How AI-powered virtual employees will transform customer experience

Organisations are fast shifting from the ‘pioneering’ phase of testing artificial intelligence (AI) applications into implementation as a way of digitally transforming their businesses, IPsoft’s chief commercial officer, Jonathan Crane, believes.

IPsoft is one of the world’s largest privately owned AI and cognitive computing companies.  The US-based group has been creating virtual engineers for the last 15 years, which manage IT infrastructure by learning what’s required from humans in order to remediate problems. About 30,000 of these are in the market today.

In 2015, the company extended into what it calls ‘virtual employees’ and created a virtual customer agent (VCA) named ‘Amelia’. The VCA is designed to tackle tactical and repetitive business practices and tasks performed by white collar workers, freeing them up for more strategic roles.

Starting in call centres, Amelia now sells car insurance, assists with life insurance, works in retail, helps with wealth advice, and also provides new account information for one of the world’s largest financial institutions. More recently, Amelia became an HR employee and is helping a large mobile carrier’s 60,000 staff for example, with information on days off, or conditions of leave.

Here, Crane shares insights into AI’s maturity, explains the difference between chatbots and virtual agents, and addresses the potentially detrimental implications of AI on both customer and employee engagement.

How do you define AI as a business?

For us, AI is about trying to take that simplistic, tactic, boring work and giving to automation so we can use our strategic abilities.

What’s the difference between what you do and chatbots?

With chatbots, you are basically taking known processes, rewriting them, then putting a digital supervisor over them. It’s automating processes and very narrow in scope. The important difference is that chatbots cover questions or revise processes you already know and use scripted answers.

We’re trying to problem solve. Say I have a customer who wants car insurance: There’s no way to script a chatbot to have a full conversation about car insurance; you’d be surprised at how many ways people ask about car insurance.  

Amelia is trained in conversational intelligence, problem solving and context. She can discern what you’re asking then converse with you on how to help.

Chatbots have a useful cycle in the life of companies building their way, but the goal for us is to get to where you are able, in a semantic interface, to take that conversation, work with it, and solve the customer’s problem. If you can’t, then you get a human on and collaborate to get the answer.

So just what AI is capable of today?

What AI can do is build an understanding of what in fact the customer/prospect or employee is trying to do then problem solve as part of a multi-faceted process. For example, if you call in with a billing dispute, an organisation could take that problem and start thinking about the fact you’ve had two years’ experience, it’s the first time you’re going over your data usage plan, you’re a loyal customer, yet you’re also on an old phone.

In that instance, I’d want Amelia to be performing a customer experience task, but I’m also looking at her to be a revenue generator. It’s not just inbound answers, it’s selling perhaps another year of service to that customer.

Which industries have been early adopters of such capabilities?

Financial services and insurance were first, and the reason is simple: They have large inbound centres facing repetitive questions that lend themselves to the kinds of things we can do with AI. For example, we have one company who was trying to handle 1.2 billion calls per year. They could respond by using human capital, yet those calls typically aren’t revenue creating. They need an appropriate solution instead of that volume of people.

Retail is another as these organisations try to figure out how to go from Web-based purchasing and connect it back to bricks-and-mortar experiences. What makes people come back into the store comes back to information on the Web, associated with activity in the store. It’s knowing a customer’s preferences, previous buying behaviour, and being better able to serve them.

Yet if I’m not armed with that information as store staff, then you as a customer walk in with weapon: A smartphone. You’ve already researched and come with an intent to buy. You want someone with the same level of intelligence and knowledge about you and products you are interested in. Unfortunately, our ability to support in-store staff is so limited – turnover is tremendous. Matching that kind of intent is where a virtual agent could come in.

With AI, we can guide people, help them make decisions and have informed personnel available. If staff don’t know the products, they could wear an Apple Watch and ask Amelia. Or the retailer could have a kiosk set up in-store.

We need to think about how to interact with prospects, customers and our employees in very different ways. This technology gives us a chance to have high-touch capacity without breaking the bank to invest in it.

How would you describe maturity levels for AI adoption locally and globally?

The last three years have been the pioneer phase, where we’ve seen organisations starting to try chatbots, IBM Watson, Microsoft’s Cortana. There are opportunities out there that allow you to sample and buy.

