AI Services
AI consulting and automation from an agency that runs its own AI systems
You can buy AI strategy from a lot of places. Almost none of them use what they sell. We do, every day, and you can check it before you talk to us. The systems linked below are live on this website, they run without us watching them, and they are the reason our AI work is worth your time.
Collective Brain, a creative and AI marketing agency in Hamburg, Germany. Founded 2018, Google Partner. Services for business customers, all prices net of VAT.
Most AI consulting stops at the slide deck
The usual engagement goes like this. A consultancy runs a discovery phase, produces a maturity assessment, ranks twenty use cases in a matrix and hands you a roadmap. Then the roadmap sits in a shared drive, because nobody in the room has ever built the thing being recommended.
The gap is not intelligence. It is that advising on AI and operating AI are different jobs, and the firms that dominate the first one rarely do the second. Models change every few months. Costs move. A workflow that was expensive in January is trivial in July, and an agent architecture that demoed beautifully falls over the first time it meets your real data.
We took the other route. Every AI system we offer to clients, we built for ourselves first, put into production, and kept running through the parts that break. That is a slower way to build a service line. It is also the only way we know to tell you honestly what works, what costs more than it saves, and what is still a demo.
What we run in production, in public
Four systems, all publicly accessible, all built and operated in house.
A blog machine that writes and publishes our own magazine
Our magazine is produced by a pipeline we built. It researches topics, drafts articles against a house style, runs them through automated quality and compliance checks, generates the imagery, and publishes. Editors approve and correct. This is where most of what we know about production AI came from: prompt versioning, model routing, cost control, and how much human review a system like this genuinely needs.
Read the output (German) →An AI model leaderboard, rebuilt daily
Our AI leaderboard ranks the leading models on independent benchmarks, output speed and cost. It is rebuilt from its data sources on a daily schedule, and each entry carries the date of its last update. We built it because we needed the answer ourselves: choosing the right model for a task is the single largest cost lever in most AI systems, and the right answer changes month to month.
Open the leaderboard →Cost calculators for Claude and ChatGPT
We publish calculators that compare subscription pricing against API pricing, so a company can see what a given team size and usage profile actually costs before committing. AI budgeting goes wrong in a predictable way: companies buy per-seat licenses for work that should run on the API, or the reverse. The tools are currently in German. The underlying model applies wherever you are.
Claude calculator (German) →More than 50 published Claude Skills
We publish a catalog of reusable instruction packages that extend what Claude can do, covering design review, brand work, content production, SEO, automation and office tasks. Our team uses them daily. They are free to use, and they are also a fairly complete audit of how this agency operates, which is not something most consultancies are comfortable publishing.
Browse the skills →What you can hire us for
Strategy and use case selection
The first question is never which model. It is which of your processes is worth automating at all. We look for the same signals every time: the process repeats, it has a clear trigger, it costs measurable time, and the output can be checked. Processes without those four properties fail, regardless of budget. You get a shortlist of use cases with an estimated effort and cost for each, and an honest note on the ones we think you should not start with.
Automation and AI agents
Typical work: proposal and bid preparation, email triage and routing, meeting documentation, invoice processing, internal knowledge search across systems that do not talk to each other, and automated reporting. The backbone is usually a workflow platform such as n8n or Make, connected to one or more language models, integrated with your systems through the Model Context Protocol where available. Multi-step agents that act with limited supervision are now realistic for a defined class of tasks. They are not realistic for everything, and we will tell you which side of that line your process sits on. Strategy and architecture stay with Collective Brain. The build is delivered by our sister company WhiteFox Automations.
Visibility in AI answers
A growing share of your buyers now ask ChatGPT, Claude, Perplexity or Google's AI answers instead of running a classic search. Whether your company appears in those answers is measurable, and it is influenced by how your content is structured, sourced and published. We measure it, we report what changed, and we work on the inputs. What we do not do is promise you a citation or a position. Nobody can promise that.
How AI visibility works →Training and AI literacy
Article 4 of the EU AI Act has required a sufficient level of AI literacy among staff who work with AI systems since February 2, 2025, regardless of company size or risk class. There is no prescribed curriculum, which is exactly why most companies are unsure what satisfies it. We run practical in-house workshops for teams, in English or German, on site in Germany or remote, built around your actual tools and processes. We deliver training. We are not a law firm and we do not give legal advice on your compliance position. Workshops start at €1,290 net.
One group, three companies
Three companies, one group, three jobs. You deal with one team and get one quote.
does strategy, consulting, creative and marketing. This is where a project starts and where the responsibility for the outcome sits.
builds. Software development, workflow engineering, integrations. When an automation needs to be developed rather than configured, it is built here.
runs it afterwards. IT operations, monitoring, support. An automation that nobody maintains stops working within months, so someone owns it after launch.
The split exists because building software and running infrastructure are real disciplines that deserve real specialists, not a side activity for a marketing agency.
