Given that Gartner is expecting businesses to be capturing 85% of customer contact using artificial intelligence (AI) by 2020, it is reasonable to think that we should be clicking on that chatbot with a bit more confidence by now. Is that the case for you?
When I open a conversation with a chatbot it’s with a slight feeling of trepidation? How quickly will the chatbot understand what I need, will conversation be meaningful and relevant, and – if we are still talking by this point – how well will it resolve my query?
According to Renat Zubairov, CEO and co-founder of elastic.io, a company that supports organisations of all sizes in their digital strategy initiatives by helping them spend less time on integrating and monitoring various data sources across the business, and more time on using this data, the challenges that companies face in deploying effective AI solutions tend to follow common themes. In fact, of the three biggest issues companies experience with these applications, Zubairov says three of them relate to one key ingredient – data.
Context through learning
Ideally, companies want the chatbot to answer as many questions and handle as many scenarios as quickly and accurately – and meaningfully – as possible from the day they go live. And therein lies the first challenge, because AI is a learning technology not a plug-and-play solution.
A chatbot or AI can only be as good as the data it is fed from the outset, and the better they are integrated with other applications across the business, the more prepared they can be when they meet a customer and the more effective and intelligent their actions.
Humans don’t interact within set parameters; we tend to have several words with similar meaning, and pepper sentences with colloquialisms and expressions. As such, every query is different, even when the answer they seek may be the same. Chatbots need to have a solid background of context on which to draw in order to predict outcomes, learn formula and extrapolate algorithms for the most common queries they handle.
Using an integration platform as a service (iPaaS) enables developers or organisations to link AI programmes to data sources across the enterprise. Working as part of the overall system structure, instead of in isolation, provides the bot with access to contextual data that creates stronger algorithms and establishes known pathways it can follow for query resolution.
Outside the test environment
As already noted, chatbots are expected to be ready, willing and able from the day they go live, which puts enormous pressure on the testing phase. Integrating with several data sources during the learning phase can be costly – though critical – and tends not to be the area on which developers and companies want to be spending their money.
Ideally testing should be on the same live data that the chatbot is expected to run on, but in isolation from the day to day business so that customers aren’t exposed to the learning phase.
Achieving viability
It has been estimated that one in seven conversations with a chatbot ends up being routed to a person and one in eight is abandoned altogether. Is AI technology viable in the commercial world if organisations are risking the delivery of their frontline customer experience?
For a meaningful interaction between human and machine the chatbot must first be enriched with lots of clean, relevant data. Without this data input, AI systems will lack the training to expand and extend the models on which they operate and the ability to deliver meaningful results that satisfy users.
When referencing multiple data sources, errors can easily throw the chatbot off course and enterprise IT structures are notorious for carrying discrepancies between related fields in different software. If the online sales interface receives an update, such as a new delivery address or order amendment, is it communicating with the logistics software or inventory platform to feed the right information to a chatbot if the customer has a query?
In most instances, probably not. But creating integration across the full range of the enterprise architecture can easily be achieved by deploying an iPaaS solution that links common fields across multiple applications.