Panel: The possibilities and pitfalls of AI in CX
- 13 August, 2018 08:58
Marketing and customer experience professionals looking to artificial intelligence (AI) to improve their organisation’s customer game are going to have to find a way to work with IT or risk missing out on the real benefits of the emerging technology.
That’s the advice from founder and principal of Singularity Research and former Forrester principal analyst, Tim Sheedy, who spoke on a panel about the implications of AI in CX during last week’s Freshworks in Sydney.
For Sheedy, utilising AI and machine learning needs to be led from within the organisation. “The issue with AI is you can’t go to an external agency to make it happen most of the time. You need data from own systems, processes, and typically data sitting outside the remit of the applications you’re responsible for,” he told attendees.
Yet research Sheedy has undertaken on how CX and IT teams are working together, shows CX staff often going beyond IT to agencies, for instance, to realise technology outcomes.
“You’re going to have to work with IT for a lot of your AI initiatives. You’re going to have to find new, better ways of working with them,” he said. “Take out all that escalation we might have, for example, of my boss speaking to their boss, as that all takes time. In the CX world, you’re often talking about 2-4 week turnaround on initiatives, versus the IT world, which has a budget for 12 months and anything outside of that falls to next year.”
Another big challenge is IT teams aren’t often held to the same KPI metrics as line-of-business. Then there’s the disconnect because of the way teams connect on projects.
“I’ve seen instances where the CX person was catching up with the IT project owner of their specific initiative once every two weeks, and only discovered after two weeks they were doing the wrong thing. That’s because they didn’t have a daily injection of knowledge and information and knowing the right or wrong direction,” Sheedy continued. “So even the way we work together as teams is often broken.
“Nearly all the great CX initiatives people talk about are where they had a ‘superhero’, such as people who bridged IT to join the CX team.”
How to think about AI in CX
Understanding how AI can help CX is clearly key in this process. Sheedy positioned AI is an enabler that can take friction out of the customer process to improve it.
“It shouldn’t ever be seen as more than that. Yet often we complicated AI to make it out to be a scary, futuristic initiative,” he said. “When you’re looking at starting on AI… and looking at your customer journey map, you have a current state you’re working through to a future state. You need to ask questions around this, and this is where AI will help you.
“For example, if I predict at that point, will that make the experience better for the customer? If I make it personalised at that point in the journey, will it make it better for the customer? If I make it move, see or sense, will it make it better for the customer? If you start to ask these questions around specific customer experiences, and the answer is yes… then there is an opportunity for AI to improve that customer journey to make that process better and take the pain out. You’ll have a more delighted customer happy to come back to your company again. It could also be a delighted employee or partner too.
“If AI isn’t enriching that experience and making it better, then don’t do it.”
Freshworks general manager for Australia, Sreelesh Pillai, also positioned AI in terms of personalisation, and said the ambition should be to give something to a user upfront without the user really expecting it. His advice was to map any AI project with such ‘wow’ moments.
“When do that, you’re helping the AI project being adopted in the enterprise, because you are supporting the end customer,” he said. “At the end of it all, it always is about the experience you leave with the user.”
Yet in adopting AI, Tanna Partners principal, Greg Tanna, saw a need to manage executive expectations.
“You need to foster realistic expectations as to what can be achieved in a realistic timeframe,” he advised. “Everyone is getting ahead of themselves at the moment. The technology [AI] is very flash but it’s not quite there yet at a top layer. So start small, manage expectations up the chain in terms of what’s acceptable in terms of failure.
“What you don’t want to do is embark on an AI project that doesn’t work and it then sets you back two or three years. No one has 2-3 years. The ultimate investment here is about survival and relevance.
“The biggest danger is blowing out of control, setting a bad tone and turning everyone else off innovation. Everything about technology is exponential – putting your head in the sand only guarantees you’re going backwards at a faster rate.”
Panellists also pointed to the culture element around CX in many businesses, again reflecting the cross-divisional element required to harness AI for experience improvement.
“We can have the best technology, but if there’s a culture problem – like the financial institutions have, bearing in mind they have the largest CX teams across Australian organisations – then it won’t have any impact in the end,” Sheedy claimed. “Customers still leave as there’s a cultural problem.”
In terms of examples of successful AI in Australian organisations right now, Sheedy pointed to digital-only bank, Ubank, which has deployed a chatbot specifically to help with the mortgage process. Importantly, the use of a chatbot came from the insight that customers were contacting them at all times of night around process application, he said.
A different example is at Carsales, where AI is being used in image recognition on the platform to help car sellers optimise images in their advertising by recognising the car brand, age and other distinguishing features.
What could halt AI in its tracks
There are, however, big trends on the horizon that could cause big issues to AI innovation. A big hurdle Sheedy noted was the European GDPR legislation and the need for organisations to be able to unpack every decision being made by a computer in order to justify each one.
“Using Google image recognition for example, how do you unpack that it thought an image was one thing or another?” he asked. “We need to backwards engineer the decisions we’re making. That challenges deep computing – you simply can’t reverse decisions in that space.
“AI also introduces new risk profiles to your business as well. Sometimes you don’t know if it’s going to tell a customer the wrong thing for your business. This might require just a new pitch to people across teams around why it’s important to have increased risk to improve CX. Then there are other challenges we don’t know about yet. And machines telling us what to do – at what level are we happy to accept that?”
But with many businesses still struggling to do basic data mining, Sheedy questioned how many were in fact ready to run with AI.
“There’s no way an organisation that can’t control its data or doesn’t know how to make decisions on data today can take advantage of deep learning,” he added.