Headlines featuring “AI” and “chatbot” have dominated the discussion in the customer service space for the past year. Naturally, it’s tantalizing to imagine a world where no one needs to call a contact center, but the sizzle around AI and chatbots obscures a deeper question: How much customer support can actually be done by self-service? This question has been with us since the earliest web sites and IVRs started to offload the simplest tasks from agents. Each new wave of technology revives the dream that we can eliminate (or drastically reduce) human agents. We can see the current excitement around AI and chatbots as an extension of this long-running dream. Is it different this time? Are we reaching the promised land? Or are we chasing a receding horizon?
Self-Service vs. Human-Assisted
The best way to think about customer service is to divide the whole universe of possible interactions into two realms: Those that are self-serve (e.g. “Where’s my shipment?”) and those that require human assistance (“I don’t understand my bill”.) Within those realms, you can have further division into channels and strategies. For example, a self-serve interaction that may have been done over the web last year may now be available via chatbots, or human assistance may be done over SMS instead of voice.
Such changes and improvements within each realm are common. Movements from the human realm to the self-serve realm is harder. Most interactions that lend themselves easily to self-service are already in that realm. The low-hanging fruit was picked long ago.
With this in mind, one can read “AI/chatbot” stories more critically. For example, if a company is boasting about banking-by-chatbot, ask yourself how much of that interaction was previously done over their website, and thus already in the self-serve realm? (Note that there still might be value in switching from web-based to bot-based banking: increased engagement, higher containment rate, or simply catering to certain consumers.) What’s more impressive, though, are projects that bring “net new” interactions into the self-serve realm. One such example is the sommelier-in-a-bot launched by Lidl. Hats off to the team at Aspect.
Reaching the Promised Land?
Restating the primary question: Will the current wave of bots and AI-powered solutions actually move new interactions into the self-serve realm, or are we just introducing new form factors for interactions that are already self-serve?
It’s possible that we are at, or near, a kind of saturation, where everything that can be “self-serv-ized” has been. I talked more about this here: Will We Always Need Call Centers?
But let’s assume for a moment that “this time it’s different”, and we are indeed entering an era where more and more interactions are taken off the plate of the call center. What would the impact be? Analyst Ian Jacobs explored this topic in his recent post, Plan Now for the Contact Center Agent of the Future.
We’re looking at a snowball rolling downhill. Automation could soon easily handle one-third, or even more, of customers’ problems…. If you have hired your agents to do L0 or L1 work, but now need them to do L2 or L3 tasks, you’ve got a real problem. You could try to “upskill” those agents, a dubious prospect if you have not specifically recruited people with the ability to quickly learn and improve… And the skills deficit will only grow…Few companies today are hiring agents with those skills in mind…
(In Ian’s language, the “L0” work is the stuff that can be fully self-serve, what I was calling the “self-service realm” earlier in this post.)
Chasing the Horizon?
Ian’s prediction and warning is logical (and good news for the agent labor market!), but it rests on an assumption that chatbots can indeed take over the L1 tasks. That’s not a given. What we’ve been living through in the past decade is the removal of L0 from the call center, as self-serve has gotten better and more popular. It follows that this decline will bottom out at some point, but it doesn’t follow that L1 and L2 are next.
It may be that the net result of chatbots, even say 5 years from now, is that they’ve displaced other forms of self-service (like IVRs) and have boosted completion and participation rates; but, at the same time, not moved the line between what can be self-serve and what needs a human.
We hope AI will change this but, so far, this is just conjecture. The skills that current AI tech is good at (“Is there a car in this picture?”) don’t really map onto to the skills needed to move from L1 to L2 (“Here are all the edge cases in booking a flight.”).
A Growing Number of Agents
So, is new technology lowering the amount of phone calls? Or lowering demand for agents? As important as these questions are, it is hard to get solid answers. One useful measurement we have is the total number of people employed as call center agents (which we can get from the Bureau of Labor Statistics). Here we see a steady growth going back at least 5 years. (For some of the caveats around this data see this post.)
Regarding the agent labor market, Ian writes:
Start investigating new labor models that could help you attract more talented team members, such as work-at-home models… Start noodling on which skills are actually required for this new type of work and how you would test for those.
In fact, it seems like the future is bright for this form of employment. Not only is the quantity going up, but the skills requirement is going up too. That should lead to higher wages and less monotonous work.