When it comes to the introduction or evolution of bot solutions, many customer service units are looking to the stars – at least if you can and want to afford it financially. Also, both technology vendors and integrators deliver a potpourri of arguments and important features for their supposed new bot solution without which you as a customer service unit miss the supposed future and connection in the topic of AI – gladly garnished with great case studies full of impressive results.
And yes, it is of course sensible to orient oneself to the top, but it can also make sense to subject the announced solution to a fact check and to re-evaluate it from an economic pragmatism.
First of all, not all bots are the same, and here we are not talking about different touchpoints or text- or voice-based bots, but rather about the different development levels of bots:
Level 1 – Basic Bots (BAB)
These are simple bot solutions with question-answer functionality, which provide defined standard answers to certain keywords. Sometimes even the questions are predefined based on event trees and no individual questions can be asked. There are no interfaces to backend or further frontend applications or touchpoints and there is no „learning“ of the bot based on the conducted dialogs.
Where can such a bot make sense? Ideally with very standardized customer concerns such as FAQ or selective topics (e.g. temporary technical restrictions or delivery delays). The advantage of such a solution is the simple and very cost-effective implementation and maintenance, which quickly generates positive ROI in relation to CEX and possible cost savings through contact avoidance.
Level 2 – Virtual Assistants (VAS)
This level of bot is probably the most familiar to all of us, because we probably also use them unconsciously on a daily basis in the form of ALEXA, SIRI, GOOGLE or other providers. These systems already have connections to backend or frontend systems as well as speech recognition or the possibility to conduct individual question dialogs for customers, whereby these must remain very simply structured. An independent „learning“ of the machine can, but does not necessarily have to be given. Likewise, there is usually no transfer of dialog data to a further bot or for further processing by an agent.
Where can such a bot make sense? Since back-end and front-end information is already available here, general and simple personalized use cases can be implemented well and show good results. The important thing here is that identification/authentication functionality is available for these operations, which complies with current data protection guidelines. The personnel effort for the maintenance and adaptation of the VAS increases, so that a small team, which also acts as an interface to the affected applications and APIs, is necessary here.
Level 3 – Virtual Agents (VAG)
From this level on, already highly developed speech recognition and dialog software in interaction with Natural Language Processing (NLP) are used, also for complex personalized use cases. At the same time, such solutions require their own dialog, information and database architecture, which documents not only the customer data itself, but also all dialogs along with classification, interpretation and results. These are used by VAG for independent learning and prediction models. The interaction with RPA hyperautomation of processes or the direct intervention and processing in the backend data/systems are also given here.
Where can such a bot make sense? Here we are entering a level where, theoretically, the largest proportion of customer concerns can already be processed completely automatically. The key question is then whether it is desirable for other reasons (customer loyalty, possible sales approach with higher conversions, etc.) to also process these contacts automatically or whether it is better to have them processed by a human.
Since we are talking about far-reaching functionality and overlaps here, a dedicated team is needed that, in addition to use case development and technical and content support for the functionalities, also keeps an eye on adjustments in the area of customer contact strategy and company processes and can provide insights from the bot systems/dialogs.
Level 4 – Digital Humans (DIH)
Now we come to the extreme end of today’s available bot technology and into areas that would have been dismissed as science fiction just a few years ago. Bots at this level are able to process highly complex solutions on their own. They continuously learn and develop themselves, deliver predictions including analyses/interpretations without human intervention and also correct them.
In addition to the touchpoints used in the majority of cases, further touchpoints such as those in Metaverse or other visual interaction platforms are provided in order to get as close as possible to the „real“ human being.
They are able to read emotions on the basis of optical, sensory or linguistic analyses of the contact person and also to mirror them, i.e. to show emotions themselves and to react to them. Currently, this is still at a very rudimentary level, but who would have thought ten to 15 years ago that it would have been possible to query orders, transfers, bookings or statuses via a bot using speech alone, that I would be able to unlock my smartphone using biometric data, or that a bot would be able to detect cancer much more accurately than a doctor on the basis of an X-ray. And the development remains extremely fast.
Conclusion
And even if these technologies either seem appealing or possibly cause discomfort, the question remains: Which bot solution or which level makes sense in my specific case?
As always, there is no general solution; customer service organizations and the products and services they support are too individual for that. However, this applies to all approaches and considerations:
So it doesn’t always have to be “the best or nothing“ for a bot implementation or evolution in customer service. Sometimes the „better than nothing“ approach is the more sensible one and still offers the possibility to build on this.
Carlos Carvalho – Senior Consultant
junokai