Personalization Context Virtual Assistant
April 23, 2020 • 4 minute read

Sorry, do I know you? The role of personalization and context in virtual assistants

Modern user interfaces are sometimes customizable but are rarely truly dynamic; you may be able to change your phone’s accent and what it calls you, but otherwise its responses don’t change whether you are 18 or 81, a new user or an old friend. This is not, however, the way humans interact naturally with each other. We alter both content and delivery when switching from talking to our preschoolers to addressing our work colleagues (a distinction made all the more apparent by our current work-at-home environment when those switches are happening more frequently and not always successfully). To evolve from offering a “bot” to building an Intelligent Virtual Assistant in the customer engagement space, we therefore need to build dynamic interfaces that take into account user knowledge and context if the conversation is to be natural.

There are multiple ways in which we can leverage this kind of personalization with Conversational AI:

Customer journey context.
When we engage with a customer, is this the first time, or is there a history we can leverage to optimize this interaction? Recall how it felt (before social distancing) when we could walk into a coffee shop and the cashier knew what our order would be because we always ordered the same thing. Offer the same great customer experience with a virtual assistant. Greet the customer with the option of going straight to the same use case they invoked last. Or, if that use case is not one that tends to repeat, leverage business rules to predict the natural next step. Show the customer from that first greeting that they are known, and that we wish to get them quickly to what they want.

Dialogue context.
When our customers realize that a virtual assistant can handle truly open-ended natural language, they drop their guard and start speaking in natural ways – and this includes using pronouns and incomplete references. Gracefully handling this could be as simple as building out the capability for understanding, “I want pepperoni on that,” or, “Change the second one to a large.”
Natural language understanding needs to take into account where the customer is in a process and what has been talked about recently in the conversation.

Customer data.
For a more sophisticated application of context, integrate with CRM and apply analytics to truly understand customer patterns. Although they may not always call about paying their balance, customers consistently called about that topic during the third week of the month, and that is where we are today. Or, perhaps, other customers similar to this one have also invoked a certain subset of use cases. Make those possibilities the easiest to achieve and deliver the efficiency that so many customers prize above all other factors in their customer service experiences.

Expert mode.
Design for savvy repeat users who know all of the prompts and want to shortcut the procedure by providing all of the information in a single utterance. Do not force them into a long sequence of individual questions; simply round up any information that was missed, and then proceed. Recognize that in some channels (particularly text), all of this data could potentially be packaged with the opening intent before the virtual assistant even gets to the first prompt.


Personalized interaction is a huge part of bridging the gap between natural and robotic conversation; and when you feel like an enterprise knows who you are as an individual, and is invested in that knowledge, you will in turn be more invested in that relationship. This is precisely the state of customer engagement we should be trying to achieve with our virtual assistants.

Want to learn more? Let’s talk.