The Anatomy of Stagnation: £15m Revenue, 20 Person Marketing Team, One Email a Week
The leadership team at a well-regarded UK direct-to-consumer (DTC) retailer, with a turnover of around £15 million, knew their email marketing was leaving money on the table. A lot of it. Despite a highly capable in-house marketing team of twenty, their engine of retention and repeat purchase was sputtering, managing just one, maybe two, mass-market email sends per week to their 300,000-strong subscriber list.
The process was a familiar story of friction and manual labour. It involved a weekly cycle of stakeholder meetings, data analysis paralysis, laborious copywriting, iterative design feedback loops with an overloaded graphics team, and finally, a white-knuckle session on Klaviyo to schedule the broadcast. The entire workflow consumed approximately 40 hours of team time per week. For one email. The opportunity cost was immense. Segmentation was basic, personalisation was a pipe dream, and the sheer effort involved meant that high-potential customer segments were routinely ignored.
They weren’t failing; they were stagnating, throttled by their own operational drag. This is the silent crisis in many marketing departments: headcount does not equal output. The traditional, human-centric model had hit its ceiling.
The default solution is usually twofold: hire more people or buy more SaaS. Hiring adds another salary to the payroll (circa £45,000 for a decent Email Marketing Manager in the UK) and increases coordination overhead. New SaaS tools promise automation but often just create another dashboard to manage and another set of templates to populate. They automate tasks, not outcomes. The team was already using Klaviyo, a powerful platform, but they were only scratching the surface of its capabilities because they lacked the bandwidth to feed the machine.
This is where the narrative shifts. Instead of adding more cogs to a broken machine, the brand took a radically different approach: they built an agentic AI workforce.
Building the Agentic Stack: An Orchestrated AI Workforce for Under £3k/month
Agentic AI is not just another tool. It is a system of systems—a coordinated team of specialist AI ‘agents’ that collaborate to manage complex, end-to-end workflows with minimal human intervention. Unlike passive AI tools that require constant prompting, an agentic system can strategise, create, execute, and learn.
The goal was to build a system that could autonomously manage the entire email marketing workflow, from strategy to deployment, at a scale impossible for the human team. The total monthly operational cost for this stack? Approximately £2,850.
Orchestration Layer: crewAI: The brain of the operation. We used the open-source framework crewAI to define the roles, goals, and processes of the AI agents. It acts as the project manager, ensuring each agent performs its task in the correct sequence and that they collaborate effectively. Think of it as the digital equivalent of an agency account director.
Reasoning & Creative Engine: Claude 3 Opus: For all high-level cognitive and creative tasks, Anthropic’s Claude 3 Opus was the clear choice. Its 200K token context window allows it to process vast amounts of information—brand guidelines, past campaign performance data, customer personas—and its reasoning capabilities are second to none. This was our ‘Strategist’ and ‘Copywriter’ agent.
Vector Database (Memory): Pinecone: An AI agent is only as good as its memory. We used Pinecone to create a long-term vector database. Every campaign brief, every piece of copy, every performance metric, and all brand documentation were embedded and stored here. This allowed the agents to learn from past activities and maintain brand consistency without needing to be retrained.