Everything with AI runs here.
This is the engine room of TheSEO. Here we train teams, build AI colleagues, and ask every question again as soon as the previous answer is older than 25 days, which in practice comes down to roughly monthly per question, to measure how visible Dutch brands are in the answers of a language model. Not a vision deck, but things that are switched on.
Three rooms, one lab
reg. AI.001What we build, concretely
reg. AI.002Every job in this lab falls into one of three build tracks. We start at the work that comes back every week, not at the technology. What you have built here is yours: it runs on your systems and your team can change it themselves after handover.
Workflows
Recurring work mapped out and then put into code. Quotes, reports, content production: from hours of manual work to minutes of checking.
ratio of steps, not a record of hours
AI colleagues
A digital colleague with one job and a name. Answers leads, watches your inbox or writes first drafts, and escalates to a human the moment it gets delicate.
a routing decision, not a lead time
Automations
Connections between the systems you already have. Data flows on its own from form to CRM to invoice, at three in the morning as well.
connection chain, example systems
The chain that keeps running after you go home
FIG.03 above draws the connection as three boxes: form, crm, invoice. What a still drawing does not show is the only thing you are buying, namely that the sequence keeps repeating without anybody pressing a button. Below it runs on. Five steps, always the same order, and then the next enquiry. The systems are examples and there is no client data in it.
Five rows, because that is the chain from FIG.03 plus the standby. There is deliberately no timestamp beside it: a time would suggest this is today's log, and this is the shape of a connection and not a measurement. What you can read off it is the difference between the three build tracks above. With a workflow you remain the button, with an AI colleague there is still a moment where a human looks along, and with an automation this row is the whole story.
What AI automation actually returns to your business
You hear everywhere that AI takes work off your hands. Yet at many companies it stalls at loose experimenting: a chatbot here, a prompt there, and after two weeks everybody is doing the work the old way again. A shame, because AI only works once it has a fixed place in your processes. Not as hype, but as a system layer that genuinely returns something. That is exactly where our work sits. We have been building digital systems for businesses since 2017, and AI workflows and automations are the logical next step in that. The difference: everything we build for clients, we use ourselves first. Every day, in content, sales, research and administration.
How we approach AI automation in your business
We do not start at the technology but at your work. Where is the recurring work that costs a lot of time and gives little energy? Writing quotes, putting reports together, following up leads, rewriting copy for every channel. Those are the tasks where automation proves itself first. Our method is deliberately simple:
- Map it out. We go through your processes and pick the tasks where AI adds value immediately.
- Build. We set up AI workflows that connect to the tools you already use, from WordPress to your mailbox.
- Hand over. We teach your team to work with it, so the automation keeps running once we are finished.
Want your people to learn it themselves? Our AI training connects to that directly: practical courses on AI, content production and more. Not theory for its own sake, but practising on your own work, with your own half-finished documents on the table. The training is given in English or in Dutch, whichever suits your team; say which when you book.
From free Claude Skills to a system of your own
The way in is smaller than you think. You can start today with our free Claude Skills: ready made AI workflows for quotes, analyses, strategy and decision making. Download, install, use.
Ordinary work counts here, not just the strategic kind. A spreadsheet formula you do not know by heart is written for you by the AI Excel Assistant, a page that gets visitors but no enquiries goes past the AI Website Check, and a cold email that must not read like a template is written with the Cold Email Writer. That way you find out without any investment what structured AI does to your output.
Unsure which thinking framework or canvas to reach for? two guides from the knowledge base point the way: deciding under uncertainty, eight models side by side and comparing strategy frameworks. For copy that has to persuade there is writing persuasive copy, the piece that bundles the writing models. Outgrow that and we build further. With Jarvis you get one brain between your AI tools, so knowledge and context are not lost with every conversation. Why that context keeps falling away and what an external brain changes about it is in AI memory between chats. And if you want to know how your organisation compares with other Dutch companies, look at the AI Index.
Who this works for
As a business you get the most out of AI automation when you think in processes. Owners who want to work faster. Marketers who want more output and more structure. Agencies and teams that want to scale up without taking more people on every time. Recognise yourself? Then the next step is small: email sales@theseo.nl and we will arrange a call with no obligation in which we pick one process together to automate first.