How a project actually runs
- 01
Analysis, one to two weeks
We sit with the people who do the work, map the process as it really runs, and identify where AI changes the economics. You get a written recommendation with cost estimates, whether or not you continue with us.
- 02
Build, typically four to six weeks for a first use case
Development at WhiteFox, with a working version in front of your team early rather than at the end. Scope is fixed before we start.
- 03
Handover and operations
The system moves to Progressive Mind for monitoring and support, and your team gets trained on it. Ongoing operations run on a monthly retainer.
- 04
Next use case
Almost every client starts with one process and expands once the first one is running and the numbers are visible. We would rather you did it in that order.
Data protection and the EU AI Act
We are a German company operating under German and EU law, which is the practical reason a number of our clients chose us over a provider outside the EU.
For each project we document which data leaves your systems, where it is processed, and which provider handles it. Where the data class requires it, we work with European hosting or models deployed inside your own infrastructure. Contracts include data processing agreements as standard.
On the EU AI Act: we classify what we build with you, we document it, and we cover the AI literacy obligation through training. For binding legal assessment of your obligations, work with your counsel. We will give them what they need from the technical side.
What it costs
The first conversation is free and carries no obligation. You get a fixed quote before anything starts, so there are no open-ended day rates.
- Workshops start at €1,290 net.
- A first automation use case is typically a five-figure investment net, depending on how many systems it has to touch and how clean the data is.
- Ongoing operations and support run on a monthly retainer, sized to what is being maintained.
- If your project also involves a website rebuild, our website redesign packages start at €18,500 net, with a paid audit at €2,900 net as the entry point.
We publish these numbers because most agencies do not, and because a conversation that starts with a shared sense of budget is a shorter conversation for both of us. All prices net of VAT, for business customers only.
Who we work with
A good fit is usually an established company with real processes, real data and a real reason to move, often an international business that needs the work delivered inside Germany or the EU. A poor fit is a company that wants an AI announcement rather than an AI system. We are not the cheapest option and we do not try to be. Selected projects are documented in our work.
Let's talk
Tell us which process is costing you the most time. We will tell you whether AI is the right answer for it, and if it is not, we will say so.
Collective Brain GmbH, Kattrepelsbrücke 1, 20095 Hamburg, Germany. We work in English and German.
Frequently asked questions
What does Collective Brain actually do in AI?
Three things. We advise on which processes are worth automating, we build the automations and AI agents through our sister company WhiteFox Automations, and we train your team to work with what gets built. We also work on how visible your company is in AI-generated answers. Everything we offer, we run internally first, including the pipeline that publishes our own magazine.
How are you different from a large consultancy?
Large consultancies have deeper benches and larger delivery organizations. What they generally do not have is a set of AI systems running in production under their own brand that anyone can inspect. Our AI leaderboard, our cost calculators, our published Claude Skills and our magazine pipeline are all live on this site. If you want a maturity assessment and a change program, a large firm may be the better call. If you want systems that run, look at what we have already built.
Do you only advise, or do you build?
We build. Strategy and architecture stay with Collective Brain, development is delivered by WhiteFox Automations, and operations afterwards by Progressive Mind. You get one quote and one point of contact.
Where is our data processed, and is this GDPR-compliant?
We are a German company working under German and EU law. For every project we document which data leaves your systems and which provider processes it, and we use European hosting or self-hosted models where the data class requires it. Data processing agreements are standard in our contracts.
What does the EU AI Act require from us, and can you help?
Article 4 has required a sufficient level of AI literacy among staff working with AI systems since February 2, 2025, for companies of any size. We cover that through practical training, and we document the classification of what we build with you. We are not a law firm, so binding legal assessment stays with your counsel.
What does an AI project cost?
Workshops start at €1,290 net. A first automation use case is typically a five-figure investment net, depending on how many systems it touches and the state of your data. Ongoing operations run on a monthly retainer. The first call is free and you get a fixed quote before anything begins. All prices are net of VAT, for business customers only.
How long does it take before something works?
Analysis takes one to two weeks. A first use case is typically live four to six weeks after that. You will see a working version early in the build rather than at the end.
Do we need to be based in Germany to work with you?
No. A good share of our AI clients are international companies that need work delivered inside Germany or the EU, often for a German subsidiary or a DACH market entry. We work in English throughout, and workshops can run on site or remote.
We already use ChatGPT and some Zapier automations. Is that a problem?
No, and it usually helps. It means your team has practical experience and you already know which tasks are painful. We generally start by reviewing what you have, keeping what works, and fixing the parts that break under load or cost more than they should.
Can you guarantee our company will appear in ChatGPT or AI Overviews?
No, and neither can anyone else. Those systems are not controllable from the outside. What we can do is measure whether your company currently appears, work on the inputs that influence it, and report honestly on what changed. Any agency promising a guaranteed citation or ranking is selling something it does not control.