How AI is improving CX at each stage of the customer journey lifecycleby
For an enterprise to get the full value of AI, it must identify the best use cases for its specific operation. Zhecho Dobrev shares real-world examples of how companies are improving customer journeys with AI.
An HBR article dating all the way back to 2017 calls artificial intelligence (AI) and particularly machine learning (ML) “the most important general-purpose technology of our era”. Machine learning refers to the machine’s ability to keep improving its performance without humans having to explain exactly how to accomplish all the tasks it’s given.
Machine learning has improved significantly in recent years and is now much more accessible. Systems that can learn how to carry out tasks on their own can now be created. Even if it is now utilised by thousands of businesses worldwide, the majority of significant prospects are yet untapped. The implications of AI will intensify over the next ten years as they affect industries like finance, retail, manufacturing, transportation, healthcare, entertainment, law and education.
Forward-thinking executive managers and business owners actively explore how they can use AI in order to get a competitive edge on the market. And as it often happens with disruptive technologies, the first movers will have an advantage and reap tremendous rewards while latecomers will risk going out of business.
In this article, I’ll look at examples of how AI is used to generate value for organisations in each stage of the customer lifecycle.
Brand & advertising
The customer lifecycle journey typically starts with the customer becoming aware of the organisation, what they offer and stand for. Unless it comes from the word of mouth, this is typically done through the brand and advertising.
A good use of AI in this domain comes from OrangeShark, a Singapore-based digital marketing startup, who uses machine learning for programmatic advertising, thus automating the process of media selection, ad placement, click-through monitoring and conversions, and even minor ad copy changes.
Because of the efficiency offered by its system, OrangeShark is able to offer a pay-for-performance business model, whereby clients only pay a percentage of the difference between customer acquisition costs from a standard advertising model and the OrangeShark model. By completely automating a previously semi-automated task, the company has created a new business model that makes monetisation of massive efficiency gains possible
At the other end of the spectrum, Affectiva, which calls itself an “emotion measurement” company, houses the world’s largest image database of sentiment-analysed human faces. The company analyses and classifies a range of human emotions using deep learning models that can then be made available to clients. Some applications study emotional responses to ad campaigns, while others help people relearn emotional responses after a stroke. Affectiva has built a business model based on providing intelligence as a service in an area where nonhuman intervention was previously impractical
An example of a good use of AI comes from the company Quantcast, which helps organisations to profile their customer base and find the behavioural patterns that lead to conversions. Quantcast builds a custom model using millions of data points available about the customers (e.g. demographics, presearch behaviours, past purchases) and finds the drivers of conversion. Then, they find audiences that fit this profile and deliver the organisation’s message to them at the “perfect time.”
IKEA, for example, used Quantcast to build audiences interested in kitchens and reaching them with household goods as well as audiences interested in redesigning their living room with couch and table options. Automated campaign delivery and optimisation ensured the right ads influenced the most relevant audiences. With their targeted audience selection, Quantcast was able to bring prospects to IKEA’s website that were 16 times more likely to buy something than average IKEA site visitors. This helped IKEA double the efficiency of customer acquisition.
Udacity cofounder Sebastian Thrun noticed that some of his salespeople were much more effective than others when replying to inbound queries in a chat room. Thrun and his graduate student Zayd Enam realised that their chat room logs were essentially a set of labeled training data — exactly what a supervised learning system needs.
Interactions that led to a sale were labeled successes, and all others were labeled failures. Zayd used the data to predict what answers successful salespeople were likely to give in response to certain very common inquiries and then shared those predictions with the other salespeople to nudge them toward better performance. After 1,000 training cycles, the salespeople had increased their effectiveness by 54% and were able to serve twice as many customers at a time.
Notice that Udacity didn’t try to build a bot that could take over all the conversations but one that advised human salespeople about how to improve their performance. The humans remained in charge but became vastly more effective and efficient. This approach is usually much more feasible than trying to design machines that can do everything humans can do.
Another very good use of AI in this customer lifecycle stage is in providing personalised recommendations. My Amazon home screen and the suggestions I get don’t look the same as yours, but nobody is picking those suggestions manually - they were created with machine learning. Similarly, automated bookseller Bokus.com in Sweden reported that its average turnover of customers increased 100% for each open of its digital, personalised recommendations newsletter.
Organisations use AI bots to automate the registration and data gathering process. Bots are now coming of age (at least with companies who learned how to strike the right balance between what interactions to be handled by bots and when to transfer the interaction to a human or where to redeploy the human element). When you hear about bots, you might be thinking of the not-so-helpful bots on the webpages but there are also bots that help by bringing data to existing workflows in as low-impact a way as possible. For example you may be typing away in Slack or Salesforce Chatter, asking someone how many customers you have in the retail industry. The best bots today will see that question, and jump in with an answer.
Organisations can also use bots to create automated learning experiences with behaviour-based nudges, the way Coursera and Duolingo uses them, to ensure new customers learn how to get the best of the organisation’s system and services.
