Incoming invoices at a tax advisory firm
Staff only check the fields flagged as uncertain. Supplier, VAT ID, date, net figure, tax rate and gross total come out of a PDF or smartphone snapshot and reach BMD or RZL as a proposed booking.
Any one of them makes a valid entry point. For a company with no AI experience yet, the cheapest sequence runs like this: decide what deserves attention, prove it on genuine workload, give it a protected home, then coach the colleagues who will depend on it daily.
A basis for board-level decisions: which tools staff already use (usually more than management thinks), which ideas would earn their keep, and where GDPR or the AI Act draw a line. You receive a staged roadmap that can be signed off as it is or trimmed.
We clock a genuine task, add AI to familiar software such as Outlook, Teams, BMD or your Shopware or WooCommerce store, and clock it again. If the pilot does not deliver the saving we agreed on, it stops there.
An isolated set-up held in an account your company owns, whether on Azure or AWS in EU regions or at Exoscale, Hetzner and similar European hosts. GPU capacity rented by the hour, search across your own files, costs kept on a short leash. We do not sell hardware.
Department-sized video workshops using cases lifted from everyday work. Whatever the group, two reflexes get drilled: verify output before relying on it, and recognise information that must stay out of any assistant. This also supports AI literacy under Article 4 of the AI Act.
Promising tasks share three traits: they recur several times weekly, an occasional error is cheap, and a human reviews the result before clients or the ledger see it. Should one trait be absent, you will hear that from us before any quote exists.
Staff only check the fields flagged as uncertain. Supplier, VAT ID, date, net figure, tax rate and gross total come out of a PDF or smartphone snapshot and reach BMD or RZL as a proposed booking.
Nothing goes out until someone approves it. Incoming mail gets linked to its order, Austrian Post or DPD tracking is looked up, and the prepared answer gives the actual delivery status.
Each answer shows where it came from. Assistants at the practice look up the hygiene plan or the rule for short-notice cancellations, and the tool cites the matching passage from in-house documentation. It never offers a diagnosis.
Reception still has the final say. Messages in German, English, Dutch or Italian are sorted, checked against free rooms and package deals, and prepared as a reply in the guest's own language.
Nothing is published unchecked. A technical data sheet becomes a first draft shaped for Amazon, Kaufland.at or MediaMarkt Marktplatz, each following its own length and attribute rules.
Customer data has no business in a free chatbot opened with a personal account. Ahead of every trial we settle with you what may travel to an external vendor and what remains within Microsoft 365 or a dedicated EU-hosted environment. Personal data brings paperwork: a data processing agreement, a line in the processing register and, in some cases, an impact assessment. We document every decision so your data protection officer has solid grounds should the Austrian Data Protection Authority (DSB) ever ask. General guidance on the AI Act is available from the RTR's AI service desk.
A brief pilot on live data always comes before any rollout. Workshops, configuration and testing take place in video meetings, Monday to Friday, 8:00-17:00 Austrian time (CET/CEST).
We time how long the task takes today and agree a success criterion with you, for instance “nine emails out of ten reach the right department with no manual correction”.
A small group uses the tool on the day's actual cases for an agreed period. A person reviews every result, and we log hits and minutes saved.
We compare the results with the baseline. If the criterion is missed, you hear it straight and the pilot closes. A failed trial costs far less than a rollout nobody uses.
The tool is opened up to everyone else, backed by a short usage policy, regular spot checks and the paperwork your data protection officer will ask for.
Model developers are not the only audience of the EU AI Act. Which obligations apply depends on what a system is used for and what role your organisation plays. One duty affects virtually every business that deploys AI: making sure staff have sufficient AI literacy. We stick to the hands-on part. We list which tools are used, for what and with which data, and help you draft internal rules. The legal assessment of your situation remains with your lawyer or data protection officer.
You choose, within the limits a given vendor sets. Microsoft Azure, AWS and Google Cloud all offer EU regions, and there are European options such as Exoscale in Vienna, Hetzner, IONOS or OVHcloud. Wherever possible we choose services whose business terms pin processing to the EU and rule out training on your content. Because those terms change, we read them again before each project starts.
Yes, though not by pasting it into a free service. You need a legal basis, a defined purpose, a data processing agreement with the provider, an entry in your processing record and, where required, an impact assessment. Under Article 9 GDPR, health information counts as a special category, so we recommend a dedicated environment for it. Frequently the easiest remedy is transmitting less: sorting a complaint needs only an order reference and the reason, never a name or home address.
For drafting, summarising and searching your own files, sometimes yes, and we will tell you if that is the case. Bear in mind that Copilot sees everything the signed-in user is allowed to see. If SharePoint and Teams permissions have sprawled over the years, it will surface documents never meant for everyone. So tidy the sharing settings and brief the team first. Tasks tied to BMD, your ERP or the web shop usually call for something built to measure.
A pilot concentrates on one process and is charged as a fixed project fee or at €110 per hour plus VAT, with a ceiling agreed in advance. Model usage comes on top and is billed by the provider according to consumption: we estimate it beforehand and monitor it throughout. If you prefer an ongoing arrangement, AI work can sit inside a Start, Business or Premium plan. A dedicated team is quoted per project.
Tell us which job your team repeats every day and which program it happens in. We will look at the data available and suggest a pilot with a measurable goal.
Your enquiry has arrived
Our reply reaches you within one working day. Outages that leave your staff unable to work are dealt with first.
We could not find that town. Try another spelling, or choose whichever provincial capital lies closest; as everything is handled remotely, you get the same service in all nine Austrian states.