on 03-10-23 12:01 PM - edited on 12-04-24 01:20 PM by Michael_S
Each month a municipality receives about 8000 incidents about parking offenses, with photos and additional information. This is stored in a web application where the manual processing was done by a team of clerks. This was very repetitive and mundane activity which was ideal for robotization.
Discovery
First we needed to understand the process and the steps which were performed. It was clear that there is a judgment needed as clerks evaluated the photos to decided whether the fine should be issued or not.
Design
The repetitive part was pretty simple as we worked in web application, however, the judgment part was a challenge. We searched for a partner who could help us with the photos assessment.
Build
We automated the part of activities performed in the web application and we agreed with our partner on integration with the AI model for photo assessment through REST API. The model was running in docker. The bots downloaded the pictures from the web application and passed it to the AI model for assessment.
QA
We tested the integration and it worked quickly and reliably. The reported incidents have been evaluated and if the case reached certain percentage of confidence it was processed automatically without human touch. In case the confidence level was not sufficient it went to the team of clerks for manual processing.
Deployment
Deployment to production was simple. We created a robotic account in the web application and it became a part of the team for processing the incidents. In the web application, the robotic account can transfer cases to other clerks so it is now valuable team member.
As usual, people are afraid of what they are going to do in case robot does all the work for them. The reality is, there are so many additional activities which they need to do so, we highlighted to explain that not all cases will be processed by this solution and that there is plenty of other work which needs to be done by them. This enabled the team to do human-centric tasks, like meeting with people who appeal their offenses.
The impact was pretty big. Our team can now talk about first digitalization/automatization projects at the municipality levels, people saw that robots did not take their jobs, and instead they have been released from the repetitive work. This is being now shared among other municipalities.
The only recommendation is to connect RPA with some additional tools as it will have much bigger impact than RPA itself.
------------------------------
Zdeněk Kabátek
Head of Professional Services
NEOOPS
http://www.neoops.com/
Europe/Prague
------------------------------
I absolutely love this one @Zdeněk Kabátek 💙
Did you have to do much training of the AI model? Were you involved in that step?