AI SYSTEMS FOR BRANDS

I design how brands work with AI. And I build the systems myself.

In plain terms: I build software that handles whole stages of marketing work on its own. It researches, drafts and checks everything against the brand's rules. A person always has the final say.

Idea, build, runBrands of every sizeHamburg · DACH

Fabian Roser in an enterprise workshop with a leadership team

Brands I've worked with

Selected agency references

01Where things stand

The will is there. The how is still open.

70^%

want to lead on AI in 2026^*

30^%

consider themselves ready to scale AI^*

42^%

use AI only as an assistant for single tasks^**

{}

Most respondents were large companies in North America and Europe. Every brand knows the questions behind the numbers, perhaps yours too: Where do we start? What comes next? And who do we bring in for it?

02An honest look

You started long ago.

AI is in the building, as it is in almost every marketing organisation: licences are running, a few people work with it really well, and there are first wins. That's a real start, and it counts. The path from this start to an operating model is the part very few have taken so far. Not through neglect: day-to-day business rarely leaves room for it.

How you can tell you're exactly at this point.

A lot depends on the same two or three people.They're genuinely good, and that's exactly what makes them the bottleneck. Everyone else works as before, or quietly gets by on private accounts: a sign that the official path doesn't carry yet.

Single tasks got faster, but the results barely show it.The gain stays inside the teams and gets absorbed elsewhere, as long as the process around it stays the same.

Nobody can say for certain what it delivers.There's movement, and there are good examples. But without checkpoints, the contribution can neither be steered nor defended in the next budget round.

None of this is a reproach. It's the normal state of play, a year or two in. The difference between organisations is made by what comes next.

03The solution

Impact needs both: system and adoption.

Most of the market covers one half. There's plenty of training without a system and plenty of systems without users. I do both, from the first decision to a system that runs day to day. So you don't get a concept you then have to implement yourselves. That I build things myself is proof and pace, not the actual service. You bring me in for three things: deciding which work belongs in a system; setting the quality bar a result has to clear before it goes out; running it, so the system actually gets used. The quality bar is where my background counts most. When production costs next to nothing, judgement becomes scarce. I've been deciding what good work is for 25 years: first as a designer, then as a creative director and managing director, today in the systems I build.

What I'm working towards with you.

Your recurring work runs as a chain, not as busywork.Carried through from brief to final asset, with your decisions at every handover. And carried by the whole team, not just the two or three who can.

What AI contributes, you can show, not just sense.With checkpoints instead of good examples. That gets AI through the next budget round. And it shows you which system is worth building next.

Your team comes along, and you can see it.No rollout against the team, but with it. Without fear, with more room for the idea. And in the end the official path is the most convenient one, so nobody needs the workaround any more.

04From practice

From product to visibility.

Marketing here means more than advertising: everything a brand creates value with, from product and range to communication and service. In these six areas I've built systems or developed methods, each time with a real commission behind it.

Product

Observations become tested product ideas.Market and trend signals turn into ideas, which the system tests against audience and feasibility until a business case stands.

Jägermeister: 25 leads from 20 markets, more than 3,000 concepts, nine business cases, four MVPs. The same chain now runs in Veya, the product-innovation software I build as a co-founder.

Market research

Synthetic audiences test an idea before it costs money.Simulated customers comment on concepts, claims and campaigns. It's a pre-test for the early phase. It doesn't replace real research.

Netto Marken-Discount: a market research platform with synthetic audiences.

Product range

Product knowledge software can understand.Descriptions, ingredients and attributes are prepared so that a system can answer questions about the range.

dm-drogerie markt: a knowledge system covering more than 45,000 products.

Communication

Campaign ideas and every format that follows.A virtual creative team develops several routes in parallel. A second chain turns the approved idea into every format, with people deciding at every handover.

Creative Studio, built within an engagement for a Hamburg agency network.

Visibility

So the brand shows up in AI answers.Many customers now ask ChatGPT and co. before they buy. The method checks whether and how a brand appears there. It also shows what needs to change in content and product data.

Own method (GEO), taught in 2026 as training for an agency.

