Michiel Gaasterland
Michiel Gaasterland
Chief Brand Officer
Dec 11, 2018

Artificial Intelligence and the eCommerce customer service conversation: what NOT to bot!

Girl with wings in lotus position floating in front of colourful wall

In What to Bot, the first of this two-part mini-series, we cut through the AI hype and ended by offering a simple formula to help you decide whether to bot or not. Overall, the answer was a resounding, yes!

But not everything that can be botted should be.

In this post, we put some of the more extravagant claims for chatbots to the test by examining some real-life service conversations from our customers.

And to close, we share our advice on how to prepare your company for a future that will be won by those who know how to get the best out of both technology and humans.

As ever, we’d love to hear your thoughts!

— Michiel Gaasterland, Chief Brand Officer, ROBIN

Tackling incredible chatbot claims

If you’re anything like us, you take claims such as, “80% of chat sessions can realistically be resolved by a chatbot” (Accenture, 2016, pdf) with a huge pinch of salt.

First, these sort of claims are typically focused on chatbots in the big global messaging apps like Facebook and WhatsApp, which are, in most cases, more limited than proprietary bots. Second, they hardly ever specify the specific companies that these statistics stem from. Third, and perhaps most importantly, these claims rarely mention the specific types of questions being handled.

So we decided to test the 80% claim for chatbots by applying it to some real-life eCommerce customer service questions. In selecting the questions, we looked at these conversations in the context of the customer experience (CX) and the eCommerce shopper journey.

The customer experience in context

The eCommere customer experience in context

Some believe customer experience will become the #1 loyalty factor. Some also believe that, by 2020, it will become the competitive differentiator.

While the hard business-case data on CX is still a little thin on the ground, numerous CX surveys make several things clear:

  • Customers are desperate for an improved experience;
  • 71% of companies see CX as a competitive differentiator; and
  • Only 13% of brands self-rate their CX delivery at 9/10 or better. 

Perhaps the main reason why so few brands rate their CX highly is that the topic of CX covers a lot of ground. Customer experience incorporates omni-channel strategy, web-store user experience and interaction, service and after-service conversations — even the tone and style of customer communications.

The eCommerce shopper journey is complex

The journey a shopper goes through to place an order on an eCommerce site

As we know, an eCommerce shopper journey is made up of various phases: orientation, decision-making, paying, delivery and unpacking, after which there's a possible additional process of returns. 

To negotiate this journey from beginning to end, customers need an awful lot of stamina. They will have to traverse multiple phases and switch between multiple search and recommendation engines, channels, various media and more.

This is what makes the full experience so difficult to manage: you are only as good as your weakest link. And if you have several weak links in a row, the impact on a customer’s experience can be devastating.

Individual touchpoints may perform well even if the overall experience is poor

Understanding this complexity, we’re going to take one question from each phase of the shopper’s eCommerce journey in order to test the claim that 80% of these questions can be resolved with a chatbot manning customer service.

The answer may not be as black and white as you think.

Five eCommerce shopper questions 

We asked our customer base of 120+ leading Dutch eCommerce players about the typical questions they encountered in each phase of their customers’ journey. While it wasn’t a full-on scientific selection, we did look for questions that were representative.


"Hi, my kid is going to secondary-school, and I want to know which size bag is right for him. He is 140 cm tall”

Could we solve this sizing issue with a bot? Well, you could build something around size and height tables to match your products. So yes, you could probably automate something.

However, this shopper is most likely a relational shopper — a shopper actively seeking expert advice to help them decide. As we wrote some time ago, these relational shoppers typically constitute 20% of your store visitors and yet they are usually responsible for a whopping 80% of your business!

This presents us with a lovely opportunity to win this customer over, build the customer relationship, and maybe even cross or up-sell in the process. Therefore, while in theory you could bot this, we'd say that this is a ‘desirable’ shopper question to have in your inbox!


