AI and IBM Watson – Are we nearly there?

Perhaps it’s stating the obvious, but the sheer number of applications to use IBM Watson technology is the main out-take from their World of Watson conference in Las Vegas last week. In the past five years IBM’s Artificial Intelligence fuelled Watson has become the topic that everyone in digital wants to talk about.

The big headlines around Watson will always be reserved for the ground-breaking results in healthcare. Such as, finding cancer-linked gene mutations (Watson found the same mutations as human experts with 99% accuracy, and then went on to find 30% more). But it’s the other use cases, both large and small, which are going to reshape our day-to-day lives.

I have a tendency towards both pragmatism and scepticism, especially when it comes to things that promise the world, as evidenced by my spending (and losing very quickly) $20 in the casino last week. So, it has come as some surprise that following IBM’s World of Watson conference, there is no doubt that cognitive technology will completely reshape our ways of interacting with businesses, and it’ll happen quicker than you think.

At the easier end of the AI spectrum you have Chatbots, or a “Digital IVR” as simplified by the Chief Digital Officer of RBS when he spoke about their Bot which is due to public release this December. Chatbots have been touted as one of fasted growing technologies that will soon be integral to our everyday lives. “By 2020, the average person will have more conversations with bots than with their spouse” according to Gartner.

At the more complex end of the AI spectrum you have the bringing together all your structured and unstructured data, to find new insights, solve problems and deliver intelligent personalised experiences that are simply not possible without the interpretation and learning that cognitive technology can provide.

An introduction to AI

The IBM Watson technology has been around since 2007 with the infamous Jeopardy story, where Watson beat two of the most successful competitors of the TV gameshow. 2007 was an incredible year for tech. It was the year of the iPhone, Facebook, Twitter, GitHub, Android, Airbnb, Google acquiring YouTube, and Watson. Whilst Watson approaches its 10th birthday, cognitive technology is still very much in it’s infancy, but it’s maturing rapidly.

IBM tout Watson as Augmented Intelligence rather than the Artificial Intelligence. To put it simply, it’s man with machine, not man vs. machine. So, there’s nothing to be scared of, unless of course you’re worried about a machine taking your job.

There’s a lot of vision and intent when it comes to businesses working with IBM Watson. You need to pick away relentlessly to find examples of those organisations who have fully deployed live solutions which are solving problems today.

What kinds of problems is AI good at solving?

Repetitive tasks. Every repetitive job can be done by a machine. Pilots, taxi drivers, software engineers. We still live in a world where most of our work is repetitive, so the potential is huge. When you introduce cognitive technology. Which can understand intent, reasoning and implications, it opens a whole world of tasks which have, until now, been the domain of humans. Now clearly some of these repetitive tasks are more repetitive then others. But if we’re already seeing AI writing better newspaper sports reports that journalists, then there probably isn’t a future of thousands of call-centre staff answering queries about parcel tracking.

The big challenge for cognitive technology is understanding the ‘signal’.

There’s a big difference between understanding that there’s a billionaire called Bruce Wayne who spends his free time flying around solving crimes, and knowing it’s Batman. If we can narrow down the breadth of challenge we’re using AI to solve, and clearly articulate what a tool is good at, a user will quickly learn the best ways to work with it. Think about the last time you were in a department store on a Saturday morning and got the trainee. He knows that if you utter the words “I have a complaint”, that he needs to escalate the situation to his manager sharpish. If we build these signals into the tools we create, we can quickly have tools which might not be great at everything, but can certainly be good at something. Then it’s easier to seamlessly integrate with the rest of your organisation where the requirement is above its current pay-grade.

The UX challenges

User Experience used to mean screen interactions. Now the number of Zero UI channels are growing fast (Echo, Siri, Cortana etc.) but these have always existed, take Call Centres for example; just because there isn’t an interface it doesn’t mean you can’t control the experience. It takes lessons from industries for whom it’s more familiar. Take Zappos for example, famed for delivering fantastic customer service, and wow in every experience. Last month, you may have heard of the 10 hour 43 minute customer service call by Zappos employee Steven Weinstein. Amazing customer service. By using a chatbot, that is available 24/7 and never gets tired, delivering a personalised customer experience could be much longer - an infinite length of time. Well, until the customer falls asleep.

When designing UX for AI, the rules are the same. Understand the problem, understand the users and determine the desired outcomes. Then learn, learn, learn. With cognitive you are defining new interaction patterns with consumers, so whilst the underlying tech is new, you’ll be surprised at the ability for your customers/employees to adapt. A great analogy learnt last week is The California Roll Rule. If the California Roll was your first introduction to sushi, you’re not alone, in fact it’s true for the majority of the western world. People don’t want something new, they want something familiar done differently. Perhaps it’s fair then to say the California Roll is a gateway drug…

Data is the backbone of AI

Data is the foundation for any modern digital solution. The challenge is connecting data to find new insights in an intelligent way.

Take the example of an airline. Your plane is delayed returning home, and you’re not going to make your connection. What if rather than offering you a hotel stay and apology, the airline could scour alternative routes and suggest if you don’t mind an hour’s drive, they can route you to the next nearest airport to your home. They can provide a hire car and even an evening meal from a top restaurant on route. It’s a nice day so they upgrade you a convertible. All the data is available to create this scenario. You just need to connect the dots, and the airline to take responsibility for getting you home, not just to the airport. Now, that’s service I wouldn’t mind paying for.

Top 6 learnings

Start by harvesting your questions

This could be from your existing website search queries (find anything which is a naturally worded question), on-site surveys, call centre recordings or live chat logs. Ultimately this will go full circle, as GSK demonstrated with their Theraflu advertising which has a Chatbot built into their display banners. This completely flips the business where previously their strategy informed their media, now their media is informing the strategy.

Laser focused on your use case

If you think by throwing everything at it, Watson can simply magic up answers, you’re sadly mistaken. RBS know this - 30% of their contact centre calls are transactional enquiries. Demand for their live chat outstrips demand by 250%. Start small, and find a simple challenge which the tech can achieve. Stuart Butterfield put it well saying “If your expectations are aligned to the current capabilities, you can get a great deal of value today”.

Collect and restructure your data now

The better your data, the better your outcomes. Even if having a cognitive solution is a while off, it’s worth taking the time and effort now to start auditing your data, both structured and unstructured. Some valuable data you might not have previously considered as useful due to the unstructured nature of it. Taking the time now in the business and technical decisions you make, will enable you to be Cognitive ready.

Make it simple

Think California Roll, and RBS providing a “Digital IVR”. The execs know what an IVR is – it halves the number of calls and associated costs. Address the anxieties – position the project as something they can relate to. It’s an awful lot easier selling in a project to a board who understand it and see the value it will deliver.

You need to diversify your teams and skills

Training an AI solution is different from programming. Some of the expertise is the same, some is quite different. One example of diversifying skills sets for chat bots development, is are looking at hiring screenwriters rather than copywriters, as they are better at constructing dialogue.

Think about this technology like a new trainee

It will require a lot of investment of time and resources, and certainly in the early stages is likely to be a drain before it’s lightening your load. Unfortunately, there are no short-cuts. On the plus side, it doesn’t repeat mistakes. So perhaps it’s time to give AI a chance, after all someone took a risk on hiring you once.

e3 will be celebrating the launch of their latest White Paper on Artificial Intelligence with a Champagne Breakfast on the 7th of December at The Fable Restaurant. Register your complementary place here.