Going from "would you like fries with that?" to "would you like to upsize your meal" implies that the customer already wants the fries. Smart use of AI helps to upsell in a similar manner.
"To upsell clients, we need to understand who they are first," said Paul Baptist, senior director of solution engineering APAC at Salesforce. "Recognition of actually being able to say who are you, what do we have to offer that you will need, and then of course understanding when is the right timing for you."
Speaking at the Tech In Asia 2018 conference in Singapore, Baptist said customers of Salesforce that rely on Einstein Discovery have been able to access data that relates to customer behaviour with an online store, purchase habits, purchase frequency, and other variables that allow the AI platform to determine whether upselling is even a viable option.
In doing so, upselling becomes contextual, which reduces the risk of being intrusive.
Behaviours for upselling rely on a variety of trigger points, with Baptist stating that the channel by which an upsell is communicated varies across verticals, buying cycles, pricing portfolios and more.
"Are you an online purchaser, are you someone that needs to walk in a store touch and feel, and then understanding the right channels that deliver capabilities," said Baptist, adding that depending on what works best, a recommendation engine will offer the merchant the option to send an SMS, an email, recommend a call, or an invitation—varying with the conversion data on what works best and when.
Across the sales, service, marketing, communities, and analytics products of Salesforce, Baptist said that customers are able to get the whole picture of where a customer is coming from, where they want to go, and what they need to get further.
The next phase of an AI-powered upsell involves being able to predict the next likely behaviour of an existing customer based on collected datasets of lookalike audiences.
"Einstein Discovery is where we're starting to look back in order to look to the future," said Baptist. "We're looking at it from a point of view that says next best offer and next best action."
The data from this allows customers of Salesforce to learn about the product extensions and iterations that represent unmet elements of the industry profit pool, while also offering an insight to the next most effective course of action that could result in a sale.
While business theory insists that an upsell only occurs with an established customer, Baptist disagrees, adding that with an AI-powered engine a business can draw conclusions in real time about what a customer may need at a given moment that will add value.
"If I'm buying a burger, I probably want a drink and fries with it," he said.
The final part of predictions relates to the Internet of Things (IoT). Baptist said Einstein's connection to internet-enabled devices allows it to draw from past data to determine when a product needs to be repaired or serviced.
By 2020, IBM believes there will over 75 billion connected devices, so this capability is in its earliest stages.
"We have an IoT platform that takes feeds of data from devices, whatever they are," said Baptist. "It aggregates it [the data] while understanding that not every problem is a problem that needs to be fixed."
On the other hand, an overheating oven or printer that uses up toner too quickly may be symptoms of a legitimate problem, he said, adding that the company's IoT platform places such data into Einstein to determine the severity of problems and the suggested courses of action.
In extreme cases, the system may simply contact the service centre and have a repair person at the customers' doorstep without human intervention instructing and intervening, he said.
In doing so, Baptist foresees the acceptance of AI-based upselling as the means towards growing the customer relationship both monetarily and in terms of brand trust.