Most of what you hear about AI is noise. A thick, impenetrable fog of webinars, LinkedIn gurus, and six-figure 'prompt engineering' courses designed to sell you a dream that rarely survives first contact with a real P&L.
They're selling shovels in a gold rush, but they've missed the drill bit. The real opportunity isn't teaching a human how to ask a chatbot a slightly better question. It's about removing the human from the loop entirely.
This is agentic AI. And it represents the most significant platform shift in business since the internet. After exiting my last marketing agency for just over £10 million, I could have packed it in. Instead, I've gone all-in on building the agentic future. Why? Because it's a £100bn opportunity that will either be captured by those who build it or missed by those who cling to the past.
This is the agentic stack. My map of the new territory. It's the playbook I'm using to build my new company, and the framework that will determine the winners and losers of the next decade in marketing.
The Great Stagnation: Why SaaS and Agencies Are Broken
For the past fifteen years, we've been sold a lie. The lie is that more software is the solution. A tool for social scheduling, another for email, another for analytics, another for customer support. A sprawling, disconnected archipelago of subscriptions, logins, and monthly fees.
This is the SaaS delusion. An obsession with creating single-player tools that require a human operator to function. Each tool solves a micro-problem but creates a macro-headache: a complex, fragmented workflow that requires a team of expensive humans to stitch together. The modern marketing department is drowning in this complexity, spending more time managing its tools than doing actual marketing.
And what's the solution we've been offered? The agency. An outsourced team of human APIs. You pay a five or six-figure annual retainer to a third party so their employees can log into your software and pull the levers you don't have time to. It's a model built on billable hours and human inefficiency. A model that is about to be deprecated.
I saw it with a FTSE 250 retailer we consulted for. Their tech stack was a museum of overlapping SaaS contracts, costing them north of £1.2M annually. Their agency retainer was another £600k. Yet, their core metrics — customer acquisition cost, lifetime value, and market share — had been flat for three years. They were spending more to stand still. This isn't an isolated case; it's the norm. It's the great stagnation of marketing, and agentic AI is the asteroid.
The only way out is to build systems, not buy tools. To architect for autonomy, not assistance. The Agentic Stack is my blueprint for how these new systems are being, and will be, constructed. It consists of three interdependent layers.
This is the engine. The foundational Large Language Models (LLMs) like those from OpenAI, Anthropic, or Google's own Gemini family provide the raw cognitive power. Think of them as a pre-trained, highly capable graduate. They understand language, can reason, and can process vast amounts of information. But left alone, they produce nothing of value. They are a resource, not a solution. The mistake many founders and CTOs are making is thinking the LLM is the product. It's not. It's a component.
This is the brain. This is where the magic happens and where the real intellectual property is built. The agency layer is a sophisticated collection of proprietary code, workflows, and decision-making frameworks that sits on top of the Intelligence Layer. It takes the raw power of the LLM and directs it towards a specific business outcome. It gives the model a goal, a memory, the ability to use tools, and a feedback loop for self-correction.