How to launch an efficient bot?

The Wow effect has passed, and bots have plunged into the Gartner Hype Cycle’s “trough of disillusionment” [1]. Their intelligence could stand some improvement. They are not replacing the customer service agents. They do not provide acceptable service except when supported by a considerable amount of human input. Nonetheless, bots promise to revolutionize the customer experience, putting them at the heart of any digital transformation strategy. So what makes them an essential tool? How can new players launch an efficient bot, avoiding the pitfalls and optimizing their customer relationship?

A must-have for digital customer relations

Software programmed to simulate a conversation in natural language, bots are spreading across brand websites, social media and messaging. They respond to customer requests in writing or verbally, assisting or advising them.

Their benefits can only convince operators to get started:

• First of all, they improve the customer experience: they provide an interface that is easy and fun to use. They combine multiple services, save customers’ time (24/7 availability, no waiting), and provide them with personalized and instant responses.
• On the other hand, bots improve operational efficiency: companies can delegate to the bot any answers to recurring and easy-to-handle requests, so as to focus its human interfaces on complex (disputing an invoice) or risky (loss of a customer, etc.) topics.

Analysis of the current market confirms a rise in the attractiveness of bots. 7 clients out of 10 prefer to talk with a bot to save time [2]. In 2020, 25% of customer service interactions are expected to occur via chatbots [3].

The feedback from operators today is positive:

• In Spain, Aura has made it possible for Telefonica to process 22% more requests from prepaid customers. 70% of these customers’ purchases were recommended via the bot.
• In the UK, Vodafone’s TOBi recorded a conversion rate 100% higher than that of a standard website and cut transaction time by 50% [4].

6 challenges to take up in order to create an efficient bot

The simplicity of deploying a bot platform, much touted by solution providers, is not always synonymous with better quality of service for the user. There is a significant risk of negatively impacting the customer experience when the familiar web interface is stripped away: customers will place much higher demands on the robot with whom interaction will be compared to that with a human.

Bot design, deployment and operation must therefore be well thought out and well organized. The road to achieving this is made up of six mandatory stages.

1. Delineate the subject

First of all, define the use cases to be covered (e-shop, TV, invoice, Internet troubleshooting, assistance, customer service, internal HR manager, etc.) and the functionalities of the bot. Advantages: give visibility to the project, establish its legitimacy and unite around it.

2. Anticipate the obstacles in order to remove them

These obstacles can be

• organizational: recruitment/training of project team members;
• budget: the budget must cover the construction as well as the continuous operation of the tool;
• technological: is the IS sufficiently structured and mature?;
• or functional, for instance: has the content of the chat tree structures already been defined?

3. Choose your solution from three operational options offering different performance

• The searchbot (aggregator of existing content). Easy to access and addressing numerous uses, it allows operators to ramp up their expertise while improving the customer experience.
• The chatbot (conversational agent). As it delivers more precise and personalized responses, the customer’s experience is better, but it also requires a mature organization and technology.
• Metabot/virtual assistant (bot orchestrator). This complex solution, reserved for large operators, is the most advanced today.

4. Implement the solution using an iterative and incremental approach

By starting up gradually, new players will minimize the risk of disappointing their customers. I recommend defining and targeting only the essential part of the scenario to be tested – the infamous “MVP” (Minimum Viable Product) – on a limited population.This scope should be optimized through successive iterations, then gradually expanded.

5. Put an operating team to work

This team, ideally organized in agile mode, must include at least 5 roles.

 

1 project manager	Responsible for budget, priority-setting, reporting, etc.
1 technical advisor	Guarantees the solution’s stability In charge of technical developments
1 cognitive trainer	Defines decision trees in the construction phase, optimizes them via analysis of conversations in the operation phase
1 UX writer	Draws up all responses to the customer
1 settings officer	Incorporates tree structures and lexical field in the back office

Typical bot operation team profile

6. Organize continuous and dynamic learning of the bot over time

It is essential that the bot continuously improve. It must be trained and take on new functionalities. 60 intentions will cover 90% of customer requests, but completion of the remaining 10% is achieved through the implementation of 260 additional intentions.

Securing a budget for continuous improvement is thus of the essence.

How to optimize the customer experience

Advances in R&D on Artificial Intelligence, automatic natural language processing, supervised learning and deep learning will eventually make bots more intelligent and autonomous (self-learners). They will free up time for humans and bring innovative use cases to the fore. Until this revolution comes about, the available solutions already offer the operator the opportunity to provide answers to the strategic challenges raised by its customer relationship.

By making it possible to carry out end-to-end pathways

With an ActionBot solution integrating multiple services in the same chatbot, the operator offers its customers an end-to-end experience: the chatbot understands the user’s need (e.g. troubleshooting a box). Instead of directing him/her to the box diagnostic application, which will force the customer out of the chatbot, it performs the task itself and informs the customer of the result.Note: the incorporation of API into the operator’s services is necessary so that the chatbot can directly step into the service provision process.

By ensuring continuity between the bot and the advisers

It is essential that the bot be able to establish a link between customers and agents:

• either with a strategy for qualifying customer requests before automatically transferring them to an agent,
• or by allowing an agent to pick up the conversation when the bot is unable to respond to the customer’s request.

In both cases, context transfer must be provided for, so that customers do not have to explain their problem again, and instead have their request addressed more quickly.

By breaking down bots’ functional scopes thanks to the metabot

The metabot makes it possible for a multiservice operator to converge the customer experience within a single interface. It is able to respond to customer requests on all subjects: bank account, mobile services, TV, etc.

Bots offer operators the opportunity to improve their customer proximity through a personalized and omnichannel relationship. Complementary to client advisors, bots are not intended to replace them, even if they will profoundly transform the nature of their responsibilities. Relieving advisors of low value-added tasks will require developing their skills in technical assistance, data processing and active listening of customers. The digital transformation will therefore require first and foremost a human transformation. It is by selecting the technology suited to the maturity of each organization and defining ambitions correlated with their own learning curve that operators will achieve their objectives for improving customer satisfaction. So, what are you waiting for?

Sources:
[1] Bryan, J. (2018) 4 Trends in Gartner Hype Cycle for Customer Service and Customer Engagement, www.gartner.com. Available at: https://www.gartner.com/smarterwithgartner/4-trends-gartner-hype-cycle-customer-service-customer-engagement/.
[2] 2017 Chatbot Report (2017). Available at: https://www.ubisend.com/insights/2017-chatbot-report.
[3] Moore, S. (2018) Gartner Says 25 Percent of Customer Service Operations Will Use Virtual Customer Assistants by 2020. Available at: https://www.gartner.com/en/newsroom/press-releases/2018-02-19-gartner-says-25-percent-of-customer-service-operations-will-use-virtual-customer-assistants-by-2020
[1] Harrison, R. (2019) ‘AI  & amp; Digital Assistants: Telco Proposals  &  Revenue’. Mobile Market Development. Available at: https://mobilemarketdevelopment.com/maximising-wireless-profit/report/ai-digital-assistants-telco-proposals-revenues/.