The Future of Customer Service Part Two: Predictive Customer Service

6 Nov 2023

8.5 min read

This is a guest post written by Sham Aziz. With 20 plus years in the Customer Service game at companies such as Selfridges, Vertu, NET-A-PORTER & Ocado, Sham is a visionary Customer Service leader. This is part 2 of 4.

“I skate to where the puck is going to be, not where it has been” - Wayne Gretzky

In the first track from the mixtape: The Future of Customer Service, The Role of Customer Service (LINK), I laid the foundation for moving from a reactive CS proposition to a proactive one. Today, we are going to build on this solid base foundation by introducing a new kid on the block, step by step (ooh, baby - if you know then you know), Predictive Customer Service

Much like Wayne, we need to anticipate where customers' needs will go and be there ahead of them. Yes, even before they know what they want. If this sounds a little like the Precogs in Minority Report, predicting crimes (against service) before they happen, then you’re on the right sort of track. But rather than a group of people in a pool having visions, this predictive behaviour will be supported through data (both structured and unstructured).

Incoming generalisation, If you've always done something a certain way, you're likely to do that same thing in the same or similar manner in the future. A subtle nod to how habits might be formed. Therefore the products we’ve purchased, when we purchased them, and what we’ve looked at since among many others – we can start to make predictions about what comes next for you.

What does this look like? Say Adidas launch a new collaboration with Kendrick Lamar. They know you’re a fan of some of their previous collaborations based on your purchases and viewing habits. Combined with multiple other data points, it wouldn’t be the biggest leap in assuming you would want exclusive and early access to the new launch. This almost works today via their Confirmed app.

Why not go one better and send you a box of trainers that arrive automagically on launch day? If you like them, they’re yours to keep and your account will be debited. If you don’t – send them back, free of charge. That’s predictive customer service in a nutshell – we anticipated that you’d like something and so we’ve sent them to you.

Where are we now?

Reactive customer service – Agents are effectively “waiting” for customers to contact them, and would rarely think to or have planned time to contact customers directly. Instead dealing with where is my order (WISMO) or where is my return / refund (WISMR), adding little value to everyone involved. 

Lack of complete joined-up data – Retailer A doesn’t know that you bought product X from Retailer B, so they keep sending you adverts and offers for it and similar products. This leads to wasted effort and spend for retailers, and a worse experience for customers. 

Unpersonalised recommendations – current recommendation engines are generic and don’t fully understand what your previous behaviour says about what you are going to do next. 

The unknown unknowns – customers don’t always know the information and updates that they need. AI could anticipate what customers need and offer it to them before they ask. 

How do we get there?

Reactive to Proactive to Predictive

Customer Service is a reactive function. Customers raise an issue, customer service addresses it. This has always been the case, and this will always be a model in the future for at least some occasions. 

However there is a fundamental problem at the heart of this: the customer has an issue. That means the customer has noticed something wrong, or has a problem and has been unable or not willing to find a solution. Self-serve isn’t for everyone. 

Brands have figured this out and through experience we’re seeing a move towards proactive customer service. This means detecting problems as they happen, or just before and reaching out to offer a solution.

For example, a crucial delivery is late, so a customer service agent reaches out to the customer to apologise, update and refund the shipping cost. The customer has had to do nothing, and has a reasonable outcome. 

Or, a customer has purchased a wooden table from a furniture brand. Customer Service reaches out after some time to ask if the table needs treating, and offers the products or services to do that. 

This kind of process can be run via AI with machine learning. Understanding when there is a breakdown in patterns, such as delivery procedures, or when tables start to show wear and tear (time based) is something that AI can learn, and then adjust based on how customers respond. 

But AI can ultimately be helping right now by automating responses to the reactive customer service issues, freeing up human agents to be proactive. Then over time, the human agent, augmented by AI can start to be predictive in their customer service. 

Ultimately the goal is to be predictive, and to reach out with a solution before a customer realises they have a problem. The journey from reactive to predictive runs through proactive. 

