency for years? With a very simple choice: either we confront the discomfort head-on, or we get comfortable being passengers. Because the truth is this: agentic AI doesn’t wait for you to understand it before it starts reshaping your workflow.
What we need now isn’t blind adoption or blanket fear. It’s discipline. Clear objectives, cleaner data, human oversight that actually interrogates the system rather than rubber stamping its outputs. We don’t need to be engineers, but we do need to be accountable.
The brands that thrive won’t be the ones with the most AI; they’ll be the ones that choreograph it properly. The ones that understand where agency belongs and where it doesn’t. The ones that don’t outsource their strategic thinking simply because the machine looks confident.
For me, Web Summit was a wake-up call. Not because agentic AI is coming – it’s already here, but because the people shaping it aren’t waiting for the rest of us to get comfortable. And if I spent the last year avoiding the conversation, I’m not making that mistake twice.
Spoiler alert, we’ve been building something at Adnami that leans into this shift rather than hides from it. I can’t say more yet, but there’s a press release landing soon and it’s going to be a game-changer for our partner integrations.
