New research finds human validation is critical for chatbot owners

Most chatbot owners are not satisfied with the performance of their chatbots and say human validation is needed to reduce errors.
Written by Eileen Brown, Contributor

Video: Google aims to embed AI-powered Smart Reply in every major chat platform

Although chatbot technology has been around for a while, many businesses do not include human validation to enhance chatbot interaction -- and consumers are unhappy with chatbot performance.

Read also: As AI floods the market, which chatbots deliver the best ROI

In a survey conducted in late 2017 by Chatbots.org, 53 percent of 3,000 consumers, who had used a chatbot for customer service in the last year, found chatbots to be "not effective" or only "somewhat effective."

US consumers were far harsher in their assessment of chatbots, with 14 percent rating them as not effective versus only 5 percent of UK consumers.

This is perhaps an indication that US consumers ask chatbots more complex questions than UK consumers.

UK-based text analytics specialists Warwick Analytics recently carried out a survey of over 500 chatbot owners and developers. Its findings showed that 59 percent of businesses that have a chatbot are unsatisfied with its performance.

The company provides machine learning technology to help maintain and improve chatbots using a human-in-the-loop platform called PrediCX, accessible via an API.

Lack of integration with human-assisted service was considered to be the biggest pain point in using virtual assistants.

Chatbot deployments that are disconnected from agent-assisted touch points caused major issues with consumers.

Fifty-nine percent (62 percent in the US and 55 percent in the UK) found that having to repeat information and context to a human agent in the event of escalation from chatbots was the biggest hassle.

The survey highlighted that the most common technical challenges chatbot owners faced were: Improving containment rates (90 percent), reducing errors (83 percent), and developing the responses for the chatbot (79 percent).

In addition, 21 percent of respondents who were yet to deploy a chatbot said it was because the performance of chatbots was not acceptable in their opinion.

The software aims to minimise the human input required to improve the comprehension rate and containment rates of chatbots.

It also analyses the topics of the interactions with customers in order to route the customers to the right department of human agents, or identify new topic areas for a chatbot response.

More significantly, an overwhelming 93 percent of respondents believed that human validation and/or curation was important to maintain and improve the performance of their chatbots.

Achieving the right level of human input is vital for chatbot owners and managers to ensure continued engagement from consumers.

Although chatbots are maturing across industries, there is still a way to go. Human intervention is still overwhelmingly viewed as a necessity for improving and maintaining chatbots.

Around 21 percent of respondents who had not yet deployed a chatbot said it was because the performance of chatbots was not acceptable in their opinion.

Dan Somers, CEO of Warwick Analytics, said: "Human validation is required for accuracy and improvement but if too much is required then a business may as well have a human service desk. It's all about finding the right technology that minimises the human intervention required but still increases accuracy."

Read also: Although chatbot use rises, we still prefer talking to humans

Successful chatbot implementations will get the human -machine blend right and provide a seamless performance across all media. Unfortunately, businesses -- and emerging AI chatbot technologies -- still have a long way to go.

Previous and related coverage

How will we cope with the AI chatbot takeover?

Ordering lunch for a group or office using an AI chatbot, searching for recipes online, or using a virtual assistant to control your home with AI. AI chatbots are becoming indispensable - whether we like it or not.

How ubiquitous AI will permeate everything we do - without our knowledge

Most of us do not know that we are using chatbots to talk to service agents, so how will we know that AI will be seamlessly interacting in with our future lives?

Editorial standards