''' The Agentic Stack: A UK Founder's Map of the £100bn Opportunity
Most commentary on AI in marketing is unadulterated nonsense. It’s written by people who have never run a P&L, managed a client relationship, or built a piece of technology from scratch. After selling my last agency for a headline figure of £10M, I saw the rot up close: marketing is a colossal cathedral of manual processes, held together by spreadsheets and human guesswork.
The industry talks about "AI" as if it's a magic wand. Slap an "AI-powered" badge on your SaaS tool and hope nobody notices it’s the same old predictive analytics model you were using five years ago. This isn’t transformation; it’s rebranding. It’s a fresh coat of paint on a crumbling wall.
My journey since that exit has been a complete rewiring. I’m not interested in building another agency or a slightly better analytics dashboard. I’m focused on the only thing that actually matters: building the autonomous, goal-oriented systems that will replace the entire marketing function as we know it. This is the shift from predictive to agentic AI, and it represents a £100bn opportunity that will leave most incumbent agencies and software companies in the dust.
Deconstructing the Hype: From Predictive to Agentic
For the last decade, "AI marketing" has meant one thing: prediction. Predictive analytics, powered by machine learning, is brilliant at finding patterns in huge datasets. It can tell you which customers are likely to churn, what product a segment might buy next, or the best time to send an email. It’s fundamentally passive. It provides an insight, but a human must still interpret that insight and, crucially, act on it.
This is the core limitation of 99% of marketing "AI" tools on the market today. They create more work, not less. They generate dashboards, reports, and recommendations that pile up in an executive’s inbox, all requiring a human team to design the creative, set up the A/B test, allocate the budget, and execute the campaign. The cognitive load remains firmly on the human.
Agentic AI is a fundamentally different paradigm. An agent is not a passive analysis tool; it is an autonomous actor in a system. You give it a goal, resources (like a budget and access to APIs), and a set of constraints. It then independently conceives of, plans, and executes the multi-step workflows required to achieve that goal. It doesn’t just suggest the next best action; it takes it. It learns from the outcome and refines its approach for the next time.
Here’s my contrarian take: Generative AI, for all its headline-grabbing ability to write copy or create images, is merely a feature. It’s a powerful and important one, but it’s a component, not the system itself. Generative AI is the party trick that gets everyone