Artificial intelligence and automation are seeing wider adoption than ever, but there are still limitations to how these new technologies can support business processes.
Is there a time on the horizon when humans can turn over their businesses to advanced deep-learning neural networks and relax while artificial intelligence does all the work? The early 2000s saw a boom in machine learning, AI and automation — and despite the advances that have given us chatbots and personalized suggestions for everything from shoes to potential spouses, we humans are still necessary for effective business processes in virtually every type of industry. That's because while artificial intelligence is efficient, it relies on humans to build its intelligence and take on tasks it can't yet handle.
Neural networks have to get the information they use to recognize patterns from humans. We have to decide which data to “feed” the neural networks to train them to handle specific tasks, such as facial recognition or similar products. This is called supervised learning and it's as far as the technology reliably goes for now. Deep-learning technology can do a lot of things it couldn't do a decade ago, according to Roger Parloff at Fortune, but it hasn't evolved enough to let networks sift through raw data on their own to consistently find relevant patterns.
AI is great at using massive amounts of human-selected data, but the knowledge and experience that allow us to select the data is still our domain. And sometimes, our human customers just want to interact with another person. What does this mean for business? For the foreseeable future, machine learning will continue to deliver efficiency that humans can't match, while humans provide the discernment and interpersonal skills that AI cannot. In other words, it's a partnership, not a competition, as demonstrated in a variety of industries.
What chatbots can (and can't) do for customer service operations
Like other AI tools, chatbots have to be trained by human interactions, and in most cases training allows chatbots to offer initial responses to basic customer questions. This frees up human customer service representatives to focus on more complicated customer inquiries. However, it's worth noting that customers may not use chatbots the way retailers and B2B merchants expect them to.
Terena Bell at CMS Wire reported that although many merchants train their chatbots to answer questions about products, US-based customers prefer to deal with humans before they make a purchasing decision. They're more likely to use chatbots after they buy to handle shipping inquiries and updates. This means chatbots need human instruction and need to be used to meet human customers' actual preferences.
Why content curation still needs the human touch
Spotify and other streaming content services generate recommendations for users with their own AI tools, but people want more. Spotify's editors manage several thousand playlists, but even that's not enough. The desire for music recommendations made by humans is so strong that there's now a niche industry of Spotify playlist curators — independent users not paid by the company — who build their followings on other platforms and make money from artists who pay for the possibility of inclusion on their lists. Why would listeners care whether a computer or a person picks their playlists? It's a social phenomenon, motivated by some of the same sense of familiarity, trust, and group belonging that draws social media users to follow influencers.
Why fraud prevention needs AI and human analysts
AI has been an asset to the fraud-prevention industry by making it easy to evaluate large amounts of customer and transaction data very quickly to look for possible fraud. However, because human behavior doesn't always conform to fraud-detection patterns, analysts still play a key role in fraud prevention. For example, wealthy shoppers' orders sometimes raise flags for potential fraud because they, like many attempted fraudulent orders, come from far-flung locations, are for expensive items and may include rush shipping. Effectively reaching out to these customers to verify their orders requires skill and finesse in the moment. Analysts must be able to spot fraud and they must treat each customer with courtesy, keeping in mind that well-heeled customers—the ones merchants most want to win and keep—are likely to stop shopping with the store if their order is rejected by mistake or if the call from the analyst feels like an interrogation.
Part of what keeps fraud prevention reliant on humans' expertise and observational powers is that fraudsters' tactics are always evolving to beat whatever new system merchants put in place to stop them, and the behavior of legitimate customers can be highly variable. These behaviors aren't limited to fraud. People adapt their interactions with businesses to meet their own needs, whether that's avoiding retail chatbot help until after a purchase, seeking out music recommendations from people online, or buying jewelry online while traveling from one country to another.
Human behavior means that learning and pattern recognition within a particular business context may be efficient to a point, but never truly finished. Businesses of all kinds, even those with cutting-edge AI and deep learning tools, still need the reasoning and real-time discernment that only humans can offer.