New: AI services from web-crossing · Innsbruck

We make your company AI-fit.

From the first automated task to a fully autonomous AI agent: we bring artificial intelligence into your workflows safely and measurably - built on webx cortex, our GDPR-compliant AI platform.

webx cortex · Live pipeline
Your request
"Summarise the 34 applications …"
PII scrubber & injection filter
m.huber@… ->[PERSON_1] · input checked
LLM router
selects the right model for each task - cloud (EU) or local
100 % local
on request - your data never leaves your premises
GDPR-compliant by design EU & local instead of US cloud Model-independent (LLM router) Since 1999 in Innsbruck
Your Benefit

What can AI actually do for you?

Not "AI for AI's sake", but measurable relief: we find the points in your workflows where automation saves time and money right away.

01

Automating repetitive tasks

Data entry, transfers between systems, recurring texts and reports: what ties up your staff for hours today will run in the background in future - checked and logged.

Example: Incoming order PDFs are read automatically and imported into your inventory system in structured form - instead of being typed up by hand.

02

Faster turnaround times

Pre-qualifying enquiries, triaging inboxes, summarising long documents: AI does the groundwork, your team makes the decisions.

Example: Customer enquiries arrive pre-sorted with a draft reply at the right team - response time drops from days to hours.

03

Better content in less time

Product copy, translations, image descriptions, metadata: consistent, multilingual and accessible - in your tone of voice, not generic AI-speak.

Example: Alt text for your entire image library - automatically generated, compliant with the European Accessibility Act (EAA/BFSG), in German and English.

04

Putting your knowledge to work

Manuals, quotes, project documentation, emails: your company's knowledge moves into a vector database and becomes queryable via RAG, instead of sitting dormant in folders - with source references and access control based on your permission model.

Example: New employees simply ask the internal knowledge base - "How do we set up a project for client X?" - and get a sourced answer.

Our Services

From your first use case to a fully autonomous agent

We don't just build alt-text generation and translations for you - we automate workflows and develop agents that complete tasks independently. Step by step, at your pace.

Stage 1 · Immediate impact

Content AI

Quick wins with low risk - the ideal starting point.

  • Alt text for images (accessibility / European Accessibility Act (EAA/BFSG))
  • Website and document translations
  • Product and SEO copy in your own tone of voice
  • Metadata, summaries, editorial workflows
Stage 2 · Processes

Workflow automation

AI takes over the legwork within your existing systems.

  • Inbox triage and draft replies
  • Data extraction from PDFs, emails, forms -> ERP/CRM
  • Automated reports and summaries
  • Integration with your systems (TYPO3, online shops, Office, APIs)
Stage 3 · Our speciality

AI Agents

Agents research, verify and complete multi-step tasks - with checkpoints where your team makes the call.

  • Preparing quotes and invoices from enquiries
  • Data and content maintenance across multiple systems
  • Monitoring -> diagnosis -> ticket, automatically
  • Every step logged and traceable

For tourism & hospitality: With AlpineChat, we already offer a ready-made AI chat assistant for tourism websites - it runs on the same platform.

The Platform

webx cortex - the secure foundation

Every AI function we build runs on our own platform. It controls which model does the work, which data it sees - and who may access what. Once set up, you digitalise more and more tasks on the cortex foundation, step by step - without starting from scratch for every new use case.

LLM Router

The right model for each task: powerful for complex analysis, fast and inexpensive for bulk tasks. Models can be swapped at any time - no vendor lock-in, no dependency on a single provider.

PII Scrubber

Names, email addresses, phone numbers and other personal data are detected and masked before a request reaches a language model. The model works with placeholders - your real data stays with you.

Prompt Injection Protection

Inputs and outputs are filtered: manipulation attempts via crafted content (emails, documents, web pages) are detected and blocked before an agent acts on false instructions.

Local Models on Request

For sensitive data, we run language models on your own hardware or in our data centre in Innsbruck. Your data then never leaves your infrastructure - not even in anonymised form.

RAG & Vector Databases

Your company knowledge - manuals, quotes, project documentation, product data - is made searchable in a vector database. The AI answers via RAG (Retrieval-Augmented Generation) based on your own documents: verifiable, with source references, instead of freely hallucinating.

Permission Model & Access Control

A fine-grained permission model controls who may use which AI function and which knowledge sources they may query - integrable with your existing user management. The AI never shows anyone content they wouldn't otherwise be allowed to see.

