Digital assistants. Businesses are increasingly using them. Their customers, often reluctantly, tolerate them! In a world where human customer service is gradually being eroded by inevitable advances in technology, Durtti wants Humley Chief Product Officer Ed Smith to explain what the current limitations are with existing digital assistants, how that can affect the overall customer experience, and most importantly, how Humley have pioneered a truly next generation AI powered solution.
What is Humley, Ed?
We’re a UK-based company that is using natural language engagement to create expert digital assistants for brands and organisations. Using the power of AI, we’re providing companies with assistants that fulfil a number of roles. This ranges from improving customer engagement – creating customer service “bots” for household name brands – to assistants that automate and smooth internal process, or even help sales efforts. Ultimately, we see ourselves as leading the charge in solving some inherent issues and challenges across industries by deploying machine-to-human conversational technology.
What’s different about us is that we’re taking AI – which many companies are talking about but few are utilising to its fullest extent – and making it practical, deployable, manageable. Essentially, we’re making it work in real life for businesses.
What exactly is Natural Language Engagement?
Natural language engagement is an AI’s capability to interact with language. It’s essentially the technical term for the power of listening, understanding and responding accurately.
Language is a complicated thing. Without even realising it, we use hundreds of variations of words and phrases to ask even the simplest questions – it is a major challenge for a machine. Natural language engagement gives customers the ability to ‘in their own words’ easily converse with the brand on any given subject whenever, wherever, however it suits them – giving businesses the power to actively encourage customer and brand interaction.
How does Humley accurately measure the performance of a virtual assistant?
There are many ways to measure the performance, but one of the key ones from Humley’s perspective is the assistant’s ability to provide a useful outcome at the first time of asking.
At the moment, it’s common practice in other services to have mixed approach of automated response scripts and triaging to human operators when the conversation travels off script. Humley is different. Through Humley’s automated training technology, unscripted conversation tooling, and ability to capture long-tail enquiries, we aim to limit the number of times an assistant has to handover to a human. This provides an enhanced customer experience by answering more questions first-time round, putting less reliance on expensive human intervention, and keeps customer queries contained within an expert channel. Humley clients have seen incredible cost savings and efficiency benefits from this approach already.
What’s the difference between unscripted virtual assistants and the traditional approach?
A scripted approach is exactly what it sounds like – the assistant is programmed with a range of possible questions and answers that attempts to predict the possible interaction flows a user may have. This approach works fine if the user asks a precise question, or the bot has a simple narrow purpose. However, once the communication veers off the script the user is either pigeonholed into a best guess or gets the dreaded “I do not understand” response – leading to miscommunication or frustration.
Humley’s unscripted approach focuses on what the user really wants from a digital assistant: an accurate and helpful response. Instead of a script, Humley has long tail data processing capabilities that work by using real-life, crowdsourced data. This means that even if the conversation has obscure subject matter, Humley’s AI and natural language engagement capabilities and unscripted conversation tools will still be able to understand what the customer is trying to ask and will provide an accurate answer, maintaining conversation as the user navigates themselves through knowledge and processes.
How does Humley train a virtual assistant? Where does it get the necessary knowledge base from?
Humley uses automated training. In contrast to other systems – that need humans to manually label data or input all possible questions in all possible formats – Humley assistants use our own machine learning processes to create ontologies that then train the various natural language processing technologies.
Humley’s platform can be plugged into or ingest from a company’s existing knowledge sources, utilising its information to act as an intelligent, bespoke, brand-focused virtual assistant, helping to answer questions phrased in the user’s own words and the user’s own language.
Can a business start small with a Humley solution and can the platform then grow organically in parallel to the growth of the business using it?
Yes, that’s the beauty of the unscripted approach. Instead of crafting thousands of individual scripts and building brand and sector training materials by hand – which is a process that can be cost prohibitive – Humley can bulk ingest knowledge and script a few narrow processes that suit a businesses’ needs. Furthermore, using our training processes and conversational modules systems, Humley can be launched very quickly. Our approach is different, and our technology is designed in a way that assistant solutions can be easily deployed to gather quick insights into customer acceptance and track business impact before full roll out of services.
Effective customer interaction is nearly always a very complex process. What level of complexity is Humley able to deal with for brands? Can you give an example of a successful solution you have provided?
Because of our unscripted approach, Humley is able to deal with more complexity than the average assistant, as it can pull information from data. This also means it isn’t industry prohibitive – our customers vary from high street banks, digital agencies, system Integrators, telcos, consumer electronics manufacturers and event organisers.
A good example is the mobile assistant we designed for the CAT S60 to provide customer support. Through Humley’s training, the accuracy of responses went from 60 percent to 90 percent in under three weeks, with the solution being delivered in four languages. Not only did this improve the customer experience, but the insights generated by the queries could be analysed by the team, and has helped to identify customer issues and address them in real time.
What characteristics of natural language engagement services are likely to evolve further to the benefit of both the (brand) user and the customer in the next 3 years?
The reality is that most virtual assistants in the market are quite rudimentary and, as a result, the use cases have remained limited. As natural language capabilities improve with greater access to data and unscripted assistants become more widely adopted, we will start to see some truly innovative artificial intelligence in the market which has the ability to transform a great number of industries, services, and our personal lives.
We are seeing the underlying natural language processing technologies evolve as we move from machine learning strategies to neural network strategies and Humley is designed to plug into these third party technology providers to always maintain a best of breed suite of technologies that we orchestrate to create effective conversations.
As we start to see effective and impactful deployments we will begin to see major disruption in many industries. Let’s just take the insurance industry as one example. Insurance is a complex purchase that is heavily regulated. Virtual assistants that can assist with customer education, knowledge discovery or provide a quote or claim could transform the whole industry, while standardising the consumer interaction to be both audible and compliant. In fact, the constant availability of a virtual assistant could change how insurance is provisioned altogether as, in theory, customers could update their insurance policies in real time – which means you could treat your insurance like a shield that you can configure for your changing circumstances on a weekly or daily basis.
Humley already has existing relationships or partnerships with IBM Watson amongst others. What does partnership with Humley entail and what sort of partnership opportunities are you keen to develop, moving forward?
Humley partners with technology providers and also service providers. As a tech-agnostic platform, our natural language engagement capabilities use a large number of AI technologies provided by the likes of Microsoft, IBM, Google, Amazon and more beyond. We are constantly re-evaluating the technology sets available to make sure our services are using the latest and best of breed technologies.
We also partner with many vendors developing assistants with customers, or developing technology suites, that will include natural language assistants. These partners bring industry expertise, customer relationships and the professional services resource to deliver to customers, and allows us to focus on product roadmap development.
Finally, Ed, how would you demonstrate the Humley platform to a class of primary school children so that they are truly inspired by its potential to make AI history?
The best way to demonstrate our technology is to let people have a go themselves. As an unscripted service it can handle the obscure, tangential and the unusual. The joy of being able to give a class of primary school children a chance to ask what they like of a nursery rhyme service, or fairy tail service would show that there is a bright future ahead of communicating freely and exercising the subtlety of their language skills when talking to a computer.
Whether you’re speaking to primary school children or a transformation team in a large enterprise, once they understand what the technology can do, the opportunities seem endless!
More at humley.com
Ed is a member of The Artificial Intelligence Group on LinkedIn.