Going beyond simple recommendations

Recommendations are part and parcel of every retail experience, and customers are used to them. Sometimes they are spot on and at other times they can be poor. 

As Justin Shanes shared on Twitter: “Amazon thinks my recent humidifier purchase was merely the inaugural move in a newfound hobby of humidifier collecting.” 

Major life changes such as moving home, having a baby, starting a new job seem to trigger and drive adverts that put more ‘relevant’ products and services in front of you. Sometimes following you around the internet or via push notifications from Amazon offering me new toys (my son using my account to browse!)

This points to a level of capability of retailers and advertisers making good recommendations when they have the right data and are processing it in the right way. Whilst this could give Big Brother vibes, in reality you can opt out. 

For future customer service, this will mean using AI models to make actual predictions for what a customer might like and then recommending them. Using machine learning, these recommendations would get better over time. A combination of self learning and direct feedback to optimise the model.

Customer Service agents could quickly and swiftly swipe left or right on which recommendations are most suitable, with the occasional wild card thrown in for good measure. Who knew I would want a single cable dock setup for my home office? Turns out Amazon did. 

Taking autonomous action

Building on our personal concierge model in track 1, in the future, customer service agents will use recommendations from advanced AI models to actually order items on behalf of customers.

When Ocado started doing home food delivery it had to work out what to do with items that were not available at the time of packing the order on the day. Rather than leave an item out, substitutions offered a suitably close / similar alternative. The default position started with a customer having to opt-in to substitutions. But a move to opt-out soon demonstrated that more often than not, customers accepted the substitutions they were given, even when they had the option to reject them at the doorstep. 

By the same logic, retailers could move towards anticipating orders and shipping them to customers before customers want them. An early and simple example of this is the move to businesses adding a subscription option. The conversation would be as follows: Hey customer, you purchased this item twice in the last month, shall I set up a subscription for you? This could save you money and it would be more sustainable. Or, I noticed you purchased this product recently, it is a known subscription product, shall I get that setup for you?

A bigger and bolder step would be to order an item that customers may not know they want for them. The customer would then be free to keep the item, and have their account debited, or return it at no cost. 

There are a few logistical challenges to make this process seamless. First of all, customers would have to allow it (more on that below), but also the recommendations would have to be so good that customers would accept the item more often than not. Finally the returns process would have to be incredibly simple, otherwise retailers are burdening customers with a box they’ll have to find time to return. 

All of these challenges are surmountable for the bold retailer, with a customer service team of Personal Concierges supported by AI. 

A journey to AI acceptance

The major obstacle to this is accepting AI’s role in customer service. Surveymonkey revealed that 90% of Americans prefer humans to AI, claiming that humans understand their needs better. 

It’s possible that a lot of customers have been burnt by poor customer service chatbots that aren’t built to address and resolve needs, but instead send customers down a variety of rabbit holes and dead ends. 

But with advances in generative AI, the ability for customers to detect the difference between AI and humans will grow smaller. 

As AI improves and tackles more use cases, and as long as humans are in the loop, ready to step in when the AI fails, then customers will accept it more. 

Top tip: don’t attempt to hide your bot (I prefer virtual assistant), be upfront with customers and make it really easy for customers to find human help. Customers will thank you for it and trust will continue to grow.


For retailers competing to win over customers and increase orders, customer service will be a key battleground. There is a real opportunity here to make or break the relationship one conversation at a time. Don’t waste it on poorly thought out customer journeys and reactive transactional queries.

Wayne Gretzky is perhaps the most successful ice hockey player ever. He saw a competitive advantage in anticipating what was going to happen next – adjusting his game and going on to score a record 215 points in 1985. A NHL record that stands today and considered unbeatable by most.

As customer service shifts over to be proactive, and then predictive, a new competition ground will be created. Customers will gravitate towards brands that make their lives easier – brands where shopping is simple and pain free. A real incentive to stop overlooking the competitive advantage of what great customer service can do for your business. 

Although you’re not going to get it right every time, by using AI to spot patterns, augmenting your customer service teams and removing friction from the Customer journey, more often than not, you will find yourself where the puck is going to be.