Chatbots not yet ready to replace humans

Read time 4min 30sec
Current chatbots are hard to use and frustrate customers with outright usability failures, say analysts.
Current chatbots are hard to use and frustrate customers with outright usability failures, say analysts.

While chatbots show great promise as a more convenient and natural means of communication between brands and consumers, most don't meet customer expectations.

This is according to a recent study by market analyst firm Forrester, which notes a confluence of factors - from high consumer adoption of messaging platforms, to advancements in artificial intelligence (AI) - has pushed chatbots towards being a viable, though limited, interaction layer between brands and consumers in 2016.

The firm defines a chatbot as a conversational interface - voice, images, or text - that streamlines tasks by allowing users to engage naturally through language.

Another recent study by Tracxn Research shows that over $140 million has been invested in chatbots since 2010, with close to $85 million invested in 2015/2016.

Customer frustration

However, the Forrester report says most bots are not ready to handle the complexities of conversation and still depend on human intervention to succeed.

Forrester says successful chatbots depend on core technology such as natural language processing, AI, and machine learning - aided by thousands of iterations to drive learning.

Julie A Ask, vice-president, principal analyst serving e-business and channel strategy professionals at Forrester, says chatbots are here, but they aren't ready to replace apps or humans.

"Outside of narrow applications, current chatbots are hard to use and frustrate customers with outright usability failures like not setting expectations or acting in unexpected ways," says Ask. "These problems are severe enough that customers must work harder to use a chatbot than an app or Web site to complete a task," she adds.

Nonetheless, she points out chatbots have the potential to improve information distribution, customer care, commerce and marketing over time by making them more natural.

However, multi-featured chatbots that surpass apps in convenience will be complex and depend heavily on AI to interpret consumer intent and reduce consumer burden, Ask explains.

"We aren't there yet. Moreover, chatbots will never be a standalone solution to business challenges. They will be a part of a company's larger portfolio of digital touchpoints."

Early adopters

A number of South African enterprises have already rolled out chatbots. Among the early adopters are Discovery Health and Absa.

Richard Hurst, director of enterprise research at Africa Analysis, says the rise of chatbot technology is being embraced by enterprises across the globe as a means of engaging with customers leveraging technology tools and services such as messaging and big data.

"In South Africa, we have seen some financial service entities deploy chatbot technology, and as this sort of service begins to gain popularity among consumers, we can expect to see other enterprises which have a strong focus on customer service deploying chatbots. The most likely candidates to deploy and become advocates of chatbot technology will be the telecoms sector.

"I doubt that the chabot will ever be able to replace human interaction, but the technology will serve to streamline the human interaction as and when called for. I think that one of the main advantages of the chatbots is that they will empower the consumer to resolve customer service issues at their own time and pace," says Hurst.

Inflection point

Jon Tullett, IDC's research manager for IT services, Africa, notes that every major AI vendor has a chatbot program in active development, and the supporting tech is accelerating too.

He points out that Microsoft just managed to achieve better than human comprehension in machine transcription, for example. "Anyone who thinks conversational AI is 20 years away needs to look at AlphaGo.

"Last year, the confident predictions were that it would take many years for an AI to beat a strong human player at AlphaGo; this year Google's AI team didn't just beat a strong player, it thrashed a world champion. That sudden disparity in expectation versus capability is a sign of an inflection point going by."

According to Jon Tullett, chatbots are not yet widespread in contact centres, but they are going to be. "Dimension Data publishes an annual report into contact centre technology trends, and right now chatbots aren't moving the needle, but when I spoke to them after the last report, they expected it to come. There are definite signs of companies gearing up for it - putting in the underlying automation and technology.

"Don't expect rip-and-replace projects, though," Tullett warns. "Look out for gradual deployment. Supporting agents in Web chat, providing suggested answers which a human approves, applying machine learning to knowledge bases, doing background analysis - that sort of thing."

He points out this provides a runway to ensure safe adoption without risking any degradation to customer experience, and also tackles the obvious labour questions that will come up.

See also