Conversational AI: Best practices to maximize ROI

With recent advancements in machine learning and natural language processing capabilities, Conversational AI has traveled a long path. Started from chatbots to answer simple FAQs and repetitive queries now organizations are adopting conversation AI to schedule appointments, order processing, and many other use cases.

“By 2022, 70% of white-collar workers will interact with conversational platforms on a daily basis” — Gartner

While there are many success stories of AI transforming the customer experience and research organizations like Gartner is predicting an exciting future for AI still mass scale adoption of AI is a far distant dream. Some early adopters have successfully implemented AI, but these are just a few of best-case scenarios. There are also instances where AI failed to meet industries expectation and the number of these scenarios are greater.

There could be many challenges with deploying and scaling AI solutions like the gap in customer data, lack of strategy and Skills to implement AI, overestimated organization ability, absence of CX governance, or poorly defined use case for AI implementation.

Since the entire point of using AI to improve the customer experience is to help your customers feel valued and bad AI will have the opposite result every time, it’s worthwhile to explore best practices to avoid difficulties with designing, building, and operating AI solution.

Below are some of the best practices to implement conversation AI to transform customer experience and extract maximum business benefits.

Not all customer conversations can be automated hence the first and foremost step is to identify potential candidates for automation. You can start with top reasons why the customer is calling, some of these calls have a linear conversation which doesn’t require creative and complex thinking, we can call them Happy paths. Use these “Happy paths” to implement Conversational AI Journey and any deviation from this happy path will be transferred to live agents.

© Vikas Sharma, (

Though the contact volume on chat and SMS based conversation will grow with time, it is recommended to start conversational AI with voice-first as ROI is more for voice however this is just a recommendation you can decide based on your business needs.

There is a thin line between what should be automated by conversational AI and what should be transferred to live agents. Hence it becomes important to set the perfect “Guardrails” to keep virtual agents in the lane where they will provide an experience as good as or better than a live agent. This requires access to customer data for insight and unique knowledge of your business to establish handling rules at any point in a conversation flow

Handling rules can be decided by mapping call types identified in the first step with how agents are handling these calls. Below are some of the call categories used to set up business rules.

© Vikas Sharma, (

To be effective and deliver Agent like experience its very important that Conversational AI should have access to customer data. This helps Virtual agents to send a personalized message to customers, handle queries to cancel orders, user authentication, track orders.

Moreover, Integration to customers’ personal information is table-stakes for a good CX. Machines can process information quickly and efficiently than humans, in an integrated system virtual agents can process customer recent activities and can predict the reason for customer calls which reduce customer effort. For human agents doing this quickly is not feasible and their responses are mostly reactive. This is the true promise of AI-powered customer service: to analyze and act on large amounts of data in a split second.

However, providing access to customer data is easier said than done, as this needs some additional efforts to keep all data points updated otherwise your virtual agents will have access to outdated data and poor AI.

Conversational AI is just a tool set, it’s organizational culture, and a group of a cross-functional team of CX experts to guide through the transformation to AI which separates successful implementation from poor implementation. Champion Conversational AI team structure will be like

CX consultant: Digs into your business needs to identify if and where the best opportunity exists for conversational AI self-service.

Solutioning and ROI: Analyze contact center data and processes to architect the solution road map for each use case and associated ROI.

Project Manager: Act as the bridge between all cross-functional team and enable flow

Analytics and tuning: Perform ongoing monitoring and analysis of call outcomes and caller behavior to improve the CX until reaching perfection.

QA: Exhaustively tests each customer application to ensure the optimal user experience

THE Conversational AI landscape is growing fast, but it has its own set of challenges in the form of designing, building, and operating AI solutions. Above are some of the best practices based on my experience of delivering AI-based self-service solutions. These are not a silver bullet but will help to improve ROI from AI investment.

I am a customer experience enthusiast with mission to discover customer experience challenges and designing programs that really delight customers.