We’re now progressing from that ‘trying phase’ to implementation. Everyone sees digital transformation as a prerogative and businesses must adopt to the new digital age, which means they have to change their cost structure, processes, and consider new competitors they haven’t thought about.

As a result, every CEO is thinking about how to transform, and most importantly, how they change the way customers and prospects see and interact with their company. The key is customer experience, and that’s where AI comes in. Yes there are operational savings, and digital benefits, but they’re givens. It’s really about how your play in this new [customer-centric] arena.

Are there instances where using AI could be detrimental?

The question people always raise is: What happens to jobs? I see AI bringing great levels of innovation, new services, businesses and ways of doing things differently. We’ll see new companies popping up everywhere taking advantage of these technologies, and you’ll have a proliferation of those supplying very niche AI-type solution sets.

A good example is in outpatient care, physical therapy and as an extension of healthcare. AI could be used to support higher levels of care we can’t currently support. It’s a tremendous assistant to things we’ve frankly been negligent about, such as taking care of older generations. The economics make it a darn hard thing to do – AI is a way to augment those people already delivering services.

What sorts of complementing technology or trends are you tapping into to improve your AI?

One area is emotional intelligence. Think about if you asked an HR virtual assistant about how many days off you have, then say you’re feeling under the weather. If you’re Alexa, you’d probably quote the temperature of the location because you didn’t understand that phrase – but Amelia does. So AI has to exhibit empathy. Without the emotional element to this humanisation of functions through digital, you’re missing a key aspect.

The second thing is learning. We expect that from humans – just think about those people in a call centre; you expect them to learn every day. For us, it’s about training Amelia rapidly, but allowing her to grow in knowledge and be able to provide analytics, marketing analytics to that learning, as well as to with emotional intelligence and behaviour.

How far away are we realistically from AI being able to do that?

In Amelia, we have already built in that first level of emotional sentiment, can read that in text and voice, and she has the ability to be empathetic. The marketing analytics means you can train her in responses – we call this ‘process ontology’. She can learn what works and doesn’t work and inputs that into her procedures. 

There are some good people doing further work in intent. Connecting voice and dialogue is another – I’d love that if you called in, I’d be able to detect if you have a Lithuanian accent then connect you to the person in the call centre that’s Lithuanian. These ways to point customers to the best place to be served are really starting to emerge.

Amelia is the front-end, so she needs to be able to rapidly accumulate knowledge in the back-end so she can respond. If she’s connecting back to antiquated systems, she’s as slow as humans. Through robotic process automation, we can make the back office rapid as well. We’re working on that through our 1Desk offering – that’s the end-to-end, and digitalisation of the entire company.

What advice do you have for organisations looking to start to harness AI?

It’s taking the people along with you to identify what key problems areas exist when interfacing with clients, prospects or employees, then developing those juicy use cases. If you were going to put a human against a problem, where would you put them? This is driving up your value chain. You’ll find the thorny issues and that becomes a good use case. 

If all you want to do is answer a few questions from the client base, then use a chatbot. Get started with that level of investment. But if you really want to think about how to interact, then this is a chance to do human-like functions with digital labour. Mix that then with your humans and provide a much more superior service level to the customers you desperately want to hold on to.

What about the ethical considerations around AI, from job loss to bias - is there a way to minimise those risks?

We do need to think about adult re-education. Young people coming through STEM understand the uses of machine learning and will be well skilled. What we need to pay attention to are people operating in jobs susceptible to being automated. It’s not unlike automobile manufacturing when robotics took over a lot of those functions humans used to do. We need to be cognisant of the jobs likely to be automated, then take those people and offer the chance to re-educate and build it into the way you’re adopting this technology.

That’s the responsibility of companies but government can help here, too. There are so many jobs that will now appear in new digital environments, such as digital workforce management and supervision, so educate people around those. Digital collaborators, keeping software updated – these are things we have to give our minds to.

Follow CMO on Twitter: @CMOAustralia, take part in the CMO conversation on LinkedIn: CMO ANZ, join us on Facebook: https://www.facebook.com/CMOAustralia, or check us out on Google+:google.com/+CmoAu