What you find here: the whole skill library
Under this path sit 96 skill pages, each with an explanation of the method, an example and a download you can drop straight into Claude. They are grouped below by the kind of work you reach for them for. Would you rather search by name? Use the full skill register; if you first want to know what a skill is and how to install it, read what Claude skills are and how to install them. Not every skill page has an English version yet; the ones that do are linked in English below, the rest are marked as Dutch.
Strategy and market analysis
The frameworks you use to put a market, a position or a choice on the table.
Deciding and taking problems apart
For the moment when you have to choose, or first have to work out what the question actually is.
Sales, copy and persuasion
Everything that goes out to a customer, from an email to a quote.
Work, planning and priority
The skills that decide where your day goes and what gets left.
Trying to make a useful routine stick? Start with the Atomic Habits Coach; it turns the method into a small plan you can actually repeat.
People, behaviour and conversation
For the work that is about people: coaching, changing and holding the conversation.
Why now
reg. AI.003Search is changing faster than most websites. More and more customers put their question to an AI assistant and get one answer back, not ten blue links. Anyone who does not appear in that answer simply does not exist for that customer. That is not a future scenario: the National AI Index records question by question which company names a language model gives, and repeats every question roughly monthly. That index measures the Dutch market, so read it as evidence that this is happening rather than as a reading of your own market.
AI answers without a click take over part of classic search traffic. How much exactly we do not measure: the AI index only counts whether a name appears in the answer, and for traffic we look at your Search Console. We have been building findability since 2017; this is the same craft with new tools. Whoever trains, builds and measures now is ahead later.
We run it ourselves too. Our own lead and client administration is not kept by people but by AI sessions, with a constitution setting out who may touch what and an audit that measures afterwards whether that is what happened. How that layer is put together is open and can be read on the explanation of our CRM layer. That layer has three phases, and each needs its own rules.
They are explained separately: the lead portal where nobody may be approached twice, the presales portal where a conversation becomes a proposal, and the client portal where it says per client what was done and why. That is not a shop window but the most honest way to show what working with AI asks for in practice: not only clever models, but above all clear agreements.
One legal note for readers outside the European Union: the EU AI Act applies to us because we are established in the Netherlands, and it does not apply in the United Kingdom. If you want to know what that means for a system we build for you, ask before we start rather than afterwards. The rules we hold ourselves to while doing this work, including the gate where a person reads and adjusts every AI output before it is published, are written down in our AI policy.
Frequently asked questions about AI and automation
What exactly is a Claude skill?
A folder with an instruction file that you hand to Claude. From that moment the model knows which method to follow, in which order, and what it has to ask you before it starts. The difference with a loose prompt is that a skill stays put: it works the same tomorrow as today and everybody on your team gets the same outcome.
Are they really free?
Yes. Every skill in the library above has its own page with an explanation and a download, without an account and without signing up. What you need is Claude itself. If you want one made to measure for your own process, that is work we do for a fee, but the library stays open.
What is the difference with an AI colleague or an automation?
A skill helps you with a task you carry out. An AI colleague takes over a recurring task inside a walled environment, and an automation connects systems so that no human is involved at all. In that order the return goes up and the number of decisions you still make yourself goes down; which layer fits depends on how fixed the process is.
Where do I start if I do nothing with AI yet?
With a skill for work you already do every week, not with the most complicated problem you have. A quote, a set of minutes, an intake: there you see within a day whether it helps. If that works, the next question is which process is fixed enough for a colleague or an automation underneath it.
Two pieces on AI and content, and one service that runs on it
The question we get most about AI and content is whether Google can tell. How Google reads AI content takes that apart: Google judges the value of a page and not how it was made, and the piece names the patterns that do stand out and what keeps a human stamp on a page. Making content for ChatGPT, Perplexity and Gemini is the practical half, with a step by step plan, the technical layer underneath it and five tests you can run on your own page.
There is also one service where AI does the production rather than the thinking. Social media management uses AI generated video for the posts, at a fixed monthly fee, and the page says what is in it and what is not.
