Core
Layer I
Layer II
Output
Software · Generative AI
Sinthex builds tailored applications and puts generative AI where the work really happens: inside your existing systems, data and workflows. From a sober assessment of use cases through a solid prototype to day-to-day operation.
What an integration looks like with us
01
Your systems
02
AI layer
03
In daily work
Clients and projects we look after
From e-commerce and logistics to manufacturing, fulfilment, insurance and open source.
Services
A language model on its own does not solve a business process. We build the application around it: data access, permissions, interface, quality assurance and operation.
Web applications, portals and interfaces that match your process precisely – instead of off-the-shelf software that fits 70 per cent and pushes the rest into spreadsheets.
We apply language models where they demonstrably save time – embedded in your data, your permissions and your approval paths.
From a single assistant to safeguarded workflows: models that operate tools, prepare cases and leave decisions with your specialists.
AI is only as good as the foundation beneath it. We put data models, infrastructure and quality measurement into a shape that still holds in two years.
Products
Alongside client projects we build our own products. They are in daily use and show how we approach software: clear interfaces, clean integrations and data kept where it belongs.
macOS · Menu bar app
Dictation in the menu bar
Hold the Fn key, speak, let go – the text lands in the active field. Speech recognition runs locally with Whisper on Apple Silicon, so the audio never leaves the Mac. Correction by a language model is optional and uses your own key.
Cloud · Web app for desktop and smartphone
Warehouse management and shipping for Billbee
Cloud WMS for online retailers working with Billbee: scanner-based pick lists, goods-in, stock transfers and stocktaking down to the bin location, plus scan-based shipping at as many packing stations as you need. beeUnity extends Billbee rather than replacing it.
Use cases
Six areas where effort and benefit can be assessed reliably – and where a prototype shows within weeks whether it works.
Suggested replies from your history and knowledge base, with sources. People approve, not the model.
Read, validate and route quotes, invoices, delivery notes and contracts automatically into your systems.
Internal search across manuals, tickets and project knowledge – with cited sources instead of invented answers.
Descriptions, attributes and translations for thousands of items, consistent with your tone of voice and data model.
Qualify enquiries, prepare quotes, structure meeting notes and keep the CRM clean.
Structure and condense free text, feedback and protocols so they can feed real metrics.
Company
Sinthex is the brand of Sinthex Holding UG (haftungsbeschränkt), based in Kircheib, Rhineland-Palatinate, Germany. The company builds custom software and integrates generative AI into existing processes, data and systems – from use-case assessment through a prototype to production. Managing director is Nikolai Sinthern. Work is remote across German-speaking Europe, in German and English; data is processed in the EU.
Approach
Four phases with clearly defined outcomes. After each phase you decide whether and how to continue.
We review processes, data availability and systems, then rate candidate use cases by benefit, risk and effort. The result is a prioritised list – including the cases we advise against.
The most promising case is built and measured against your real data. You see quality, limitations and running cost before larger investment.
Connection to your systems, roles and approvals, plus tests, logging and documentation. The result is a solution that can be operated and handed over.
Monitoring, ongoing evaluation, model and cost optimisation, and further use cases on the same foundation.
Principles
Generative AI produces impressive demos quickly. In production it only holds up with clear rules.
Every use case needs a metric that shows success. If it cannot be measured, we do not start it.
Data minimisation, processing inside the EU, data processing agreements and – where the data demands it – models in your own infrastructure.
Providers stay replaceable. Models are chosen by task, quality and cost, not by marketing.
Comprehensible architecture, tests and documentation. You own your solution and stay able to act.
Critical steps run through approvals. AI suggests and prepares, your specialists decide.
Named contacts, honest assessments and transparent effort – including when the answer is "not worth it".
Technology
Selected by task and operability – not by trend.
Questions
What companies most often want to know before their first AI project.
With a workshop of two to three hours. We walk through your workflows, collect candidates and rate them by data availability, benefit and risk. Afterwards you have a prioritised list and a realistic view of the effort – including the areas where AI does not pay off right now.
We work with data minimisation: only the information a use case genuinely needs ever leaves your system. Processing takes place inside the EU and data processing agreements are part of the deal. Where data is particularly sensitive, we run models in your own infrastructure.
We work with fixed frames: scope, outcome and budget are defined in writing before we start. After the first conversation you receive a committed range instead of an open-ended list of hours – cost depends mainly on data availability and depth of integration.
Usually not. Generative AI is docked onto existing systems – via interfaces, exports or directly at the database. We only replace or rebuild what demonstrably slows the process down.
A small team with named contacts. You talk to the people who also do the engineering – no handover to rotating project staff.
Mostly remote, with on-site sessions where they help the project. Short feedback cycles, visible progress throughout, documentation in German or English.
Contact
Describe in a few sentences what you are working on. You will get an honest assessment of whether and how generative AI makes sense there – and which next step is actually worth taking.
Usually answered within one business day
