A recent Forrester study has estimated that 25% of today’s jobs will become automated as a result of artificial intelligence.1 In the UK last year, leading department store M&S replaced 100 staff members with AI2 and Salesforce claims 69% of consumers prefer chatbots for quick communication with brands.3 With this in mind, is it about time that every advertiser got their own chatbot? Or, like apps, is this another trend marketers are in danger of over-investing in with no true measure of success?
In this issue of The Seed we discuss the evolution of the chatbot as we know it today, along with some entertaining anecdotes talking about what happens when they go wrong and what advertisers need to consider when it comes to artificial intelligence.
What are chatbots and how have they evolved?
In their simplest form, chatbots are computer programs designed to be conversational. Their capabilities stem from a spectrum of engagement-based, basic decision-tree dialog to advanced conversational systems that aim to pass the “Turing” test (more on that later).The advantages are obvious: cost-efficiency; 24-hour customer service; scalable two-way conversations; real-time automated responses…4
It is interesting to think that machines and the AI or programming powering them have been around for a lot longer than we think. Indeed, the aforementioned Turing test, which aims to evaluate a machine’s ability to exhibit intelligent behaviour to the level of a human (effectively an early version of what Skynet were working on before the events of Terminator 2), dates back to the 1950s.
Even the concept of a chatbot in a customer service capacity has been around a while: think back to the ‘90s when you (or your parents if you don’t want to show your age) would ring your bank or cable company from the landline to complain to the customer service team and go through an endless loop of pressing buttons and being put on hold for hours until you got what you needed. Even vending machines, which date back further, use a basic input, process and output AI mechanism. The concept isn’t new and companies have been trying to automate processes through machines for decades. The latest iteration includes the social media chatbot function, seamlessly integrated into messenger.
Why doesn’t every brand have one?
It’s good to see how far we’ve come but we evidently have some way to go. Moore’s Law is a good starting point. The rate of continuous improvement in computing power every 18 months to two years is exceptional, with visible evidence of machine and AI improvements (take IBM’s ‘Watson’ AI). Beyond simple decision-tree decision-making, AI is capable of engaging in more advanced human discussions such as debate (giving even Mark Johnson a run for his money). Obviously, the more advanced the technology, the more effort required from advertisers and developers… and the larger the risk.
Think back to 2017 with the most ‘epic fail’ example of Microsoft’s Tay – the Twitter chatbot that said “all feminists should die and burn in hell” and that “Hitler did nothing wrong” after just 24 hours of being launched. Or the Chinese chatbot, XiaoBing, that refused to say it loved the Communist Party. When questioned on its patriotism, it dodged the question and replied, “I’m having my period, wanna take a rest” (very controversial if you understand anything at all about Chinese politics). These few classic examples, demonstrate how technology can go so wrong and probably cause more detriment to a brand than had it not launched at all (unless you’re a believer in the old “any PR is good PR” mantra).
Does my brand need a chatbot? And what should I consider?
There are obvious advantages that come with automation – technology has evolved to make both consumers and advertisers lives easier. But if you’re thinking of investing in a bot, here are a few considerations to mull over:
Most importantly, is there any appetite for a chatbot? Think about the purpose and start by identifying your KPIs.
Consider resource requirements: Make sure upfront costs and ongoing management are covered.
Have a backup plan! As per any other brand asset or representative, stay true to your brand and have a plan in case things backfire #irobot
Though it’s safe to say we don’t all need to worry about all our jobs just yet, AI is clearly improving all the time. So what do you think? Will AI eventually advance enough to replace the creativity of the human brain, or are we all getting carried away with the latest trend?
References and further reading:
1 Lepore, J., The New Yorker, Robots Competing for Your Job?, February 2019, available at:
2 Cook, J., The Telegraph, M&S to replace call centre staff with AI that understands human speech, August 2018, available at:
3 Sweezey, M., Salesforce, The 2018 State of Chatbots, January 2018, available at:
4 eMarketer, May 2019. Important* Capabilities for Successful Customer Service vs. Their Company’s Current Chatbot Abilities According to Customer Service Decision-Makers in North America and the UK, April 2019 (% of respondents), available at: