To state that chatbots are here to stay is not breaking news. As consumers, we are becoming more and more used to interacting with these virtual assistants in our daily lives, especially when browsing the web. As companies, we know that chatbots are an essential piece on the path to digital transformation, which involves omnichannel and technological innovation. Nor is it news that implementing a chatbot is an investment of time, resources and of course money. The question “What is the mission of the chatbot?”, is the key to getting the project approach right and it must be answered jointly by the different departments that drive the initiative (Customer Service, Marketing, IT, Innovation, etc.). But the other and biggest question is: how to select the best chatbot provider?
The best chatbot is the one you get working (and not die trying).
All technology has its architecture, its functional characteristics, its roadmap and its restrictions, but these parameters are mere circumstances when choosing a provider. A technological solution is chosen by testing it, making sure that its capabilities allow you to achieve your expected objective and verifying that its limitations do not affect the development of the project. If the validation of a technology provider goes through testing, often in the context of a pilot or a proof of concept (PoC), it is worth it only if the provider allows to quickly, easily and representatively implement its technology.
STEP 1: Ask for help!
If you like to “mess around” with new technologies, you’ll probably be used to Open Source libraries and the DIY philosophy. These approaches are interesting, since they allow you to discover the latest trends but may be inefficient to carry out a specific project with a defined scope and deadlines. If instead what you need is to establish the specific requirements to start building your chatbot, then the best strategy is to contact the provider directly. In addition to giving you a technological platform, the provider must put at your disposal his practical experience in the sector, his know-how, so that you can take advantage of the technology and perfect your chatbot. Your provider should accompany you during the construction of the chatbot and will help you both to use time and resources optimally and to manage the expectations of all the stakeholders.
STEP 2: Ask for a test environment!
In such a diversified and dynamic market, buying software from a PowerPoint or a promotional video is totally anachronistic. Technology has to be tested, touched, manipulated and evaluated. The success story of another company does not guarantee success in your company. Your chatbot has to adapt to the use case that it intends to solve as well as its audience, and the vertical in question. The only way to check if the project will work is therefore to start building the virtual assistant. First of all, it is essential to impart knowledge to the machine to check its level of understanding (either by NLP, or by Machine Learning). This knowledge base almost always already exists: it can be the FAQs of the website, the scripts of the Contact Centre, etc. The first step is thus to have a brand-new environment where you can upload the answers that the chatbot will have to give. Next, it is up to test its performance, possibly with the help of some kind of debug tool, and carry out minimal adjustments to refine the AI engine. It should be noted that in the context of a proof of concept, it is not feasible to have to “train” the chatbot with “variations” or “utterances”, that is, thousands of ways to ask the same question. Devoting so much time and resources to training would make the pilot completely unprofitable, so any technology that requires this type of training is incompatible with a PoC.
STEP 3: Ask for feedback!
So that the proof of concept of the chatbot does not remain in a purely functional and technological stage, its necessary to involve more trusted users and give them access to a chatbot, as similar as possible to the final intended product. To do this, it will be very helpful if the chatbot platform includes an SDK that allows you to deploy a customised UI, in an agile way and without the need to code. You can share the chatbot in a ‘family & friends’ environment, clarifying the scope of the pilot to handle expectations, and collect the impressions, suggestions and criticisms that the virtual assistant raises. Both this external feedback and the direct experience of working with the chatbot platform will allow you to see the viability of the project, the suitability of the provider, to ultimately clarify whether you made the right choice when selecting your chatbot provider.
Selecting the best chatbot is a challenge with as many potential outcomes as unknowns. The assessment of the technology provider must be agile and easily adaptable to the requirements of the company and the use case. The investment in time and resources must be moderate and at the same time allow a general idea of each of the essential aspects. Last but not least, besides the capabilities demonstrated during the proof of concept, the potential for evolution and scalability of the technology platform will give you the key to determine which is the best chatbot for you.