Product & use
When it comes to improving the products and the customer experience using them, there are many ways ML can help.
One of those is in optimising the production process. Historical data can be used as a training set in order to form accurate predictions. The result of this work is one or more models that can predict the most likely outcome of the technical process or the set of options, among which the best is chosen.
For example, Yandex Data Factory developed a recommender service for Magnitogorsk Iron and Steel Works (MMK) that helps to reduce ferroalloy use by an average of 5% at the oxygen-converter stage of steel production. Not only it saves about 5% of ferroalloys but, more importantly, this happens with sure and steady maintenance of the high quality of resultant steel.
Another use is in product performance and predictive maintenance. Manufacturers like Caterpillar install lot’s of sensors in their new machines which provide a constant feed of data about brakes, lights, engines, hydraulics, etc. That data is then used for learning purposes in order to calculate when maintenance is needed and then schedule repairs and upkeep before a severe failure occurs. Similarly, car dealers can use ML to predict when a part or car equipment will fail so they can offer after-sales service at the right time.
Amazon and other ecommerce organisations also rely on ML algorithms for product demand forecasting and trend-spotting. For instance, Domo users who are merchants can extract data from their Shopify point-of-sale and ecommerce software, which is used to manage online stores. The extracted information can be used to generate reports and spot trends in real-time, such as in product performance, which can be shared to any device used by the company. The platform is supposed to issue new alerts and notifications for significant changes, such as the detection of anomalies or new patterns in data (similar to approaches used in cyber security already).
Billing & payment
Symend’s digital engagement platform uses behavioural science and data-driven insights to empower customers to resolve past due bills.
The QuickBooks Online chatbot, which Intuit is currently experimenting with, lets users query their phones in a Siri-like fashion to get answers from and execute tasks in their QuickBooks account. For example, a user might ask, “What clients owe me money?” and then tell the system to send reminders to clients to pay overdue invoices.
For example, AI has been applied to analyse past calls, compare them with customer history records and create models that can accurately predict a customer’s risk of attrition. That predictive analysis can be used to guide interactions with customers in real time – during the engagement – to reduce the risk of customer attrition and improve agent performance. AI has also been used to create highly accurate categorisations of customers that are likely to call back or otherwise reengage the company based on the outcome of the initial contact.
For an enterprise to get the full value of AI and automation efforts, it must identify the best use cases for its specific operation.
For example, AI can be used to find a cause effect relationship between what actually transpires during a contact (the words used, emotions expressed, questions asked and more) and the action the customer ultimately took.
AI can help identify and isolate many variables to provide better insight into cause and effect. These insights can inform process improvements, such as developing new scripts or pathways to guide contacts to the desired action.
Customer service matters to today’s businesses and American Express is no exception. The credit card issuer has been open in its efforts to integrate machine learning into its customer-facing functions, as the company’s AI director and VP of natural language processing and conversational AI told VentureBeat in October.
For American Express, NOVA represents “NLP-based automation for customer servicing.” NOVA, the presentation claims, underpins many of the financial service firm’s customer service applications, including:
- Transcribing voice to text.
- Processing travel bookings.
- Automating customer service chat.
- Enabling search in the AmEx mobile app.
- Classifying emails for delivery to the right departments.
The company has not lost sight of the need for human assistance in the AI technology it is pursuing, however. While American Express is “automating a large chunk of the queries” it receives, Madhurima Khandelwal, Head of AmEx AI Labs recently told Experian, AmEx works to integrate human-assisted AI, “for very complex queries, transitioning those seamlessly to live customer care professionals.”
Prevent customer churn
Increasing customer satisfaction helps reduce churn. AI is also being used specifically to prevent churn by analysing historical data to identify at-risk customers so companies can proactively take action to engage customers and get the chance to improve customer satisfaction.
Improve agent training
The predictive powers of AI can be used to accurately forecast how well individual agents will perform in different situations. That insight can be used to identify where additional training and coaching needed and to tailor the guidance to the agent.
But what we need in order to set up and train an automated bot are real conversations. As we stated, the speech transcriptions from your call centre are a data goldmine to train chatbot interactions.
The most effective rule for the new division of labor is rarely, if ever, “give all tasks to the machine.” Instead, if the successful completion of a process requires 10 steps, one or two of them may become automated while the rest become more valuable for humans to do
For instance, the chat room sales support system at Udacity didn’t try to build a bot that could take over all the conversations; rather, it advised human salespeople about how to improve their performance. The humans remained in charge but became vastly more effective and efficient. This approach is usually much more feasible than trying to design machines that can do everything humans can do. It often leads to better, more satisfying work for the people involved and ultimately to a better outcome for customers.
When you choose the task for applying machine learning technologies, you should choose the one with measurable results and economic effect. In addition to this, the availability of data is required, as well as understanding how these recommendations and predictions should be used practically.
The basis of accurate predictions is formed by historical data which is used as a training set. The result of this work is one or more models that can predict the most likely outcome of the technical process or the set of options, among which the best is chosen.