I told my Head of SEO his entire team was redundant. His face told me he thought I was either brink-of-a-breakdown insane or about to sell the agency. He was wrong on both counts.
That team cost me over £120,000 a year in salaries, plus another £30,000 in tools, link-building budgets, and digital PR retainers that consistently underdelivered. That’s £150,000. For what? A monthly report bloated with vanity rankings, a handful of blog posts that took six weeks from brief to publication, and a constant, nagging feeling that we were simply rearranging deckchairs on the Titanic.
This isn’t a story about saving money. It’s a story about exchanging a broken, human-powered, low-leverage model for an automated, high-leverage, engineered system. It’s the story of why we killed our SEO team and built a 620-day GEO engine in its place.
The Great Stagnation: Why the Traditional SEO Model is Broken
The fundamental problem with the search engine optimisation industry is the model. It’s built on sand. Agencies sell retainers, which incentivises activity, not outcomes. Clients get a 40-page PDF at the end of the month showing green ‘up’ arrows next to keywords that have never, and will never, convert to a single pound of revenue.
We were guilty of this ourselves. We were good at it. We could rank clients for their trophy terms. But I started to notice a fatal flaw in the logic, what I call the ‘Great Traffic-Revenue Disconnect’. We’d spend three months and £15,000 of a client’s budget to get them from position 4 to position 1 for a high-volume keyword. The traffic would spike. The client would be delighted. And then… nothing. Sales didn’t move. Qualified leads didn’t increase. The CEO would eventually ask a very reasonable question that the entire SEO industry fears: "So what?"
Traditional SEO is a craft. A dark art, some call it. It involves ‘optimising’ a page, begging other websites for links, and writing ‘quality content’ with the hope that Google’s black-box algorithm blesses you with its presence. This is not scalable. It’s not predictable. And in an age of Large Language Models and agentic AI, it’s utterly, hopelessly obsolete.
The blunt reality is that most SEO work is drudgery. Keyword research, competitor analysis, on-page tweaks, meta description writing, internal link mapping. These are not tasks that require the creative spark of a human brain. They are data-processing tasks. And machines process data infinitely better than we do.
Our team was talented. A head of SEO, two SEO managers, and a junior executive. They used the best tools—Ahrefs, Semrush, Screaming Frog. They attended the Brighton SEO conference. They talked a good game about E-E-A-T, topical authority, and all the other abstract concepts that sound important in a pitch deck.
But let’s break down the output of this £150,000-a-year function. In a good month, they might produce four high-intent blog posts, secure three-to-five backlinks from mid-tier websites, and conduct one technical audit. That’s it. For the price of a small fleet of Mercedes A-Class cars, we were getting a handful of digital assets and a report.
The breaking point for me was a Q3 review. We had successfully ranked a fintech client on the first page for "pension consolidation". A huge win, according to the team. The campaign had cost the client £45,000 over six months. I asked the simple question: how much resulting revenue had we tracked? The answer was a single, low-value lead worth less than £1,000. We had spent £45,000 to generate £1,000. It was a categorical failure disguised as a roaring success by useless metrics.
That evening, I started sketching out a new system. A system that wasn't reliant on human guesswork, manual outreach, or the whims of a junior copywriter. A system that treated search acquisition not as a marketing ‘channel’, but as an engineering problem.