Data Protection & Control

Yes to AI - but under control

Many teams are already using personal ChatGPT accounts - without a contract, without control, with real customer data. We replace this shadow AI with a platform your data protection officer(s) can actually sign off on.

Processing within the EU

Cloud models only via EU endpoints with a data processing agreement (DPA) - or run locally in Innsbruck or on your own premises.

Data minimisation built in

The PII scrubber ensures models only see what they need for the task - personal data is masked beforehand.

No training on your data

Your content is never used to train models - guaranteed contractually, or technically ruled out entirely with local models.

Logged & traceable

Every model call and every agent step is logged - you can always prove what the AI did and when.

Roles & access

Who may use which AI function with which data is controlled through roles - not trust and chance.

Documentation for your DPO

Documented ready for your record of processing activities: data flows, storage locations, legal bases - the final assessment is made together with your data protection officer(s).

Our Approach

AI-fit in three steps

No mega-project, no flying blind: we start small, measure the benefit and scale only what proves itself.

AI Potential Workshop

We look at your workflows and jointly identify the tasks with the best ratio of effort to benefit. Result: a prioritised list of concrete use cases including a data protection assessment.

½ day · fixed price

Pilot application

The most promising use case is implemented in production - on webx cortex, with real data and measurable criteria: hours saved, turnaround time, quality.

4-6 weeks

Rollout & operation

What works gets rolled out: further use cases, training for your team, monitoring and ongoing development. The cortex foundation is then in place - every further task you digitalise builds on it and goes faster than the first.

ongoing
In Practice

We don't sell anything we don't use ourselves

These applications are already running today - in our own agency and in client projects.

Accessibility

Alt text for thousands of images

AI describes image libraries automatically in multiple languages - directly in the CMS. This makes accessibility compliance (European Accessibility Act / EAA/BFSG) realistically affordable even for large media libraries.

Multilingual content

Website translations built into the workflow

Content is translated automatically as it's entered and submitted for approval - instead of being bundled up weeks later and sent off to a translation agency.

Editorial & Operations

AI in everyday editorial and agency work

Copywriting, image editing, summaries and internal automation - our own agency works every day on the platform we offer you.

Frequently Asked Questions

Good to know

Is using AI even possible in a GDPR-compliant way?
Yes - if the architecture is right. webx cortex processes data via EU endpoints under a data processing agreement (DPA), or entirely locally on request. Personal data is masked before every model call (PII scrubber), nothing is used to train models, and every call is logged. We provide documentation so your data protection officer(s) can make the final assessment on solid ground.
Does our data have to go to the cloud?
No. For sensitive data, we run language models locally - on your own hardware or in our data centre in Innsbruck. Your data then never leaves your own infrastructure. For non-critical tasks, more powerful cloud models can be used via EU endpoints - the LLM router keeps this cleanly separated per task.
Which AI model do you use - ChatGPT?
We deliberately don't commit to a single provider. The LLM router in webx cortex selects the right model for each task - from large cloud models to locally run open-source models. If a better or cheaper model comes along, we swap it in without changing anything about your workflows. No vendor lock-in.
Can the AI work with our own company knowledge?
Yes, that's one of the strongest use cases. Your documents - manuals, quotes, project documentation, product data - are captured in a vector database and queried via RAG (Retrieval-Augmented Generation): the AI answers based on your content and cites its source. A permission model with access control ensures that each person only reaches content through the AI that they'd be allowed to see directly anyway.
How do you make sure the AI doesn't do anything foolish?
Through checkpoints and technology: agents work with defined approval steps where humans make the decision. Inputs and outputs pass through filters (including prompt injection protection), every step is logged, and before going live we test every use case against your real scenarios. For us, AI never replaces responsibility - it speeds up work under supervision.
What does getting started cost?
The AI Potential Workshop is a manageable fixed-price entry point, and the pilot is a clearly scoped package - you know in advance what it costs and how success will be measured. We put together the binding quote after a short, free needs assessment. Ongoing costs (e. g. model usage) are always shown transparently and separately.
How quickly do we see results?
The first content applications, such as alt text or translations, often go live within a few weeks. An automation pilot is designed to run for 4-6 weeks - including measuring how many hours it actually saves. So within two months at the latest, you'll have it in black and white whether and how AI pays off for you.
Ready for the first step?

Let's find your first
AI use case

Tell us, in a free initial consultation, where time is being lost in your business - we'll tell you honestly whether and how AI can help.

Prefer direct contact? +43 512 206567  ·  info@web-crossing.com