Skip to main content

The rise of the ‘Conversational Architect’

(Image credit: Image Credit: Enzozo / Shutterstock)

Last year AI got serious. The increased understanding from businesses on the efficiencies of AI, as well as the convergence of AI with computer power, Cloud, mobile, big data and Natural Language Processing (NLP) mean that the technology has now become a commercially viable business tool.   

Looking at the current landscape, it’s clear that businesses of all shapes, sizes and sectors are using AI to disrupt traditional models. Lemonade, for example, is using chatbots to help their customers resolve insurance policy claims without the need of going through a call centre. Similarly, O2 recently announced at Mobile World Congress this year that they will introduce an AI system that will answer customer questions and allow account changes to be made without the need for human interaction.    

These developments will almost certainly trigger a huge amount of investment by rival companies in the same space. But they will also ignite a considerable change in the IT career and recruitment landscape as businesses rush to find the right talent to drive projects forward. In fact, according to Gartner, by 2019 more than 10% of IT hires in customer service will mostly write scripts for chatbot or digital assistant interactions. That means that one in ten people employed in customer service will have a script writing background.    

But how can businesses build chatbots that effectively address a range of complex and common business issues?   

Organisations must firstly invest in those training their AI machines. IT Technicians may be able to build great services like help desk chatbots and machines that excel at poker, but they won’t build fluid, conversational, digital assistants.   

Businesses must team them up with ‘Conversation Architects’; people with different backgrounds like psychologists, user experience (UX) consultants and script writers. Individuals in this role (which doesn’t exist today) will possess a mixture of relevant skills: UX research, language analysis, psychological understanding, creative ideation and script writing. Integrating these individuals into the IT department will help to build digital assistants with human-centred services suitable for today’s customer focused companies. 

Once you have your ‘Conversational Architects’ on board, there are various things for the IT department to bear in mind. First and foremost, building the chatbot doesn’t have to start from scratch. The brain of an AI assistant is a complex set of algorithms, a NLP engine and vast amounts of data powered by machine learning. But you don’t have to build your own machines and algorithms.   

The tech giants, and some start-ups, already offer services in the Cloud that organisations can use to develop digital assistants. Google recently acquired, Amazon Web Services (AWS) has Lex, which uses the same deep learning as Alexa, Microsoft is developing an open source bot framework, and of course, IBM has the infamous Watson. These conversational services easily plug into mobile apps, Skype, iMessage, Facebook and literally every other type of modern technology – meaning the IT guys can have an AI service up and running within a couple of weeks. 

You must then teach it how to chat, line by line. A conversational-based assistant is only as good as the inputs and outputs it has been taught. The machine needs to learn context, as well as different patterns of conversation – everything from speaking with teenagers to adults and all the subtly that this brings.   

For example, if I asked a bot, “what’s the weather like?”, the computer could respond with “it’s raining all afternoon, but dry right now.” But imagine that when I asked about the weather it was straight after I had asked “what are the best places to stay in San Diego?” A human would assume the person meant the weather in San Diego, maybe even in the school holidays, because they know they have kids. Teaching the computer to work like that is complex. It involves mapping out possible conversation flows and various options for each interaction. This takes time and specialists.   

The entire customer journey should be considered and responses created for every specific, service based question. Appropriation to each customer type is also critical as teenagers may have a whole different conversation with a computer than an older person. However, a survey by Mindshare found that 60% of people ‘would find it patronising if a chatbot asked about how my day was going’. A balance must be therefore struck to ensure the bot does not come across as overly creepy. The chatbot’s tone, grammar and use of slang must be carefully considered, requiring skills that are beyond a typical IT technician.   

There is another, more automated route to teaching a digital assistant; through reinforcement training. These sophisticated algorithms crunch huge volumes of conversational data and learn how to chat through trial and error. Like Pavlov’s dogs, positive outcomes become reinforced and failed interactions get dropped. The computer then quickly figures out what works and clarifies its correct response to a given situation. 

This approach does not come without risk as organisations often don’t have control over how and what the computer learns. Twitter users notably managed to make Microsoft’s digital assistant, Tay, a racist and a misogynist in about 12 hours, causing widespread embarrassment for those who created the bot.   

Once an organisation has built their bot, it needs to continually test prototypes in low-stake environments to ensure that the digital assistant is learning correctly. If the assistant is running on an external system, organisations should avoid being caught out by unanticipated software updates which may impact the bot’s behaviour. AI is still a young, immature field and releasing a guinea pig on customers, suppliers and vendors may have disastrous consequences for a brand. 

Ultimately AI has tremendous potential, even if its vision does not match up with reality in certain areas. Instead of watching from the side lines to see how things settle, business should seriously consider AI within the context of their current processes. If they are to be truly successful, businesses will embrace these innovations from the ground up and adjust their recruitment process to hire individuals who possess skills currently beyond that of the IT department. If 2016 was the year AI got serious, then maybe 2017 will be the year of the Conversational Architect.     

Simon Sear, founder of SPARCK – the strategic design consultancy of BJSS 

Image Credit: Enzozo / Shutterstock

Simon Sear
Simon Sear is the Founder of strategic design consultancy SPARCK. Born from IT consultancy BJSS, SPARCK helps organisations to design, prototype and plan new services, products, and ways of working.