Hi, I am considering buying the ZYxel NAP 102 as an extra [internet] access point. But where can I find how to install this? Can a layman do it?”

Bottable? Well, this is from a website that offers more than 1 million products. Many of these products also work together. So, if you wanted to bot questions like these, you would have to map all this complexity.

Although doable, it might prove smarter to make a definitive user manual available. Why? So that this question doesn’t come up again (and again and again) — no break in the overall customer experience and no service questions. So, we would probably class this as an ‘avoidable’ question. 


I am in the middle of placing an order with a discount code, but according to me 12,5% discount on € 30, is € 3,75 and not € 0,37! How is this possible?"

Are we going to solve this with a bot? Many of our customers seem to struggle with discount codes for some reason. And yes, it is totally possible to bot this. However, we recommend fixing this in the customer experience. As a top brand, with service as a key differentiator, things like discount codes must run like clockwork. Otherwise there is something broken, and it needs fixing.

So, we’d call this an ‘avoidable' question.


"Last Sunday I ordered a pair of jeans with you. [The site says] The jeans are in stock. But as of today, I haven't received a parcel. Where is it?!!"

Note the question mark and the double exclamation points. This person is clearly irritated. That’s something a bot would not easily pick up.

At ROBIN, we would not consider botting these sorts of questions. Instead, we’d allocate our best customer service people (check out our blog post 'Robots Can’t Care: Humans and Empathy in Customer Service') to handle this and, if possible, solve it in one interaction.

While this is an ‘avoidable’ question, it may take some root cause analysis to resolve it. Therefore, it is better to solve this question in the customer experience rather than have a bot handle this question.


"Hi, I have a return shipment. Tomorrow it will have been 13 days since I ordered. Postage takes 2 days. Will I still be on time, because your return policy is 14 days?"

This is a good question. It’s entirely possible that the answer is already somewhere on the website, buried in FAQs, but customers rarely want to put in the effort to trawl through all that info.

So, bottable? We think yes, absolutely! Most importantly, this would make a solid positive contribution to the overall customer experience.

Questions: Bottable, avoidable or desirable? 

The proportion of customer service questions an eCommerce store should desire, avoid or answer with a bot

Understanding the contents of your questions is critical

From our analysis, the claim that 80% of shopper questions can be resolved by a chatbot simply does not hold true in eCommerce. Our data and analyses show:

  • 15% of questions are bottable;
  • 45% of questions can and should be avoided by fixing them in the CX; and
  • 40% of questions are ‘desirable’ and should be handled by a customer service rep.

While these percentages are based on a sample of our customer base —and your proportions of bottable questions would be dependent on the specific type of store you run, your portfolio, the maturity of your service operation and so on — we think these are good principles to work from.

A little more detail on our findings:

  • The 15% of ‘bottable’ questions include any number of FAQs such as: How long do I have to send a return? When will I get my money back after a return? A company could easily bot its brand story, or even bot all the questions associated with the typical issues that arise when making an online purchase — from creating an account to managing passwords, even explaining the full returns process.
  • The 40% of desirable questions are the ones that you really want to have handled by a skilled rep. By focusing on the questions that help reduce shopper insecurities, and close more orders in the cart (which in our experience always leads to 20% conversion and up to 50% conversion for some) there are a ton of commercial and brand benefits to be had.

This is why, while onboarding new ROBIN customers, one of the first things we look for is insight into the actual content of conversations across a company’s customer service function. 

Moving forward, our goal is to help a team receive fewer unnecessary, annoying, and avoidable questions (which, by the way, are also annoying for the shopper) in order to free staff capacity to handle valuable questions for all concerned.

What’s in a conversation?

We know from experience that many retail brands are not aware of the specific nature of the questions asked by their customers and handled by their customer service team. Even fewer brands analyse these questions. In fact, many don’t think it is important. Our advice:

1. Think about the proportion of your orders that generate a conversation.

With some of our leading customers this can be as high as 75-90%, but typically this averages around 40%.

