The AI engine for fast user input
Watch the video to see how Ron (bottom left in the video preview) works without AI support, while Lea (playing laser tag) works with AI support. Find out more in the video.
Benefits
Efficient use of time
Less manual user input means working faster
Stress-free service
Less searching for data values makes the working day more relaxed for the user
Cost-effective solutions
Faster work results in lower costs
Why exactly this solution?
Time independent
the AI model does not need to be retrained over time
Independent of user changes
When a new user is added, the AI model does not have to be trained again
Independent of master data changes
the AI model does not have to be trained again when creating new master data
Motivation
In 2018, my friend and I visited a beach cafe in Heilbronn. A waitress there was particularly attentive and often knew what you wanted to order before you knew it yourself.
The next day, I was working on an SAP development project near the café. Unlike the attentive waitress, the SAP developer didn’t know the needs of the users. So I asked myself: is it possible to teach development this skill?
The first draft of a test application consisted of a drop-down field with a selection of drinks and an order button. Initially, the drop-down field sorted the drinks by order frequency. Later, we weighted orders closer to the present higher, which improved the sorting. Nevertheless, this was still not the essence of the nice waitress.
In 2019, I went to St. Gallen, where I worked on time recording. There were patterns here too:
- Holiday recorded yesterday => holiday again today
- Bosch customer order three Tuesdays in a row => Bosch customer order again next Tuesday
- Duration of 2 hours entered => weekly meeting
The relationships were more complex than in the beverage example. In 2021, I began to study AI in order to master these challenges. I came across many questions:
- How do you integrate different data types into the network?
- How does the model deal with changes to master data?
- Does the model need to be constantly retrained?
- Which architecture is the best?
After many attempts and hurdles, I finally found solutions. But one challenge remains: integrating the solution.
I need your help for this.