Think AI for Customer Support is all hype? Think again!

 It would be easy to think the benefits of AI are massively overhyped. Every day there’s another article extolling the future of robots answering customer questions and reducing the need for customer support representatives. But you could be forgiven for thinking this is all hype. So far, results have been less than impressive.

When Facebook tried to launch smart chatbots for Messenger in 2017, they recorded a 70% failure rate. Bots simply weren’t equipped to answer customer questions – meaning 7 out of 10 people that interacted with them were left wanting. However, things have changed drastically in less than a year.

A recent Forrester analysis identified customer support trends that show an increase in self-service portal usage to over 81% among responding American adults. In fact, by 2020 more than 80% of customer service will be conducted without engaging humans. Customers have begun to warm up to the idea of dealing with the bots. CX sensitive brands are also exploring options to incorporate any AI that interfaces directly with customers.

While AI is a long way away from replacing customer support entirely, it is still a useful tool for enhancing your support offerings. Ignoring the potential of AI in customer service might allow your competitors to surpass you. With that in mind, here are four reasons you should still be thinking about AI, even if you’re doubtful of the robot revolution.  

#1 Agents Love the Advantages of AI

Managers are frequently concerned with the impact AI will have on their team. If you make customer support more efficient or outsource it to the robots, are you putting your current team out of a job?

But in reality, agents really like AI because it prevents them from doing robotic tasks that they were forced to do before. Charles Myers, VP Customer at Answer IQ explains that “agents can get burned out and demotivated answering the same repetitive questions.” AI is perfectly suited to repetitive, menial tasks like tagging tickets or surfacing documentation to customers for simple how-to questions.

Aspect’s 2017 survey on the agent perception of chatbots found that 79% of agents feel that handling more complex customer issues improves their skills and offers more opportunities for career growth. Charles says that “implementing automatic answers and routine task automation actually improves the agent experience by giving them more time to deliver amazing experiences at every interaction. This helps empower your team to do what they do best – be human.”

Instead of trying to replace agents with AI, use the machines to make the agent’s work more enjoyable. Sure, they could do all the work themselves, but why would they want to?

#2 AI has Better Data Crunching Power than Humans

Companies are collecting more data than ever on their customers. Product usage, surveys and customer conversations all contain a lot of insight about what our customers want, but it’s difficult for humans to accurately analyze this vast amount of unstructured data.

Getting customer support a seat at the product table requires quantitative data. AI and machine learning can derive quantitative data from the qualitative – much faster than humans can. AI can also find the patterns that your agents didn’t even think to look for. Because each agent is only seeing a small slice of the total number of customer conversations, it’s impossible for them to determine if the questions they are answering are one-offs or symptoms of a much bigger issue.

Tools like Idiomatic and Scope.AI work with customer support and product teams to analyze the voice of customer (VOC) data across multiple channels. “Manual reviews of VOC data yields anecdotal and inaccurate views of your customer experience, and they take a lot of time.” says Idiomatic co-founder Christopher Martinez. “You can only improve what you measure, and measuring enormous amounts of data is something best left to computers.” While robots are still pretty terrible at talking to customers, they excel when it comes to data.  

#3 AI Boosts Efficiency Without Impacting Quality

Even if your brand is customer experience sensitive and firmly against the idea of AI talking directly to customers, AI can still lend a capable hand behind the scenes. Using natural language processing, AI can “read” a ticket and direct it to the right team much faster than a human triage system can.

For example, Uber built COTA (Customer Obsessed Ticket Assistant) to help route tickets better and suggest answers to customer support agents. They found that better ticket routing increased efficiency by 10%. Plus, measuring customer satisfaction through surveys, they found that CSAT stayed consistent or improved through the implementation: “By empowering customer support agents to deliver quicker and more accurate solutions, COTA’s powerful ML models make the Uber support experience more enjoyable.” By not allowing the AI to talk directly to customers, Uber gets all of the benefits of AI but reduces the risk of a terrible customer experience.

#4 Even Smaller Companies Can Benefit from AI

Much of the hype around AI has been driven by big companies like Chinese banks crunching millions of requests every day, or even Uber’s example above.

But the technology is becoming more accessible to every company, and smaller companies can see an advantage from AI too. If you’re a smaller company thinking about AI, look for something that’s very easy to set up (essentially plug and play) and integrate into your existing workflows.

For example, tools like Solvvy and MonkeyLearn don’t require the enormous amounts of data that some other platforms ask for (or come with the associated costs). But you’ll see an immediate benefit from automating some small part of your customer support workflow:

– Solvvy aids customers help themselves faster by surfacing the best knowledge base articles for their question. They estimate that their customers see an average self-service rate (the percentage of customers that are able to resolve their own questions) of 20%.

MonkeyLearn automatically tags tickets or identifies customer sentiment for better prioritization – much like Uber’s in-house product. But the good news is that you only need between 50-100 historical tickets for each tag you want to analyze in order to get started. (That’s not a typo!). Tagging more tickets will improve the accuracy of the AI, but depending on the complexity or specificity of your customer conversations, that might be enough for your use case. MonkeyLearn can also analyze other text samples from across the internet (like social, product reviews or app reviews) to glean insights from what your customers are saying.


Ultimately, the companies that are able to implement AI earlier will have more breathing room to focus on providing amazing customer service, and more insights to focus on. In fact, even if you don’t believe the hype, you can’t afford to ignore it completely.