We build AI applications and automations that take over recurring work: reading documents, creating invoices, syncing data, generating product images.
We use this in our own operations: our billing product reads invoices with locally hosted vision models, our analysis agent runs on the Claude API. Local models when data protection comes first, API models when performance comes first.
we choose the model by your data protection needs, not the other way around
we use the tools we build in our own daily operations
connected to your existing systems: ERP, invoicing, shop
every automation replaces a named manual task, not an end in itself
We run our own billing product that reads invoices with locally hosted vision models and classifies costs automatically. Local means the data never leaves the system it runs on.
We built our own analysis agent on the Claude API, with tiered models and fixed risk limits. Our rule: local models when data protection comes first, API models when performance comes first.
Daily scraper and sync pipelines via cron, automated quote and invoice creation through the Easybill API. Recurring manual steps become scheduled jobs.
Automated product image generation for a trading client: a Blender tool assembles configurations from CSV parts lists and renders them in a consistent perspective.
BIGmount is a modular mounting system whose variants combine many individual parts. Photographing every configuration in a studio was not practical: the online shop needed consistent product images for all variants, reproducible whenever the assortment changes.
The images in the shop come straight from the tool. View the project reference
That depends on the use case: extracting data from one document type costs considerably less than an agent that spans several systems. After a free initial consultation you receive a fixed-price quote for a clearly defined first use case.
Not necessarily. There are two paths: locally hosted models run on your own hardware or your existing infrastructure, so data never leaves your system. API models such as Claude run with the provider and offer more performance. We recommend the path that fits your data protection requirements.
Good candidates are tasks that recur in the same form: reading and classifying invoices and documents, creating quotes and invoices, syncing data between systems, generating product images. In the initial consultation we go through your processes and name the tasks that are worth it.
That is decided per project. With locally hosted models, your data stays on your systems and is not sent to any third party. If we use API models, we tell you which provider processes which data before the project starts. You make the call.
With a free 30-minute initial consultation, no strings attached. We then pick one small, clearly bounded use case as the first step. Once that runs in production, we extend from there.
Write to us or book an initial consultation directly. We reply within 24 hours, usually faster on weekdays.
Start Project Now