The Value of Transactional Robots

For a machine to maintain an open conversation with humans, artificial intelligence (AI) technology is required.

For a machine to assess whether cancer is detected in a radiological image, artificial intelligence (AI) technology is also required.

For a machine to detect whether there are delays in the execution of a production order, be it for a car, a bed sheet or a building, or even to check whether all the clauses of a contract are safeguarded, is not. It is even counterproductive.

In the first two cases, there is the possibility of closing the conversation or having the analysis performed by humans. We all know cases of companies that have had to compensate customers for incorrect information provided by chatbots or for stupid, foolish or even offensive responses. AI-based robots have an open and unpredictable operating component, both due to the complexity of the machine and the nature of the programmatic construction that is at its core.

To be specific, my fuel card is from Prio and not Galp because, to begin with, Galp's chatbot could not understand my "yes". At one point in the conversation, it asked me to confirm by saying "yes", I said "yes" and it did not understand, he only understood a "yes" when said by a woman. And I decided not to put up with that and opted for Prio, where everyone understood me and understood what I wanted.

Now, in a transactional robot, like those mentioned above, the aim is and is to avoid the risks mentioned in conversion robots, this is because in transactional environments the options are much more contextualized and limited, and there is no need to invoke a technology from another level with risks specific to that level, such as those mentioned. Nor are there the costs that exist in AI-level technology.

Therefore, it is like using a truck to deliver a letter with the risk of breaking all the flowerpots on the street because the street is too narrow for the truck.

And I believe that this part has not yet been understood by the management of many good companies, they can and should still benefit greatly from the world of classic algorithms to program robots, starting with the certainty of behaviors and results and continuing with the costs.

After all, why use an open solution in a closed transactional world? The issue of incompatibility is immediately apparent in this formulation of the question.

If you have any questions about how to achieve this, please contact us.