Companies must determine why and how Chat Bots can fit alongside their other digital and physical channels to deliver a great customer experience.
The launch by Facebook, Microsoft and others of tools for building and integrating Chat Bots with messaging services has generated a lot of excitement in the latest evolution of automated assistance. But what is their purpose, and where do they best fit in your customer’s journey?
As described in my previous post, financial advisors must bring all the resources available to their firm and its partners to those critical “moments of truth” with their customers. Tools for other sales roles based on personal relationships have both the potential for augmenting the salesperson or replacing them.
Can Chat Bots achieve all this? If not, what place do they have in the Customer Journey?
Chat Bots and the Customer Journey
Chat Bots can benefit customers by being available instantly, anytime, anywhere, being readily accessible from mobile devices. They also satisfy customers’ desire for self-service, giving them a greater feeling of control. For companies, Chat Bots can free up contact centre staff for exception management and more complex customer challenges. They hold out the potential of automating language and sentiment analysis, and adapting for an improved customer experience.
Typical use cases can include helping customers understand their needs, compare products and services, learn how to use or trouble-shoot a product, or understand their account and product usage.
Engaging Chat Bots to actually complete a sale is largely untried, although there is a precedent in consumer applications such as using Amazon’s Echo to re-order goods. In industries like financial services, regulators would require hard evidence before agreeing to take a human out of the customer buying cycle.
The recent advance in linking Bots with Artificial Intelligence (AI) enables a more adaptive experience, leveraging insights gained from firms’ large knowledge base of previous customer interactions. This is a step improvement over rule-based tools such as Interactive Voice Response (IVR), Virtual Assistants and Online FAQs (which require someone to codify the rules).
However, these other touchpoints are not going away anytime soon. Furthermore, customer journeys are rarely linear – customers engage with multiple touchpoints (online/mobile, contact centre, branch/store) at their own choice. A company’s channel mix must accommodate this multi-path journey consistently across all its channels.
A well designed channel architecture would enable each touchpoint to:
• Be aware of previous interactions with the customer,
• Learn from that history (rather than asking the same questions each time),
• Take insights from previous interactions (predictive: Next Best Conversation),
• Feed results of each interaction back so they are available to other touchpoints.
For best results, the customer should be identified at each touchpoint, but for some use cases (e.g. researching a product or learning how to use it) the firm needs to cater for the identified and non-identified customer.
Because Chat Bot interactions are not as pre-determined as the menus/rules/content formulas of other channels, firms need a different approach to governance. A rules engine can be readily inspected by humans to ensure their integrity, and these can be demonstrated to external parties such as auditors and regulators.
A Chat Bot built on an AI method, however, is more of a black box, with logic paths inferred based on ever-changing data. This requires governance professionals to have a much better understanding of how the Bot interprets a customer’s messages, and for fail-safes to be built in to allow for manual intervention. This is not simple if a Bot is designed to interact with hundreds or thousands of customers simultaneously in production.
UBS’s David Bruno notes that bots currently have 40% failure rates (defined as unable to address the customer’s question or to converse adequately). Even as failure rates improve, fail-overs need to be built into the process to allow a call to be transferred to a human channel (e.g. online chat, contact centre).
As Microsoft’s early experience with its
Tay Chat Bot shows, firms need to monitor and be ready to kill or halt a runaway bot. In Microsoft’s case, their latest Chat Bot, Zo, is limited in the subjects it can address.
Chat Bots model the behaviour the firm wants to project to its customers. Like all models, if left unchanged, they will quickly become obsolete, and can frustrate customers with incorrect or outdated information. Supporting a Chat Bot in production entails human monitoring of performance, tuning where required, re-validation of models, and resetting of customer engagement objectives. This requires collaboration between the marketing, customer service, development and analytics teams.
Steps to Success
The Chat Bot space is evolving – accuracy and reliability are improving with better AI techniques and increased volumes of training and testing data. For firms to succeed in this immature environment, and advance up the learning curve, they should:
• Select a use case that adds value to the customer experience, but is within the technology’s current capabilities,
• Provide sufficient training and test data for Chat Bots to learn and improve through both automated and human-guided learning (supervised and unsupervised learning),
• Integrate with other customer touchpoints to allow for a multi-path customer journey,
• Build in fail-safes, and
• Establish a team to continuously tune and improve.
The Bottom Line
To successfully apply Chat Bots to their customer journey, firms must commit to continuous improvement and adaptation as both the capabilities, and their customers’ expectations, evolve.