Munich sends us older problems than most cities do. Automotive, industrial manufacturing, insurance and reinsurance all sit on twenty or thirty years of operational data held in systems that were never built to be asked questions. The work is rarely the model; it is reaching the data at all, and doing it in a way the plant, the works council and the auditor can all live with.
Munich's industrial and financial concentration produces four fairly distinct kinds of request, and they behave differently enough that it is worth naming them separately.
Automotive and industrial groups arrive with maintenance, quality or supplier problems and a data estate that spans four decades of acquisitions. The engineering that pays for itself is usually unglamorous: getting sensor histories, work orders and inspection records into one place where a question can be asked across all three. Once that exists, prediction is comparatively easy. Before it exists, no model helps. Both halves are described under legacy modernisation and predictive maintenance.
Insurers and reinsurers arrive with documents — claims files, treaty wordings, loss reports, medical evidence — and a process where a person reads for an hour and then makes a decision that has to be defensible years later. The valuable system does not make the decision. It assembles the file, cites where each fact came from, and leaves the judgement with the underwriter, which is also the only version a regulator is comfortable with.
Aerospace and semiconductor suppliers arrive with material they cannot let out of their own network, sometimes for contractual reasons and sometimes for export-control ones. That resolves to a deployment inside their own perimeter rather than a hosted endpoint, which we cover under on-premise AI.
Everyone eventually asks the same question about scope: whether the pilot that worked on one line will work on eleven. It is the right question, and the honest answer usually depends on whether the eleven lines record data the same way. They rarely do.
No. We have one office and it is in Barcelona. We travel to Munich for discovery and handover and for site visits where being at the machine is the point. We would rather say that than describe a registered address as a presence.
Yes, and for a meaningful share of Munich work that is the right architecture. Open-weight models can be self-hosted on your own hardware, retrieval runs against your own storage, and nothing egresses. The trade-off is real and we will state it: self-hosted models are behind the best hosted ones on some tasks, and you take on the operational burden. Whether that trade is worth making depends on the task, and we would rather test it than argue about it.
No, but it changes the sequence. Any system that can be used to monitor employee performance or behaviour normally requires a works agreement, so the sensible order is to describe what the system logs and what it cannot do before anyone builds it, rather than presenting a finished thing for approval. In practice the councils we have encountered object to unbounded logging and vague purposes, both of which are also bad engineering. Your legal and HR functions own the agreement itself.
For a private-sector company, the Bayerisches Landesamt für Datenschutzaufsicht. Public bodies in Bavaria answer to a separate authority instead. The BayLDA has historically been willing to get into technical specifics, which is worth knowing before you write a two-page answer to a question that deserved a diagram.
Our ranges are on the pricing page, in euros and exclusive of VAT, and they do not change by city. The variable that actually moves the number here is integration: reaching data in older systems is where the effort goes, and we would rather scope that honestly in week one than discover it in month three.
Thirty minutes on what you actually have and whether it can be reached.
30-minute technical call · no deck
What we built, what stage it is at, and a link to the live thing wherever there is one.
yourOS is an operating system with an AI at its heart. You talk to it like a person, out loud or typed. It makes your calls, plans your days and works on your projects with a crew of agents, on your own private server. Our own product, built from the first commit; the waitlist is open.
Website, backend and app assets for Sonora, a free AI sound-therapy app on iOS and Android published by Cognitive AI.
Brand, website, prototype, apps on five platforms, the AI backend and the algorithm-intelligence pipeline for Podswap, a social growth engine that runs without a human in the loop.
In closed beta: Roxy, a private AI assistant with permanent memory running on infrastructure the user owns. Brand, site, product, backend and conversational design.
Live at coherence.com: brand identity generated with AI agents we directed, the motion system and website for the group behind Sonora, AboutAI, WakeUp and MudShadows. Agent work remains in build.
Live at domainbooth.com for DomainBooth, a premium domain brokerage: a new animated mark, every page redesigned and rebuilt with designed mobile layouts, and a custom CMS for its sales record, featured names, press and enquiries.
Apps, platforms and campaigns for these teams, long before the AI work. Every one is written up on the work page — what we did, and where our part ended.
