The Rise of Truly Intelligent Chatbots


If you think you know chatbots - or virtual assistants to give them a term that covers both voice and text interaction - think again.


They’re developing faster that many thought possible. Artificial intelligence is making them clever at a speed worthy of one of those rapidly maturing aliens in any number of sci-fi movies. 

Not so long ago they were basic, fairly dumb automated text-based systems that offered a limited set of answers to a limited number of questions. And often those answers were wrong, or it didn’t understand the question at all.

You might argue Amazon’s Alexa is still a bit like that - “Sorry, I don’t know the answer to that one” - but the most advanced conversational AI is now a self-learning context-aware helper that has a much better understanding of natural language.

Technology giants such as IBM, Amazon, Google, Microsoft and Oracle, as well as more specialised companies such as Nuance and Rulai, have taken chatbots by the scruff of the neck.

They are proving particularly useful for weeding out those “high volume, low value” customer interactions – the most common, simple questions that are easily answered. Questions to a bank or telecoms company - “What’s my balance?” or “How much data do I have left?” – can be answered in seconds.

IBM’s Watson Assistant, for example, which powers Brazilian bank Bradesco’s chatbot, is helping the bank’s 60,000 employees answer about 30,000 customer queries a day. As a result, the bank has been able to switch 94% of its customer queries to virtual agents. Customer satisfaction levels have risen, too, the bank says.

Nuance’s Nina chatbot has also been helping companies reduce call centre costs in the retail, financial services, telecoms and government sectors. Clients include Coca-Cola and the Australian Tax Office. 

Chatbot systems have had to learn all the different ways people ask the same question, not to mention the differences in dialect. The growing mountains of data machine learning can now be applied to has meant accuracy rates for answers to customer queries have risen.

Developers have been working overtime, creating chatbots for a range of uses, not just the standard call centre customer queries, but guiding customers through expenses reports and tax forms, making product recommendations, or helping them plan journeys. Connected cars – and eventually autonomous vehicles – will undoubtedly feature voice-controlled chatbots in future.

Research by CX Company suggests chatbots can already handle 51% of customer calls on average. For some businesses, that percentage can be much higher. And more than 90% of businesses surveyed said chatbot technology would “add value on their current customer service model”.


The term conversational commerce, coined by Chris Messina in 2015, is now common parlance as brands embrace the chance to interact with customers at any point through the purchase process.

Websites may receive a lot of traffic but less than 5% of them result in purchases on e-commerce websites, according to Monetate, the personalisation marketing company. Despite Amazon’s dominance of the online space with 44% of e-commerce sales in the US, it still accounts for just 4% of retail sales overall.

One reason is that websites lack the personal touch – human help when you need it. So argues chatbot provider LivePerson at least. It reckons some of its customers have enjoyed a quadrupling in sales conversions after introducing conversational commerce technology.

A digital sales assistant that talk almost like a human could be the secret to boosting e-commerce’s share of the giant sales pie.


The advantages are well known. Software is cheaper than humans. It can handle multiple conversations at once. It never gets tired so can be available 24 hours a day. There’s no need for customers to wait in a queue listening to maddening musak.

And as young people prefer interacting via messaging apps rather than telephone these days, text-based chatbots are naturally more suited to this smartphone generation - 95% of the UK population now owns a smartphone.

They are also enabling customer service staff to add value to the advice and help they give, acting as assistants, not replacements. As the most basic queries can be automated, this is freeing up staff to answer more complex questions and move into more of a sales role, proponents believe. 

Chatbots are being developed that can supposedly interpret the emotional state of the caller and suggesting an appropriate tactic for the human sales assistant to follow.

But there’s little doubt that reducing call centre costs is a key driver in the adoption of these early CRM chatbots.

In the next feature we’ll look at some more specialist roles for chatbots, particularly in healthcare.

Megan Dickie