Re-framing our thinking around conversational AI
It's seven years since Amazon launched its Echo smart speaker, introducing us all to Alexa and taking the virtual assistant mainstream.
Since then, Amazon and Google have shipped over a hundred million speakers, and it’s estimated that there are now over 100,000 ‘skills’ available for Amazon’s Echo alone.
They represented the first idea of a computer whose entire interface was based on voice. Up until this point, our primary interactions with computers were controlled by keyboards, mice and eventually touch – with everything fed back to us through graphical user interfaces (GUIs).
They are clever pieces of technology – but the most basic form of computer; a voice browser with microphone and speakers with all intelligence behind it based on the cloud. Although basic, it was an ingenious way to capture data to train the machine learning algorithms to improve the performance of the speech recognition.
On launch, the ambitions of both Amazon and Google’s devices were absolutely huge. Initial visions were of Star Trek-style ‘all seeing, all knowing’ computers. Our very own Personal Assistant’s helping to organise our lives, controlling our homes and giving us access to whatever we wanted.
But in reality, they failed to reach their full potential – instead being used today to predominantly play music, check the weather, or turn on the odd compatible lightbulb...
So, what’s stopping us from using them to engage more widely? And does our experience of home devices impact how brands address the conversational AI opportunity?
The challenge Amazon and Google had was that they underestimated and understated the complex relationships and domain expertise that was required when interacting with brands; brands that were concerned in the likes of utilities, insurance, travel or any other services that the average person/household needs.
It was just too complicated – it’s a bit like calling my bank to book a holiday. The poor person on the other end of the phone wouldn’t have the knowledge or tools to fulfil my request. The idea that a universal voice assistant could run our homes and our lives was just too ambitious...
Instead of complexity, we want our brand engagement to be quick, easy to understand and simple to conduct. If we’re dealing with an insurance firm, for example, we expect any conversational AI solution to understand what it is that we’re trying to achieve and offer a self-service option when it’s the right thing to do.
Any solution should be able to capture why I’m calling, work out whether my request can be addressed through automation, or connect me with an advisor when the task needs a human touch. We engage with a bank on financial matters, or when we have issues with our power or water we turn to a utility. At no point would we expect an insurance company or a telco company to help us sort out our holidays...
Technology is no longer a barrier for conversational AI
The good news is that technology no longer needs to be a barrier when it comes to deploying conversational AI. Speech recognition keeps getting stronger and stronger, indeed we’re now at the stage where we can synthesise speech to sound almost indistinguishable from humans. And with AI processing power now doubling every ten weeks, the computing power that’s assigned to training neural nets and AI engines is becoming more and more accessible.
This is driving both performance increases and cost reductions, enabling organisations to broaden out their speech AI capabilities and making it possible for CX teams to capture conversations from voice, video and text – wherever they take place in the customer journey. Capturing all these conversations digitally allows brands to unlock new insights and extract value from the data, particularly in the customer service world where conversational AIs can now be trained to become real experts in their specific fields or sectors.
Building conversational AI with deep, sector-specific context
Instead of attempting to be a universal assistant, the goal for conversational AI solutions across customer journeys is for them to become real experts in their own field. Effective use of intent capture and analysis techniques will give your AI precise insight into just why your customers are getting in touch. Speech AI solutions can then be trained in detail, with further content and expertise added as new customer issues and topics are raised.
This will see conversational AI move beyond the first wave of AI-powered voice and chatbot solutions. These tended to replace somewhat clunky IVR systems, and have generally been highly successful with many succeeding in automating between 30-40% of interactions. However, these solutions have often been standalone leading to silo-ed customer data that has been hard to integrate with other parts of the customer journey.
Moving towards a second wave of conversational AI
We expect the second wave of speech AI to be much more far-reaching, embedding natural language understanding and AI and automation capabilities across a broader range of applications. Conversational AI and voice recognition will increasingly be used to support CRM and mobile apps, as well as for contact centre advisor support where the AI can listen to conversations, advise on compliance, recommend relevant knowledge articles and give advisors help where it’s needed.
This kind of real-time guidance, backed up by powerful analytics tools and capabilities such as sentiment analysis will help conversational AI deliver consistent, high-quality experiences across extended customer journeys.
To learn more about conversational AI and how you can transform your customer journeys with Automation and AI, download our AI & Automation ebook.
Stuart Dorman is Chief Innovation Officer at customer contact technology specialist Sabio. Stuart is a recognised thought leader in the contact centre industry, regularly producing thought provoking white papers, speaking at industry events and judging top industry awards in Europe.