Chatbots are a popular business solution for customer services and many other use cases. But there are some times when a chatbot is not an appropriate way of engaging with customers, so consider these points before adopting a bot for your business or service.

Back in 2017, chatbots had a failure rate of 70%. In 2018, it turned out that “smart” chatbots could be corrupted, like Microsoft Tay. Over 2019, the value of the chatbot market grew 31.6% globally. And, in 2020 they’ve become an essential part of many businesses, with their profile dramatically raised by governments and healthcare using them to highlight COVID advice.

Even so, the combination of the limitations of chatbots and businesses uncertain about how to adopt or use them leaves a sizeable gap, one where chatbot adoption is not advisable or even wise. For all the many good reasons to adopt a chatbot there are a few on the flip side that might help steer you away from specific use cases or routes to adoption.

Don’t let a chatbot do the talking

For some reason, undertakers and fishmongers leap to mind as businesses that shouldn’t ever be involved with chatbots. The first for reasons of taste and decorum, and the second, fresh fish offers should be made live, loading them into a bot seems to go against that nature of freshness.

Of course, that hasn’t stopped some marketing maniacs touting bots to these types of businesses. But, in general, most bots you see today are fit for purpose and there are companies out there (bespoke services or highly personal ones) that should always provide a face-to-face service or other traditional approaches, and they know better than to make such an effort.

Also, there are plenty of companies that are only too happy to fling a chatbot onto a website or platform with no real thought or plan. Any bright spark, IT type or marketing manager should be stopped by the business leadership unless they can present a clear bot strategy, funding and staffing plan to avoid a chatbot failure that languishes unloved by customers and the business, or creates negative sentiment about the company.

Just because a business can launch a chatbot in a matter of days (ideal in a specific emergency, perhaps). It doesn’t mean they should during the normal course of operations, rushing a bot into service is rarely a good idea, expect in the simplest of use cases, say regularly changing opening hours, stock situations or to provide a specific set of limited information.

Common fail points in chatbots that have been launched (and rapidly erased from history) include:

• The business didn’t know what it wanted the bot to.
• There was no way of measuring its success or impact.
• Customers were aware of the bot, so it failed to gain any traction.
• Someone saw another company’s bot and thought “that’s a good idea!”
• The developers used AI for a bot that didn’t need it, overcomplicating things.
• The developers left the bot open to abuse, creating more than just technical problems.
• Machine-learning sounded like something the team could gain an advantage with

Another common problem is that one part of an enterprise will launch its bot using one service or software, and another will use different tools. A global company can soon end up with many bots operating to different standards, with incompatible data for analytics and no ability to combine or track the efforts. Larger companies need to avoid the “shadow IT” situation and get a firm grip on who can launch bots, following a common set of guidelines.

The rise of the customer experience

Out in the real world, end-users expect a good customer experience and the early generations of chatbots were very limited in the information they could give and how they responded to users’ questions. Now that marketing are taking more of an interest in chatbots under the guise of the “customer conversation”, bots are getting better at holding a conversation, but there are still too many stuck in the “I can’t help with that, try asking something else” approach.

One extension of better bots comes from the rise of AI in chatbots as they move from script-based tools to smarter, self- or managed-learning machines. Natural language

processing (NLP) can help bots understand what people are asking about by looking for key words or phrases in a conversation. Better, natural language understanding (NLU) helps categorise questions or can detect the tone of voice of a user.

These can be added to bots relatively easily but require monitoring and training to get the best results and to ensure they provide the right results, rather than creating another set of examples where bots fail to deliver. With these and smart technologies like text-to-voice for services like Siri and Alexa, and translation tools to communicate with people in any language, there are added complexities and a need to understand the risks and benefits before they should be adopted.

Pausing into the future of smart bots and AGI

Another business problem with bots (and many technologies) is that the planners or leaders might see articles about the upcoming leaps in abilities and features. This can lead many to put on hold any plans for a bot that they could successfully launch now.

The glossy promises of artificial general intelligence, powering a bot that could understand and respond to any question, is tempting. But in many cases, that technology will never be appropriate for the typical customer support or information bots, that will happily run on a script or existing AI technology.

Other design issues that can impact bots include the decision to add a level of personality to them. Marketing and customer engagement bots should rightly have a lighter air to them, than those for enterprise sales, some non-profits or healthcare bots. But, this needs to take a measured approach in line with the company’s tone-of-voice and what the customer would expect to see or hear.

For most businesses a bot that “does the job” is more than enough for most use cases. Over-stretching the capabilities or experience of designers, marketing and customer experience teams should be avoided, making bots a journey to take in small steps together, starting with those basic simple bots to avoid the pitfalls that many others have suffered. Take the health bot that turned a nosebleed into a pregnancy diagnosis as just one example.

Whatever your business, there are probably several valid use cases for chatbots, where they save time, speed up customer relations, or save or generate revenue. Most of the obvious uses are out there in the world, so if someone in your business comes up with a novel use for a bot, it should be investigated carefully.

Then, when developing bots, build one that is appropriate to your business and customer needs, and not one that is a too-blunt instrument or one loaded with unwanted bells-and-whistles in the form of complexity that can confuse users or make it harder to establish the actual benefits of the bot. And, under no circumstances, let anyone selling you a chatbot that one for a funeral parlour is in any way a good idea!