“I’m sorry. I didn’t understand that. Please try telling me again what you’d like to do…”
The chances are that phrase will seem familiar if you’ve ever tried getting something done by a ‘Chatbot’. Usually, it’s not long before the Bot has given up and put you through to a person.
So why don’t call centres stick to the traditional method of ‘key 1 for customer service, 2 for sales…’? It may well be due to the misguided premature adoption of (as yet) inadequate technology for the sake of appearing progressive.
Ineffectual technology is worse than none at all. It merely gives a bad name to the company using the automation, while raising the customer’s blood pressure. It also damages the trustworthiness of such technology generally.
This happens because most contemporary Bots don’t use true Artificial Intelligence, nor genuine computational linguistics. They are pre-programmed to use ‘decision trees’, which automatically react to keywords. For example, when calling or messaging a printing company to find out why your business cards haven’t been delivered, you might constantly find a ChatBot or WebBot routing you to billing or accounts, because it interprets the word ‘card’ as a credit card issue.
Despite such difficulties, consumers are forced to put up with the situation. According to research by Chatbotslife.com, back in 2019, there were a reported 265 Billion customer requests in USA, and businesses spent nearly $1.3 trillion servicing them. The findings state that using Chatbots has helped businesses to save about 30% of that expense in the last two years. But as to how many telephones have been thrown out of windows by frustrated and potentially lost customers, who knows!?
When using Bots, to attain true satisfaction for customers, to reduce costs and increase revenue, a leap forward in the underpinning technology is imperative. Consider that businessinsider.com stated recently that consumer retail spend via chatbots worldwide will reach $142 billion by 2024; increasing FIFTY-FOLD from just $2.8 billion in 2019. What’s more, as far back as 2016, according to a survey by Venturebeat.com, 50% of consumers stated that they expected any business to be ‘open’ to deal with enquiries 24/7/365.
Fountech.Solutions are at the forefront of this drive for improvement. We specialize in Advanced Conversational AI as opposed to unproductive and inefficient keyword detection.
Our recent ground-breaking work in EdTech, with Soffos.ai (a corporate conversational knowledge transfer platform) and other deep-tech startups, especially in the healthcare sector, has produced results even beyond our own expectations; not least those of our customers. Not only do our algorithms use long-term memory from previous conversations to understand the true meaning of a person’s intent, we also use such techniques as emotional and sentiment analysis. Our AI can detect anger, it can understand written and spoken acronyms, slang and dialect. It can output in many languages.
To do this our AI interrogates Knowledge Graphs (KG’s),which are far more advanced than simple databases. Our KG engineer, Zahra Hossaini Pozveh (PhD) said:
“I think conversational AI is one of the most important technologies being developed at the moment… However, the capabilities of current Natural Language Processing systems are limited. Speech is merely converted to numbers and machines are instructed to interpret those representations in specific ways… Instead, what differentiates us? We understand that there is a missing underlying factor which is crucial to taking AI conversational systems to the next level, and that's the ability to reason. In other words, enabling the machine to mimic thought processes…”
Following Zahra’s system, if a banking customer were to state: “I lost my credit card ”, our AI could suggest a hierarchy of intent. First: “Do you want to deactivate the missing card?”, next- “Would you like a new card issued?” Then -“Are there any transactions that you might wish to dispute?” But if responses aren’t as expected, it can ‘cope’ and investigate further.
70 years ago, leading British computer scientist Alan Turing designed a famous test, intended to assess if a human could not distinguish a computer from another human in a text-only conversation. In seven decades that feat has never yet been satisfactorily achieved.
But we’re working on it. Watch this space…