2. Consider the makeup of those conversations and what the smart management of them would mean in terms of potential:

  • Costs savings through reducing, for example, returns;
  • Revenue generation; and
  • Contributions to customer experience.

That’s precisely what the ROBIN Program can help you do.

Now let’s wrap up by discussing what you can to do today and tomorrow to reap the benefits of AI and bots. While doing so we’ll also take look at the longer-term future as well.

Getting started with AI and bots today

If you’re looking to get started with AI and bots today, here’s our simple advice:

1. Keep it real!

We eCommerce folks need to keep cutting through the marketing BS around advanced tech and focus on tangible use cases. Let’s see what is being solved successfully, for whom, and with what product portfolios. Then we can make an informed choice about where to focus. There is no magic bullet.

2. Start experimenting with marketing/assistant bots

Virtual assistants that cater towards new shopping behaviours can generate additional revenue, and are easy to build. There are any number of ways you can use a virtual assistant. Go experiment!

3. Focus on customer experience gains, not just cost-cutting

Many people focus on cutting costs, but where you really win in terms of service is in making gains in customer experience.

Looking at the actual content of your service questions — and digging in deep — will help you make that transition. This will also free-up capacity to shift your service focus from after-sales to pre-sales, where it generates much more business value. 

Preparing for tomorrow 

Here’s three things to consider if you want to stay ahead of the pack when it comes to service and leveraging tech:

1. Build a sustainable customer service operation

The customer service managers who focus on delivering a genuinely end-to-end customer experience can integrate front- and back-end eCommerce processes to create significantly more business value than a ‘service-only’ focus.

Building a sustainable service operation begins by embedding, and then monitoring, your customer service key performance indicators. Once under control, you’ll need a service vision and ambition.

2. Get structured about data and content

Structure is key. Thinking structurally about business rules, data, and content is the key to capitalising on advanced tech. Introducing the concept of a business rule across your company will help you build a strong foundation.

We know from experience that there are loads of learnings to be had from analysing shopper conversations. These insights can be used to optimise both CX and operations — tangible proof of the benefits of thinking structurally and tackling root causes rather than symptoms.

3. Automate and Augment

Currently there’s a strong focus on automating the front-end of customer service. And yes, identifying repetitive, low-value tasks that can be automated is a good way to improve efficiency and productivity.

Augmentation, however, means equipping customer service staff with increasingly intelligent tools with valuable data and information — effectively turning them into super-reps. We recommend that’s where you focus on developing use cases, and look to scale.

And so, on to our closing thoughts.

We have covered a lot of ground in these past two posts. And technology has been firmly at the forefront. But there is something to keep in mind at all times.

Never underestimate a human

Guy making a V for Victory signImage: Eddie Knop

We all know that customer experience is both hugely important — and tough to manage.

We know that our shoppers’ journeys are complicated. And they’re only going to get more complicated over time. More competition, more tools, more channels, more content. Just more.

So many touch-points… so many customer questions… especially if you have a broad range of products.

For a shopper, service conversations are ‘make or break’ moments with your brand. These instances are ‘make or break’ moments for the customer relationship.

That's why we believe that service conversations are critical touch-points for the entire customer experience.

Nowadays only the most skilled human reps can navigate this complex environment while delivering great customer service with a smile.

Current technology is not capable of matching this level of service, at least just yet. And it may never be.

As more and more bots appear online, and customer service interactions increasingly become automated, personal service will become an even more scarce commodity than it already is… and therefore; even more of a USP.

Elon Musk, Tesla and SpaceX CEO, isn’t everyone’s cup of tea. But as one of the world’s biggest advocates of advanced technology, he said something very interesting:

“Humans are underrated”.

Tweet by Elon Musk saying: "Yes, excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.

We wholeheartedly agree.

Yes, robots rock, but you should never underestimate a human.

New call-to-action

Customer service, eCommerce, chatbots, AI


Recent posts