Insight of the Week: Applied AI and Amazon's CX Wizardry
15 years ago, Amazon's first SVP of Customer Service, Bill Price, wrote a book: The?Best Service is No Service.
The general idea was that customer don't really want to have a 'relationship' with your customer service department. They want to get on and use your product, or service. When they call (there wasn't so much chat back then!) it's often a sign of dysfunction.
He introduced a way of thinking about the different kind of contacts you receive that is still super-useful today: The value irritant matrix.
But the tech has changed. We've got Conversational AI and Generative AI now. So is this still relevant?
Yes. In my view, advances in AI make the matrix?more?valuable, not less.
You can deploy AI powered call routing that asks customers "How can I help", so you know exactly why they called, not just which menu option they choose.
You can deploy a website chatbot that lets customers type in their question, so you know exactly what question they have, not just which part of your website or FAQ they browsed.
AI enabling the initial touch points is the first step in the?Applied AI playbook?I introduced in last week's email.?
The second step, is to decide how to use the data you get from AI enabling the initial touch points. For that, we use the value-irritant matrix.
Here's how:
1)?Rank each of the things your customers ask for their in order of frequency. Let's call these 'contact types'. We want to work on the highest volume contact types first. It helps to group queries that are very similar, but don't get too broad - one of the brilliant things about our AI enabled touchpoints is that they allow us to handle even quite similar contacts, in different ways.?So combine things like: "Have you sent my order yet?" and "Has my order dispatched yet?", because they really do mean the same thing. But keep something like: "what's your returns policy" and "I want to return an item" separate - because the first is a question that might occur prior to purchase, and the second is a request to actually do something.
2)?Think about contact type from the perspective of the customer, and the perspective of your business. Is the contact valuable to your customer? Or is it irritating? Is the contact valuable to your business, or is it irritating? Score each contact, on each dimension. Pick whichever scale works for you, but 1 to 5 is a good start, with 1 being really irritating, and 5 being really valuable.
领英推荐
3)?Finally, you're going to plot these contact types on a grid. We typically use a spreadsheet to plot this out, and use the size of the dots (or blobs?) to denote the volume of each contact type, so the higher volume contact types stand out.
Now, in the past, we'd have focused on the bottom right hand corner, things that are valuable to customers, but irritating to the business, because that's where the biggest opportunity for automation lies. As you can see in this example:
But AI has moved on a lot. We can do more now, so the Applied AI playbook looks at all four segments:
Simplify?(valuable to you but irritating to your customer) - make these interactions so simple the customer hardly notices. Use Biometric authentication. Make it easy to switch to a digital channel like web, mobile or social, if that's going to make things easy. Streamline the feedback process so it's just a click of a thumbs up or down button, or a single, easy to answer question.
Eliminate?(irritating to you and your customer) - remove the root cause of these contacts. Improve product design & back-office processes. Use AI to identify issues before your customer notices, and provide pro-active updates.
Automate?(irritating to you but valuable to your customer) - this is the classic automation opportunity. Use AI to give your customers what they need, without wasting expensive human agent time on tasks that don't add value to your business. Assuming you already AI enabled the initial touch points, this is where you can expand the capability of your website chatbot to answer more questions and automate simple transactions. Same with your NLU call routing system. You can add answers to frequently asked questions, and route customers to new or existing self-service capabilities.
Leverage?(valuable to you and your customer) - You want to leverage your best agents to engage, retain, upsell and satisfy customers in high stakes contact types. This is an area that's really opening up as a result of advances in AI. Newer, generative AI powered solutions can provide contact history summarization, a chat powered knowledgebase, and highly relevant suggested responses. Whereas older tech struggled to justify ROI, recent research has demonstrated that Generative AI powered Agent assistance tools like this can improve overall productivity by 14%, and help new agents achieve the same performance in 2 months, as agents without AI assistance achieve in 6.
The beauty of this approach, and the power of modern AI, is that the first step: enabling the initial touchpoints, has a dual function. It gets you the data you need to drive your strategy - to decide whether it's a contact you want to simplify, eliminate, automate or leverage - but it also enables you to dynamically select the right kind of AI at the right time.
So if you haven't already AI enabled your initial touchpoints... what are you waiting for?
And if you need a little helping hand... you know who to call ??
Kerry Robinson?is an Oxford physicist with a Master's in Artificial Intelligence. Kerry is a technologist, scientist, and lover of data with over 20 years of experience in conversational AI. He combines business, customer experience, and technical expertise to deliver IVR, voice, and chatbot strategy and keep Waterfield Tech buzzing.