(3 min read)
It’s around 8 AM as Thorsten steps into his car. He’s tired, but his mood is lifted by the fact that he doesn’t have to spend all day dealing with order requests and RFQs via email. He has Hackstrid, his virtual in-car assistant, to help him take care of that. On his morning commute, Thorsten is the human in the loop.
A Usual Morning with Thorsten and Hackstrid
Thorsten: Good morning
Hackstrid: Good morning, Thorsten. How are you doing today?
T: Great, thanks! What’s new in my inbox?
H: You had 30 new emails this morning. I have processed most, but there are still five emails which require your input. Should we take a look at the RFQs first?
T: Yes, let’s do that!
H: OK. Firstly, our customer Satair wants to order 100 x 500 ml units of propeller oil. What shall we do?
T: You can approve that order. Anything else?
H: Our customer AviationSynergy wants to order 4 x 700 ml units of propeller oil, 3x 500 ml units of steel coating, and one unit of absorbent cloth.
T: AviationSynergy? I don’t know that customer very well. Can you perform a quick credit check?
H: OK. Running online credit check… I see that the customer’s liquidity ratio is 52.57%, which is below our average customer’s rate. How do you want me to proceed?
T: Let’s approve it. But I would like to add a note to the reply.
H: No problem. What would you like to add?
T: That the payment should be cleared before we ship the products.
H: OK. I have approved the order. I will also include the following message in the reply: Clear payment before shipment. Shall we move on to the next RFQ?
And that is how Thorsten’s mornings begin!
What’s Under the Hood?
In the recent SAP iRPA Hackathon, we demonstrated positive and intelligent robotic process automation experiences. To understand this, you need to know what is going on under the hood: We have integrated our digital human platform, SAP Conversational AI, and seven iRPA bots to incorporate webpages, SAP S/4HANA, machine learning, and email.
The conversational AI allows Thorsten to use his voice to communicate in his natural language. The iRPA bots carry out the hard work upon Hackstrid’s command.
To recognize and categorize the requests in Thorsten’s inbox, we deploy state-of-the-art natural language processing techniques. This meant that we could also capture email semantics and entities. We used transfer learning from a two-million-word deep learning model with 685,000 unique vectors of 300 dimensions. From here, we retrained the last layer of the deep learning model to learn classification schemes for our particular problem.
Innovations like these help companies to improve the employee experience considerably, create cost savings in the long term, and alleviate repetitive manual tasks. An ever-increasing number of companies are interested in how intelligent robotic process automation simplifies workflows. Our success at the SAP iRPA Hackathon was evidence of this. itelligence picked up the 2nd prize at the event organized by SAP Coil EMEA North. Our colleagues in the Netherlands won 1st prize for a separate life science-based project.
Human in the Loop
Although this story is fictional, there is a strong case for intelligent robotic process automation. We have found the solution to the problem. Human-in-the-loop concepts are becoming more and more relevant and new smart technologies provide tools for completely new automation cases in a whole range of industrial and business environments.
We should continue enhancing the interface between humans and machine, like we did at the SAP iRPA Hackathon! And never forget: the most powerful and universal interface we have ever had is the human face.