AI Customer Support

The 10-Step Guide to Scaling Customer Support — with Conversational AI

by admin

We’re at a tipping point for customer support operations. In the past year, you must have witnessed a drastic increase in customers contacting your business over a greater number of channels for quick access to reliable information.

To handle such sudden fluctuations in support requests, we have been contriving to an unsustainable strategy of increasing the headcount over the years.

But with the unprecedented spread of COVID-19,  it’s impossible to scale the support operations with a human-only agent team. This has led businesses to examine future-proof and scalable conversational AI solutions for automating their support system.

There has been an acceleration in global deployments of conversational AI solutions post-pandemic, especially in the financial, hospitality, and retail sectors — to manage unexpected customer engagements, provide consistent information, drive efficiency, and maintain business continuity.

If you are getting started on the journey toward deploying conversational AI, I can understand that you might be struggling to define a strategy for gaining true business benefit from it. The steps I’ll walk you through will ensure you’re able to leverage this technology for delivering the best customer support experience while saving cost, rather than just having it for vanity. 

So let’s get down to it.

Step #1: Identify top reasons for customer contact and the channel preference

As per the Gartner survey, the most preferred channel for customer service interactions is phone (44%), followed by chat (17%), email (15%), company website (12%), and search engine (4%).

Phone being the dominant channel, you must understand the root cause that generates these inquiries in the first place. As per Gartner’s research, there are six resolution types into which customer service inquiries can be grouped in the phone channel: Transact, Confirm, Discuss, Workaround, Validate, and Vent.

A leading grocery and meal delivery retailer in North America was struggling to meet the sudden surge in customer support requests over the phone, fueled by COVID-19. On analyzing their calls, it was found that most customers inquired about order status, delayed deliveries, coupon redemption, cancellations, and refunds. They realized that for most issues, they could provide resolution to customers over voice in a self-service mode and easily scale their support operations.

Step #2: Define the business objective

Define what business problem you are trying to solveIt could be,

“To reduce the total volume of calls handled by the customer service reps by X%, by implementing conversational AI to handle Y% of initial inquiries. This is to deliver high value to the customer base and optimize costs.”

Step #3: Choose the right use cases to automate

Once you identify the top reason customers contact you through the phone, identify the right use cases to automate. You need to :

A. Understand why these use cases are important, the mechanism that they currently run, and the volume handled. You could easily collect this information from your contact center.

B. Determine the complexity of the use case inquiries:

  • If the customers are calling in for general information, confirmations, and transactional queries, resolve it with a conversational AI-powered virtual voice agent.
  • If the call intent is complex and requires deeper human support, redirect users to the right live agent via conversational IVR.
  • If phone calls could be deflected with proactive support, identify such shortfalls and resolve them via outbound calls, text/SMS, etc

C. Extrapolate the baseline metrics like how long it is taking to resolve customer queries, what is the average waiting time, and how to achieve success for these use cases

D. Analyze how these calls and variations are handled by the agents on a day-to-day basis

Step #4: Design, build, and train conversational engines

Once you have all the answers, start creating conversation workflows that mimic the way agents deal with queries. You need to consider many elements to build a conversational AI engine — speech recognition, NLP/NLU, a dialog manager, and text-to-speech. You can use one out of many DIY platforms such as Dialogflow, Lex, Rasa, etc. to quickly build one. But they have limitations, are expensive, and require specialized talent.

Step #5: Deflect and automate the resolution of customer requests in seconds

Start deflecting transactional, confirmations, and other routine queries. You can leverage conversational AI-powered virtual agents to:

  • Provide end-to-end resolution for select use cases instantly, with zero human intervention, and eliminate the need for creating additional tickets on other channels
  • Focus experienced and expensive agents on sensitive, high-value conversations
  • Instantly scale up phone support capacity while continuing to deliver exceptional customer service

Step #6: Personalize every conversation to the caller

Integrate with your technology stack that enables AI to curate data, pull information in real-time, and deliver tailored responses.

  • CRM systems (Zendesk, Salesforce, etc.)
  • Telephony systems (Amazon Connect, Nice, Genesys, etc.)
  • ERP systems

Step #7: Seamless transfer to human agents based on the issue, sentiment, and customer

Your USP should be how you handle every call in the best possible way. Well, for that, I recommend that you define call transfer scenarios beforehand so that AI can seamlessly hand-off conversations to human agents based on sentiment, your business rules, and customer profile.

Step #8: Scale proactive communication

Proactive communication is one of the most important pillars for scaling customer support. Not convinced? Here are a few stats to get you interested.

  1. Proactive support increases customer retention by 3-5%
  2. For 12 months, being proactive has seen a 20-30% reduction in customer support calls, lowering contact center operation costs by 25%
  3. 87% of customers want to be contacted proactively, in issues related to customer support

You should use AI to send out automated messages or schedule outbound calls to keep your customers updated in ways that save them time and effort. You may keep your customers informed about ETAs, order status, new promotions, payment dates, loyalty points, and subscription renewals.

Many retail companies are using automated outbound calls to notify customers when wishlist products are back in stock or on sale. This also comes useful when product deliveries are delayed due to lockdowns or weather conditions.

Step #9: Engage beyond voice conversations across channels

You should also think about integrating your conversational AI-powered virtual agent with other channels of communication. For instance, use emails and SMS to provide a seamless experience to your customers. So if your customer is calling in to update his address for an upcoming delivery, it is a good idea to send a confirmation email too!

Step #10: Analyze, measure, and optimize

The last step for you is to build a dashboard to monitor the call performance and the ongoing support. It should include metrics related to the use of your virtual agent and the conversations it engages in. Is the call volume going up? Are customers spending more time on calls? What are the most common queries?

Conclusion

Conversational AI over voice is a relatively new and dynamic trend. You need to learn from the ongoing support experience and measure the effectiveness of your conversational AI-powered virtual agents. If you follow these ten steps, you will be well on your path to leverage conversational AI for scaling customer support!

Related articles

AI talking to customer
Voice Automation – 8 key questions on getting started

Most brands see the value in automating customer interactions with Voice AI. The benefits emerge from the following value points:…

Artificial Intelligence
How to Use AI to Predict Churn If You Are An Ad-Driven Digital Publisher

In an increasingly fragmented market, reader loyalty is notoriously difficult to build and maintain. You know that all too well…

Creating Custom Design
AI-Enhancing Your Product | Machine Learning

When it comes to data, we tend to think in dichotomies: quantitative vs qualitative, objective and subjective, messy and curated,…

Ready to get started?

Purchase your first license and see why 1,500,000+ websites globally around the world trust us.