AI web applications: development and integration.
We develop web applications with built-in AI and integrate AI into your existing software: from customer portals and internal tools to an assistant that knows your documents. If you wish, everything runs on your own hardware.
What we build
From prototype to operation.
- Web applications with AI features: chat, document search, forms that understand input
- Integrating AI into existing systems: shop, CRM, ERP, ticket system via interfaces
- Customer portals and internal tools with login and permission management
- Local models on your own hardware or cloud models, depending on data protection needs
- Prototype in a few weeks, then operation and further development
What is an AI web application?
An AI web application is a program that runs in the browser, such as a customer portal or an internal tool, and uses a language model in one or more places. The language model reads texts, understands questions in plain language, summarises, sorts or writes drafts. The rest of the application is conventional software: login, database, permissions, forms, export.
The difference from a chat subscription such as ChatGPT: the application knows your data and your processes. It is connected to your ERP system, your shop or your ticket system and carries out a specific task instead of just answering questions.
Typical examples from small and medium-sized businesses
- Knowledge assistant: Staff ask "How was the warranty for customer X arranged?" and get the answer with a reference to the relevant document.
- Enquiry sorting: Incoming emails and forms are read, classified by topic and urgency, the key details are entered into the database and a draft reply is prepared.
- Quote generator: A quote draft in your own layout is created from a short description and your price list.
- Product data: Manufacturer details, photos and data sheets are turned into consistent shop texts in several languages.
- Customer portal: Customers see the status of their order, upload documents and ask questions that the AI pre-sorts.
- Analysis: free text from feedback, complaints or reports is turned into categories and key figures.
Why not simply use a ready-made tool?
Ready-made tools are fine when your process fits the tool. Often it is the other way round: you adapt your business to the software, pay per user and month and hand data to yet another provider. A lean application of your own does exactly what you need, and it belongs to you. Where a good ready-made product exists, we say so honestly and prefer to connect it rather than rebuild it.
Local or cloud: how we decide
Every application can work with a model on your own hardware or with a cloud model via an interface. Cloud models are the most powerful and cost money per request. Local models are good enough for most office tasks today, cause no ongoing fees after purchase, and the data does not leave the building. A mix often makes sense: internal documents locally, public product texts via the cloud. We build the application so that the model can be swapped later.
From first conversation to running application
- Initial consultation (free): Which task costs you time today? Which data and programs are involved? What must never leave the company?
- Concept: One to two pages: what the application does, which interfaces, local or cloud, rough effort.
- Prototype: Within a few weeks, a clickable version using your real sample data. This shows whether the AI solves the task reliably.
- Extension: login, permissions, logging, tests, connection to your systems.
- Rollout: Installation on your server or with a German hosting provider, training for your staff.
- Operation: On request, we handle updates, new models and extensions at a fixed price.
Important: AI makes mistakes. That is why we build applications so that a person checks results before anything goes out, and every answer shows which source it relies on.
What we build with
Python for server and AI, TypeScript with React for the interface, PHP for existing websites. Open standards, you own the source code.
Frequently asked questions
Local model or cloud?
Both are possible. We process sensitive data with models on your own hardware, and for everything else we use the interfaces of established AI providers. We recommend what suits your data and your budget.
How quickly is there a result?
A clickable prototype with real sample data is usually ready after a few weeks. Then you decide whether and how to continue.
Who owns the application?
Yours. You receive the source code and documentation; we handle operation and further development on request.
Request a project
Describe in a few sentences what the application should be able to do. We will get back to you with an initial assessment.
By phone: 030 417 17370
Briefly describe what you have in mind. We will get back to you with a concrete assessment.
Danziger Straße 78a
10405 Berlin
Berlin-Prenzlauer Berg