If I built Brath again in 2026, I would not hire analysts. Not fewer analysts. None.
I would probably build a smaller agency overall, spend considerably more on senior specialists and use AI for much of the work that previously created leverage around them. That is the biggest organisational difference between the agency I built in 2012 and the one I would build today.
At Brath, I was the CEO and also the most senior SEO specialist. We had senior specialists who were responsible for customer projects, and we had junior employees we called analysts who did much of the labour-intensive work. They went through pages, assigned keywords, collected information and prepared material for the senior specialists. We also had Key Account Managers. They sold our services and held the customer relationship together, especially when a customer bought several services. A company using both SEO and SEM might have several specialists involved, while the Key Account Manager made sure the customer did not have to coordinate everyone themselves.
It was a sensible organisation for the technology we had. I would not rebuild it.
The analysts were useful because they gave senior specialists leverage. If thousands of pages needed to be reviewed, someone had to review them. If keywords needed to be mapped to pages, someone had to work through them. If competitors needed to be compared, someone had to collect and structure the information. A senior SEO specialist could then spend more time on the difficult part, deciding what the customer should actually do.
Today I would try to create that leverage with machines instead. Going through pages, classifying them, matching keywords, extracting information, comparing datasets and preparing a first analysis are all tasks where AI can remove large amounts of manual work. I would rather spend that salary on another experienced specialist.
There is an obvious problem with this. Analysts were also how we created future senior specialists. You learned by doing the boring work, seeing hundreds of websites, making mistakes, watching senior people correct them and gradually understanding which problems mattered and which ones only looked important.
If agencies stop hiring juniors, the industry eventually has a problem because someone still has to become senior. I am aware of that, but if I am building an agency, solving the industry’s training problem is not my job. I would not employ people to perform work that software can do efficiently because the industry needs an apprenticeship system. The industry will have to find another way to train them.
Senior specialists get the leverage
I do not think AI makes senior specialists less important. I think it makes them much more valuable.
At Brath, even the best specialist had limited capacity because every decision had a lot of work around it. Before deciding what to do, someone had to crawl the site, collect Search Console data, map keywords, compare competitors, investigate individual pages and prepare the material. The analysts handled some of that work.
Today I would build software around the specialist instead. A senior SEO specialist should be able to receive a structured view of a website, its rankings, competitors, content, technical problems and opportunities without spending most of the day creating that view. Then the specialist gets to the part I actually want to pay them for: deciding whether something is really a problem, whether it matters enough to fix, whether the AI has misunderstood the page and which of ten possible changes is most likely to improve the customer’s business.
The best specialist was never the person who could produce the most spreadsheets. It was the person who understood which number mattered. The old model was essentially one senior specialist supported by junior labour. The new model can be one senior specialist supported by machines.
That changes what I would spend the salary budget on. I would rather have several very experienced specialists than one experienced specialist surrounded by people doing preparation.
The Key Account Manager has to become something better
I would also change the Key Account Manager role. At Brath, the role made sense because someone had to sell, keep the relationship together and coordinate the specialists. This was especially important when the customer bought several services.
But a surprisingly large part of traditional account management consists of moving information between people. Ask the customer a question, send the answer to the SEO specialist, ask the SEM specialist for an update, prepare the meeting, write the notes, send the summary, follow up on the actions and update the CRM. I would automate as much of that as possible.
A person whose main value is moving information between the customer and the specialists has much less reason to exist today. The role I would keep is closer to a client strategist. That person needs to understand how the customer makes money, understand enough about the specialist areas to challenge recommendations and be able to prioritise across them.
If a customer buys both SEO and SEM, it is not enough to arrange a meeting with both specialists. Someone should be able to discuss whether the next dollar belongs in SEO or SEM. AI can prepare the meeting, collect the data, summarise previous discussions and produce the follow-up. The human should understand the customer.
I would hire an automation specialist early
This is one role we did not have at Brath and one I would seriously consider very early today. I would want someone who understands APIs, automation, AI systems, data and enough development to connect the tools the agency uses.
Their job would be to find work that should stop being manual. If a specialist exports the same data every Monday, investigate it. If a specialist spends two hours preparing the material for a ten-minute decision, investigate it. If someone copies information from one system to another, investigate it. If every customer receives essentially the same reporting process each month, investigate it.
At Brath, an inefficient process usually meant someone had to spend more time doing the process. Today, the first question should be whether a person needs to do it at all. This role is not there to replace specialists. It is there to remove work from around them.
Production becomes cheaper but judgement does not
The agency still has to produce things. Ads need to be created, pages need to be built, articles need to be written, campaigns need to be launched, tracking needs to work and reports need to reach customers.
I would use AI aggressively for the first version of much of that work. It can produce drafts, variations, structures, code, images, data transformations and reports. That does not mean quality control disappears. It means quality control becomes more important.
If the machine can create fifty ads in the time it previously took someone to create five, the valuable skill is recognising which two are worth running. If AI can suggest a hundred pages, the important question is whether those pages should exist at all.
Production becomes cheaper. Judgement does not.
I would keep the rest of the organisation small
Sales still needs a human decision-maker. AI can research prospects, prepare meetings, draft proposals, update a CRM and follow up on routine communication, but someone still needs to decide whether the customer is a good fit, whether the problem can be solved, whether the price is right and whether the agency should walk away. Early on, I would probably keep that responsibility with the founder.
Project management should also become smaller. Modern systems can create tasks from meetings, follow deadlines, identify overloaded specialists and combine project data with financial data. I would still want someone responsible for operations, but not someone whose primary contribution is keeping a project management system updated.
Finance is similar. Invoicing, bookkeeping and payroll can largely be automated or bought externally. What needs to stay inside the company is the decision-making. If a customer becomes unprofitable, someone has to decide whether to change the price, reduce the scope or stop working with them. Everything that does not require a full-time internal role would have to earn its way onto the organisation chart.
What the agency would actually look like
At Brath, the organisation looked roughly like this:
- CEO and senior specialist
- Senior specialists responsible for customer projects
- Analysts doing labour-intensive preparation and production
- Key Account Managers selling and coordinating customer relationships
- Supporting administration around the organisation
If I started again today, I would build something closer to this:
- One founder responsible for positioning, early sales and the important commercial decisions
- One client strategist responsible for understanding customers and prioritising work across specialist areas
- Two to four very experienced specialists
- One person responsible for operations and automation
- External support for accounting, legal work, payroll and narrow specialist needs
- AI doing a large share of research, classification, preparation, administration, first production and reporting
There would be no analyst layer. Depending on the services, that could mean starting with five to seven people, but the important difference is not really the headcount. It is where the humans sit.
I would put people where judgement matters and mistakes are expensive. I would put machines around them to remove as much preparation, administration and repetitive production as possible.
When we introduced the six-hour working day at Brath, we learned that productivity did not come from making people work faster. We had to remove work that should not consume their time in the first place. If I built an agency in 2026, I would apply exactly the same principle to the organisation itself.
Before I hired anyone, I would ask one question.
Why does a human need to do this?
