Advise
We first work out soberly where AI actually has an effect in your process — and where it doesn't. We also classify the planned use case against the EU AI Act early, so you know which obligations apply before a single line of code is written.
AI & LLM Applications
AI that actuallyeases the workload.
AI/LLM applications are software that integrates language models or machine learning into an existing business process — for automation, assistance systems or data analysis. We build them securely integrated and measurably effective, not as a loose AI add-on: privacy designed in from the architecture up, assessed against the EU AI Act upfront, with EU hosting on request.
Who it's for
When AI applicationsactually pay off.
Not every idea needs a language model — we'll say so if a classic automation approach is enough. AI pays off when one of these applies:
- Staff visibly lose time on routine work like screening, summarising or sorting documents and requests.
- Large volumes of unstructured data — emails, PDFs, tickets, free text — sit unused because nobody evaluates them systematically.
- An assistance system should support staff on recurring decisions without replacing them.
- You're already planning an automation and want a sober assessment of whether and where AI adds real value — not just because it's trending.
The open ledger
Off-the-shelf or custom AI —compared honestly.
| Off-the-shelf AI tools | Custom AI application | |
|---|---|---|
| Entry cost | Licence per user, from day one — the cheaper start. | PoC to pilot typically starts in the low to mid five-figure range. |
| Time to value | Usable immediately — but generic. | First productive pilot in eight to fourteen weeks. |
| Process & data integration | Doesn't know your systems — knowledge ends at the copy-paste boundary. | Wired directly into your systems and data sources, with a human checkpoint. |
| Privacy & control | Data flows to the vendor; model choice and logs are only partly yours. | EU hosting through to on-premise — you decide data flow, model and logging. |
A read from project experience — often the honest order is: max out off-the-shelf tools first, then build deliberately.
How we work01 02 03
Advise, build,operate.
Build
We integrate AI building blocks directly into your existing systems and data sources instead of creating an isolated add-on. Model choice, prompting or fine-tuning stay documented and traceable, not a black box.
Operate
After go-live we track whether the application actually eases the workload — with monitoring for quality and the cost of model usage. EU AI Act governance duties stay in view during operation too, not just at launch.
What you can rely on
Accountability,not a black box.
Privacy by architecture
Personal data only reaches a model where it's legally sound and actually necessary.
EU AI Act assessed early
We assess the risk class and obligations of your use case before we build anything with AI.
EU hosting on request
Operated on AWS infrastructure in Frankfurt when your data shouldn't leave the EU.
No offshore
Development is owned on-site in Hamburg — fixed contacts, not rotating teams.
Frequently asked
AI applications —answered briefly.
What does an AI/LLM application cost?
PoC from low five figures
As a guide: a clearly scoped first use case — from proof of concept to a productive pilot — typically starts in the low to mid five-figure range. For a clearly scoped project we work fixed-price; otherwise time-and-materials with transparent billing. The first step is always a free, no-obligation intro call — after that you know your concrete range.
How long until an AI application is live?
Pilot in 8–14 weeks
Depends on scope, and on how much groundwork the EU AI Act classification needs. A first productive pilot is often achievable in eight to fourteen weeks. Timing is settled in the proposal, not discovered mid-project.
Can AI be integrated into our existing software?
Yes — the normal case
Yes, that's the normal case. We connect AI building blocks to existing systems and data sources instead of building a separate add-on that nobody maintains.
What does the EU AI Act mean for our project?
Phased: 2026–2028
The EU AI Act phases in over time — and the deadlines shifted again in June 2026 via the Digital Omnibus: obligations for stand-alone high-risk systems (Annex III) now apply from 2 December 2027, and for product-embedded high-risk AI (Annex I) from 2 August 2028. Already from 2 August 2026, however, the Art. 50 transparency duties apply — such as labelling chatbots and AI-generated content — along with the penalty regime. We classify your use case against this timeline before we build, instead of fixing it afterwards.
Where is my data hosted, and do you work with offshore teams?
EU possible · no offshore
Data fully within the EU on request, on AWS infrastructure in Frankfurt — which models or services are used, we agree with you transparently upfront. Development and accountability sit throughout with our team in Hamburg, not an offshore team.
AI application or custom software without AI — which fits better?
It depends
Depends on the task. If it mainly needs clear, deterministic logic, plain custom software is often the more robust and cheaper choice. More on our custom software development page.
Related services
Let's size up your AI projectin 15 minutes.
A no-obligation intro call — you describe your plan, we'll tell you honestly if and where AI really helps.
Book an intro call