Choosing the process
We look at how a department actually works and hunt for repeated steps, such as retyping orders that wholesale customers send over as PDF or Excel files into the sales system.
The winners are rarely glamorous. This is what a typical project covers, from choosing the spot to the team signing it off.
We look at how a department actually works and hunt for repeated steps, such as retyping orders that wholesale customers send over as PDF or Excel files into the sales system.
Are there examples from recent months, and are they usable? A hundred orders in five layouts is good material; three emails is not enough.
One scenario, a small group of users and a threshold fixed in advance: how many lines must be recognised without edits and how many minutes must be saved.
Results land where the employee already sits: a draft document in the ERP, a note on a ticket, a ready reply in Outlook or a post in a Teams channel.
A colleague reviews every result for as long as the hit rate stays below target. Human oversight is also an AI Act topic worth discussing with a lawyer.
If customer or employee details flow through the model: a processing agreement with the model provider, an entry in the record of processing activities and, where needed, a DPIA agreed with your data protection officer.
How long it takes from the first call to a go or no-go decision depends on the process and the state of the data. Everything runs remotely, and the pilot figures set the direction.
We talk to the people doing the job and time how long each automation candidate takes them today.
One process with a clear measure. Running several pilots in parallel is a reliable way to get no clear answer from any of them.
For the first days the model only suggests while the employee works as usual. We compare both versions and refine the instructions.
When the numbers back it up, the rest of the department gets access, and we set up quality monitoring and hand over documentation.
Clean data is only half of it; the other half is whether staff believe the tool. Anyone who doubts the output will quietly return to doing the job by hand, since that beats double-checking software. So the eventual users sit in the pilot from the first morning, flag every miss and watch their comments reshape the behaviour.
Often, yes. Many popular systems, such as Comarch ERP Optima, Subiekt nexo, enova365 or Business Central, offer an API or an import route, and we check what your version provides at the outset. With an old, closed system we start with an interim step, such as an import file that someone approves with one click.
We are open about what changes: retyping and sorting disappear, while judgement and customer contact stay. We do not promise headcount cuts, because the gain usually shows up as time for the backlog. Letting the team take part in the pilot calms those worries faster than any slide deck.
We can take it on under an ongoing support contract, keeping an eye on model costs, output quality and changes on the provider side. Alternatively you receive full documentation and your own IT team runs it. This is agreed before we begin, so the system never ends up without an owner.
A pilot on one process is quoted as a fixed fee, normally covering several weeks. Further work follows only if results justify it, charged at PLN 190/hour excl. VAT. If the pilot fails to show savings, closing the project is the honest move, instead of pressing ahead on momentum.
Which routine chores eat most of your team's week? Send a short description and we will come back with a pilot idea based on your data.
Your enquiry has reached us
You will hear back within one working day, and if you have reported an outage that is holding up work, it goes to the front of the queue.
No match for that name. Try a different spelling or pick a bigger town nearby - all our support is delivered online, so your choice has no effect on the service.