Profile Picture

ICS Heart of AI report: don't force the customer, let them choose AI

I am frequently asked to contribute to publications or to speak at events, for example the Institute of Customer Service “Heart of AI” report, published on 17th July this year. This is a good state-of-the-nation type AI overview report, diving deeper into 1000 consumer and 1000 employee views on AI, from a customer service viewpoint (108 pages, £500, or you can get the 14 page executive summary via Capita, one of the four main sponsors, for free here.

I like a lot of the material in the detailed report; it goes through the consumers’ usage, exposure and awareness of virtual agents, and it identifies direct contact channel customer service as the areas employees most see their organisations starting to work with. I think Mark Bernard at Capita summarises some of the material quite nicely. I didn’t like so much the ICS splash page which altered slightly their research wording to create an attention-grabbing headline quote of

“Half of UK consumers (48%) would stop using – or switch – organisations if they were forced to speak to robots over people."

Frankly I don’t think customers are going to answer a survey positively on being “forced” to do anything, and I don’t think that’s the message that comes through from a lot of the report material, indeed it clashes with other numbers in the report on consumers’ willingness to talk to a virtual agent if offered through chat or smart speakers. But, regardless, we need as an industry to make using AI a preferred customer choice as part of an overall experience, not an enforced chore.

When we offer a great user experience we see people choose conversational AI experiences above the other options they have. By the end of 2017, 20 Million people had bought an Amazon Alexa. [1] I have colleagues, family and friends who have them in their children’s bedrooms. That’s a huge step! You wouldn’t allow your children to phone a random human and ask questions of them without your supervision. But they do let a conversational AI do that, and indeed prefer it over a tablet or a PC in their room. Over half of US smart phone owners in 2017 chose to use their virtual agent, Siri/Google/Cortana on their phone [2]. There is no need to do that if you don’t want to; you could press buttons, type a Wikipedia search, or ask a friend on Facebook messenger and achieve the same thing, but they choose a conversational AI experience. Users are very good at working out what they like, what’s quick and easy and differentiated, and what’s not. So if the survey asked a consumer:

“If you had to go through yet another additional ‘Interactive’ Voice Response (IVR) step to talk to a real person when you phoned your XYZ company, what would you do?”

It seems to me that’s the kind of the question that the (1007) UK consumers felt like they were actually being asked in the ICS survey. Reflecting the frustration of the years and years of salami slicing costs in customer service, and the resultant creeping implementations of barriers to customer service, rather than enablers. Things like IVRs and 1st, 2nd, 3rd, n… support levels, that feel anonymous and low value, almost like the organisation doesn’t want to talk to us. Actually, in most cases that couldn’t be further from the truth: organisations would love to speak us, to establish a conversation with us. But the way we’ve set up as a contact industry sales and service, and the way we manage the times we try and achieve tasks with users just doesn’t enable that to happen.

If used right, AI has the ability to make those barriers redundant and to create time and space for a real conversation to start to happen. But that’s only possible if we’re willing to empower AI and make it transparent and insightful enough for people to select that experience. It’s not about forcing it on people. It’s also so often not about the choice between talking to a human and talking to robot - customers are almost never offered that choice. Normally instead of a choice to talk to a human the users choice is:

  • Download our app, and try and work out how to self-serve this task
  • Go through this 5 level IVR before you’re allowed to talk to anyone
  • Could you phone back when we aren’t so busy or (if you’re lucky) maybe we could call you back.
  • Please talk to this person in 1st line support to restate all the information again, who you absolutely know can’t help you.

If the choice for AI is against one of those, and if you’ve got an AI reaching out to you who knows you, that you can try out instantly, and who has shown insight in the way they approached you: would you not give it a go?

If you had a good experience would you not go back?

If it couldn’t help, if it handed you over fully-qualified to exactly the right person who could help, quicker and easier than going through any other way, would you really be upset with the organisation? Or is that not a big improvement on the way most organisations approach this currently?

In creating new AI experiences, we need to look at the advantages which AI gives customers in these situations and use it to help them. Three example advantages of AI:

Contact staff are generally not given the time to and don’t have much inclination to wade through a lot of repetitive background reading before every case. On the other hand, an AI doesn’t get bored and can read everything it’s provided in milliseconds. AI should turn up informed, every time, and it should use that information to help the customer. If the first thing your AI says is “Can you give me your name”, and “What can I help you with today”, you’ve already lost. You know your customer, you have an established channel with them, and you’ve probably got a pretty good idea of why they are trying to reach you – the AI needs to also have that information and use it to get off on the right foot. We must use the data we hold as organisations to differentiate the experience for users, and you can trust your AI to always use that data in an approved and consistent way.

Not all contact staff are the same. There are awesome people out there (Thank you! We really need you in the industry). But in a lot of call centres, especially outsourced ones, there is often more than 50% chance the person you are talking to has been there less than 3 months, and the experience you get from person to person is going to wildly differ. Sometimes that’s to the customer’s advantage, and they exploit that, but often it’s to their disadvantage. AI is always consistent about the way it approaches the tasks it’s been trained on and the training isn’t lost as individual staff members come or go.

Finally, we all know we have wait time issues in the contact industry, especially at peak times. But the whole way we approach contact with our users makes that worse for us. We make people contact us on completely different channels than those they use in the rest of their lives, and at times which are inconvenient for both us and them. That’s something we need to solve. AI works particularly well on the messaging channels which are the primary way most users communicate in their personal lives. AI then lets us give an instant, personal response on that, their preferred channel. It may or may not be the final answer after the case is fully resolved, appealed and closed, but it’s the best answer we have right then, right there, and should always be hyper-personalised to the individual.

If the survey question read:

“Next time this organisation has an issue affecting your service would you like them to reach out and let you know what’s going on? On your preferred channel? With detailed information unique to your situation and how it’ll be sorted for you personally?”

Would people really abandon that company? Or flock to it?

That’s why at ThinJetty we see so much potential for AI in proactive outreach. There are so many situations where, if you can use AI to get on the front foot and reach out to customers, you can scale with a speed and personal touch you simply cannot achieve using traditional contact measures.

[1] Estimates from CIRP for 20M match with Bezo’s quotes of tens of millions in the <a href=>earning release

[2] You can get widely different numbers on this depending on the demographic you survey, the location you do it in, and the means by which you conduct the survey. So here I’m quoting Voicebot survey. The ICS survey puts it lower at nearer a third depending on type/channel. You can see that a telephone conducted survey to people’s home landlines (unfortunately often the methodology though not always) is going to select a very different sort of user, with very different contact preferences, than if you were reaching out via Facebook messenger.

Back to articles list

We build brilliant bots

Humans and AI working in partnership