As a marketer for a large B2B company, imagine you’re presented with a scalable solution that would improve the efficiency of your customer service, reduce overhead, and could operate 24/7. Or rather, bots. Now, thanks to Facebook launching chatbot support for its messenger service back in April 2016, more companies than ever are embracing the automated customer service platform to reduce some of the human burden on the more tedious elements of their content strategy. Ideally, the technology should be able to do three things: Convincingly mimic pleasant human interaction Solve problems and provide answers as presented by a user Improve in its ability to do the above two points over time In practice today, chatbots are primarily able to do number two. Most approaches involve machine learning (number 3), which is a process by which a chatbot’s program is supposed to optimize over time, given enough input from users. So in terms of fully mimicking human interaction, chatbots today have to be far more limited in scope to minimize the chance that they’re ability to optimize goes off the rails and breaks the program as a whole. The result is often a sort of uncanny valley problem, where chatbots are very close to seeming human, but in their few errors they break the illusion in an uncomfortable way. First, they’ve embraced the idea of using a bot, publically announcing it and making the novelty of the marketing tool a draw for their content strategy, rather than something they try to hide. Some brands are seeing great successes with the technology, however. This is perhaps one of the most powerful moves for brands to take note of, because it makes it possible for companies to actually customize a tool for their own B2B experience—experiences which are often more complex or particular than a standard B2C customer service interaction.
The lives of marketing technology officers, chief marketing technologists, and other martech professionals are often complicated and draining. Whether they’re in control of buying decisions or implementing new marketing tools, the swing between the excitement for a new solution to the drudgery of actually supporting those solutions can take a large toll. From budget burdens to committing their team’s man hours—the payoff for every new implementation has to be worth it.
As a marketer for a large B2B company, imagine you’re presented with a scalable solution that would improve the efficiency of your customer service, reduce overhead, and could operate 24/7. Oh, and it would be entirely in-house—no third-party service or call center that’s likely to muddle your brand image. Sounds like a dream, right?
Only catch: how do you feel about robots? Or rather, bots.
Over the past couple of years, chatbots have become an increasingly popular tool for companies looking to improve their capacity to answer customer questions and requests. Now, thanks to Facebook launching chatbot support for its messenger service back in April 2016, more companies than ever are embracing the automated customer service platform to reduce some of the human burden on the more tedious elements of their content strategy.
Are chatbots the next necessary thing for your marketing mix? Or do they still have a limited use case?
Novelty and the Uncanny Valley
The first important thing to understand about chatbots is what the technology can and cannot do. Ideally, the technology should be able to do three things:
- Convincingly mimic pleasant human interaction
- Solve problems and provide answers as presented by a user
- Improve in its ability to do the above two points over time
In practice today, chatbots are primarily able to do number two.
Chatbots that can hold complex conversations are difficult to produce. Most approaches involve machine learning (number 3), which is a process by which a chatbot’s program is supposed to optimize over time, given enough input from users.
The problem, however, is that people aren’t always pleasant or precise online. And machines are really good at learning. Even the bad stuff.
Microsoft learned this the hard way when they invited Twitter to try and help its chat AI, Tay, learn how to speak. In less than 24 hours, the bot went from being “cool with humans” to spouting racist rhetoric, and was subsequently shut down. This was a natural evolution from a similar problem that IBM’s supercomputer service Watson also faced back in 2013 when it accidentally memorized all…
COMMENTS