Customer Service with AI

As a business grows, one of the common challenges its faces is keeping up with its customer connections. While there are many different mediums available for customers and brands to interact with each other, contact centers remain one of the most popular mediums for customers to have conversations with the brand especially for industries as Insurance and Banking where the subject or customer queries are more complex in nature. The key challenge, in this regard, is as the business grows the need for scaling contact center agents fell quite naturally and thus a near linear relationship with input cost. Despite these increases in costs, other means of customer touch points such as emails and agent based web chats, remain a source of annoyance and drive customer satisfaction down.?

To make things a bit more interesting, imagine the business being multi-country, multi-language, that drives the language specific and region specific staff recruitment further challenging both from quality of service delivery as well as its cost effectiveness.

Let's take an imaginary scenario, depicted below in the table, of a multinational insurance company having operations in 3 languages - English, Mandarin and Arabic. A simple customer service metrics of current state may look like as mentioned below:

In this scenario, time taken to respond to customers, most often, drives customer satisfaction. Response time is dependent on volume of queries as well as complexity of queries. Thus not only the current business situation would need to be improved, in aforesaid example scenario, additionally,? as business grows, keeping up with these metrics, especially phone call volumes, email service and web chat would require sizable investment in contact center staffs especially if the expansion happens to a new region requiring an additional language support to be added. A mere single digit or low two digit improvement in these metrics could stand a significant gain for the business and surely a huge tick in customer experience.

Let's now look at a potential chat-bot solution to help ease the work and also not let into the trap of organic increase of agent staff.

1. Data analysis is the first step and identify what are the basic tasks that customer would need support with -?
2. Next comes the complexity of languages and common proverbs. There are region wise common terms that need to be identified and understood apart from language proficiency of the solution. A simple case of localization!! Additionally, insurance itself is a complex subject and thus model training requires significant domain expertise. Refer my previous article on AI in business applications to understand a relationship on AI complexity with respect to domain expertise and how should you choose an AI initiave for implementation.
3. Most importantly, the solution should not just provide information rather it should be able to complete a transaction, say, ability to say you have $300 left of Dental service. This ability will actually close the customer query
4. Lastly, ability to have a seamless transfer or integration between chat system and core CRM system thus there are no residual manual work for agents in identified transactions while interacting with customers. Summerization of conversation, follow-up actions and reminders etc are all actioned through the model.

Aforesaid is a list of first few items which take an organization towards leveraging AI in customer service and it will surely be encouraging to see the value delivered.?

There are many other possibilities in this space, as next steps e.g.:

1. Summarizing phone conversation and building insights out of it as, currently, most phone agents need to summarize the calls manually.?
2. Summarizing calls,, building insights and creating actionable items for escalation handling
3. Sales summaries for leads to all the way conversions and providing likelihood of conversion
4. Assistance in composing replies based on customer query and its association with the brand
5. Cross learnings of expert insights from closed deal transcripts, which may work as virtual coach to sales and services agents
6. Impromptu contextual information available to agents while on call with a customer to provide an effective and relatable service

And the list may go on. The effective use of AI in contact centers opens many avenues of scaling operations and brings measurable benefits to both brands as well as customer experience.

Finally, how do you measure the success of AI investment?

After all, a solution itself may not take the entire credit of improvement and vice versa. This also differs per business and thus careful definition of success metrics is essential. While some simple metrics could be that a portion of Web Chat or Email Service agent staffs have now been repurposed in telephony call center agents, a more outcome based approach would be to measure the performance of this channel / solution as:?

1. What is the service containment percentage that this new channel has serviced over a period of time? I.e. a request initiated in this channel (if you consider this as a channel otherwise model / solution), has been successfully closed 30% of time (say) month-on-month. This may also have another aspect of deflecting away from human intervention and measuring the volume accordingly.
2. What is the percentage share that this channel (or solution / model) supports the customer base out of all channels or touchpoints