Don't let your Chatbot land you in Court

There is a universal agreement that ChatBot's are totally the wrong way to keep customers. Everyone hates them, and comes to hate the companies who, for whatever reason, believe that a chatbot is the solution for a vibrant and happy customer base.

Let's consider Amazon, a massive company with a massively hated chatbot for support. Its virtually impossible to get past it to someone who can help, and it has the added feature of infuriating customers before they get to a real person.

Ever tried using the 'chatbot' at UPS, Parcelforce, your chosen Airline, or any other public facing company trying to replace real customer support staff with a poorly written, badly implemented software program? You know what I mean then.

In a survey from 2021, the following feedback was provided in response to chatbot interactions:

Customers	Feedback
5.5%	It feels strange
17.2%	It takes too long
11.6%	Had a bad or adverse experience
17.2%	It is annoying
23.4%	It can't answer my questions
5.6%	Concerned about Privacy
11.7%	It is impersonal
7.6%	I don't know how to make it work

Is this something you can to inflict on your customer base?

Programatic or LLM

A well implemented chatbot should be programatic, having a defined path for all possible outcomes, with quick exit points to allow customers to reach a real person if their enquiry falls outside its remit. These are the chatbot's developed by GEN, but more and more companies are trying to use LLM's to save time writing the code, and this presents a massive risk that cannot be understated. You cannot just throw your FAQ at an LLM and expect it to answer customer questions reliably.

GEN have our own experiemental 'customer service' LLM, but its exactly that, experiemental because LLM's are not ready to be placed in front of customers to answer their questions, LLM's whilst invaluable for some internal functions like transcription, or analysis, are still far from perfect when left in charge of your livelihood.

As a direct example of this, asking our resident AI, that is trained on our companies products and services, how much it would cost to get support, it responds confidently that "If you're experiencing issues with your computer, I recommend seeking assistance from a reputable IT specialist or a computer manufacturer's support team." Instead of pointing the customer to the HelpDesk pricing. We can fix this of course, but you cannot test every possible combination of words thrown into it from customers.

VOICE chatbot's are even worse! Barclays Bank, are the proud operators of a voice chatbot that is absolutely tragic both in its design and in its ability to hang up on customers when it get's confused. How a business with the resources of a BANK can produce such a badly designed and pooly implemented service and then unleash it upon their customers is beyond me. We know of at least one customer who moved away from Barclays bacause of this electronic annoyance.

In the News

A recent article in the news, from a well-known Canadian airline, highlighted a court case where their 'chatbot' entered into a contract with a customer, leaving the airline out of pocket. This was based on the fact that a chatbot is indeed acting as an agency of the company, and therefore, anything it says or does can be construed as binding - regardless of how you try to disclaim it. This is because it's acting on behalf of the company with a consumer, had the customer been a business then the laws are different. Citation

Another massively entertaining article reports that DPD's chatbot used bad language, and told a customer that DPD were the worst delivery service. Seriously, you can't make this stuff up - using an LLM (Large Language Model) is the worst idea on the planet for a customer-facing chatbot. Citation

The Future

DO NOT go boldly down the chatbot route unless you have a specific use case that cannot be served with a properly crafted programatic algorithm. Chatbot's are universally hated, yet a professionally developed support platform with its elegant design and fast time to resolution(TTR) are reveered to be the future of customer support.

Blog Index   PermaLink   Comment           13 Votes

Comments (3)