As the use of messenger apps have overtaken social apps and more customers want to interact with their brands through this medium, the use case of bots have been on a meteoric climb. It is no longer the case that a customer will phone or go to a brands website in order to interact with the brand about a product, but would rather continue that conversation within their messenger service of choice.
Machine learning and bot agents are creating plenty of opportunities for organisations to provide quicker and better service when their customers need it. Customers use many channels and brands need to be relevant across all channels – bots can be especially useful in providing the kind of agility required to offer a seamless, Omni-channel customer experience.
Consequently, bots are changing the ways that brands support their customers. The most common of the initial applications involve deploying bots to “chat” to customers in call centre environments. Other implementations are more sales- and lifestyle-oriented: bots can be used to make purchases, deliver gifts, make restaurant bookings, and arrange travel itineraries.
One of the big advantages of bots is that they are quick and easy to implement; however, delivering true value and a convincing customer experience remains a challenge.
How do bots work and where does machine learning fit in?
Bots are automated software “agents” that are deployed as part of intelligent, interactive self-service chat solutions. They can be used to solve customer problems, deliver information and provide an increasing range of functions tailored to specific customer experiences and environments.
- Simple bots which evaluate an input and provide a response. They typically have limited functionality and live within a defined set of rules.
- Advanced bots which can accept natural language input and formulate a response. Advanced bots are derived from advanced analytical algorithms better known as Machine Learning that requires sufficient data in order to provide an informed response based on the input from the user.
Depending on the business application and functionality required, one of the above options can be used in order to create a solution which can meet customer needs.
Springbot in Facebook Messenger is a good example. It acts as a shopping concierge, to which users teach their preferences. It then delivers texts or rich media messages about products it has found that match these.
Why is having a bot a good thing?
Customers increasingly expect accuracy: we all have a decreasing tolerance of service mistakes, technology problems and administrative errors. Bots can provide a degree of reliability and accuracy that humans cannot match.
Machine learning and bots will free up human resources to focus on more strategic activities, like customer service planning. More emphasis can be placed on skills diversity, upskilling and training.
Machine learning and bots will free up human resources to focus on more strategic activities, like customer service planning.
Another key benefit is that sales and after-sales service can become much more automated, simultaneously increasing revenue and return on investment.
Why do bots fail?
When bots fail, it’s usually because we overestimate the capabilities of machine learning models, and are seduced by the idea of finally arriving at machine artificial intelligence (AI).
The obvious problem is that bots function in the most challenging area of AI:human conversation. We are a very long way from a world where bots can understand and successfully mimic social and personal nuances.
What should businesses do to leverage bots successfully?
The answer to the challenge of conversational bots is to recognise the issues that can only be addressed by humans, and employ a blended bot/human strategy.
Additionally, many customers still show resistance to the non-human contact of automated mechanisms, so a fair degree of acclimatisation is still required. It is up to companies to virtually take their customers by the hand and help them to become accustomed to the efficiencies that bots introduce when handling the simpler issues.
The crucial approach to take is to view and treat ML and human as part of a team, each bringing specific abilities and strengths. Humans need to guide the bots, in a number of ways, from creating the right tone in scripted responses, to constantly correcting errors to expedite machine learning. More complex queries and issues should be left for human-only response.
It is this combination of human and ML resources that will improve an organisation’s ability to responds to its customers swiftly, intelligently, accurately – and empathetically.