Organisation

So the whole organisation works with it.Platform, rules and enablement: who may use what, how data stays protected and who brings the others along.

DHL: rollout with the works council, legal and IT. Plus the ongoing engagement with a Hamburg agency network.

„He is one of those who do, who kick things off. A creative leader with a digital heart, always fully committed.“

Christoph Pietsch · Chief Growth Officer, Publicis Groupe Germany

05Working together

Idea, build, run.

Three building blocks, one after the other. Discovery and system sprint come at a fixed price; running it continues for as long as you need it.

Block 1 · Idea

Discovery

Workshop · fixed price

A joint workshop that leaves you knowing which problems AI should solve for you.

→Where does work repeat itself and cost time and quality?
→Which use cases are worth it and which comes first?
→At the end: prioritised use cases, each with a quality bar and a checkpoint.

Block 2 · Build

System sprint

per use case · fixed price

One use case becomes a system that runs.

→Designed around your brand's rules, built by me
→Handed over with a quality bar, fixed checkpoints and rules for data and approvals
→Several sprints in sequence or in parallel, each at its own fixed price

Block 3 · Run

Ongoing support

ongoing

So the system gets used and gets better instead of gathering dust.

→Fixed sessions each month: sharpen quality, measure use, plan the next chain
→A quarterly look outward: what is new, what is ready and what you can safely ignore
→Bringing leadership along, in confidential 1:1s or as an interim AI lead with fixed days on site

Prefer to listen first? Keynotes and talks →

Common questions

Who is this for?

For brands of every size, from mid-sized companies to global groups: businesses that create value through brand, product and customer experience. It's aimed at whoever owns marketing in the broad sense there, meaning product, range, communication or service. Decision-makers at management and leadership level.

What is an AI system?

Software that works through several steps in a row instead of answering a single question. A chatbot writes you a text when you ask it to. A system takes the brief, researches, drafts several proposals, checks them against the brand's rules and puts the best ones in front of you to decide. The technical term is agentic AI. It works to your rules. A person has the final say.

Does this work for mid-sized companies too?

Yes. The need is the same as in a large group, just without an AI department of its own. Because I own everything from the decision to the running system, the effort pays off even for a single system.

Do you sell software?

I'm a co-founder of Veya, a software for product innovation. If Veya could be right for you, I say so openly before recommending it. You decide. Most systems I build for the specific case, to your rules and, wherever possible, on the tools you already have.

Who runs the system once it's finished?

Your team uses it; your IT or a service provider looks after the technical side. If you want, I stay on: I sharpen quality, measure use and keep building as your work changes. I don't run data centres or servers.

What does it cost to build an AI system?

Discovery and system sprint come at fixed prices. We agree the frame for the discovery in a first conversation; the price of each sprint is set once the use case is clear. You get a transparent quote within the same week.

Do you also offer AI training?

Yes, as part of the work: whoever is meant to use a system learns it on their own case. The AI-literacy requirement under Article 4 of the EU AI Act is covered along the way. Blanket training with no system behind it is something I deliberately don't do.

What about data protection and the EU AI Act?

Both belong in the groundwork. I design rules for tools and data to be GDPR-compliant and account for the obligations under the EU AI Act, from AI literacy (Article 4) to labelling (Article 50). In the DACH region, data sovereignty is an advantage when you shape it actively.

What are you not the right person for?

Prompting courses: it's a craft that shifts with every model release. And I'm not the one to push your creative tooling forward or tell you whether model X or Y is currently the better one. Nor do I run servers or roll out an AI platform across an entire group; there are large firms for that. What interests me is what sits underneath: which processes you run again and again, how they should run in future and how people and machines work better together in them.

What sets you apart from a strategy consultancy?

A consultancy writes up what I've done. I ran an agency with full P&L and was responsible for the creative work. Today I build AI systems myself, from the first draft to running them. Since 2024 I've been a Certified AI Officer of the German digital industry association BVDW. What you get from me in the end is a system that runs.

You don't have to know where you stand to talk. Talking is how you find out.

Book a call