Okay, you’ve done the initial analysis and determined that there is an opportunity to reduce your costs AND delight your customers by providing them additional self-service capabilities. Callers like it since they can reduce their wait time for an agent; companies like it since they can service a customer at a much lower cost per call.

But automation takes time, money and skill to create a self-service transaction that works well and provides a good return on investment. If you’re like most companies, you have a myriad of transactions to choose from to determine where to start implementing speech automation. Starting with the wrong tasks can waste resources and alienate your callers.

We always start with several "rules of thumb" to help determine the best automation opportunities.  Not all of these may apply to your business, and sometimes there are trade-offs to consider, but it’s a good way to start to evaluate where you want to end up from both a customer satisfaction perspective as well as from a cost savings perspective.  Here are the first two:

Follow the Traffic:
One of the first items to look at is the amount of agent calls that are handled for each of your transactions; how many balance requests, how many make a payment, how many tech support, etc. (Having an accurate picture of all your transaction types and the traffic they get is important for so many reasons, but we frequently work with clients who don’t have a good handle on this).  Then start with the top of the list – those transactions that account for the most calls – and further evaluate how self service would work for each of these.  The important point here is that it usually does not make much sense to automate functions that get a tiny percent of the traffic – even if it would technically be possible.  You need to budget your development dollars to get the most bang for your buck, at least in the beginning.  If a function gets insignificant traffic, we move it to a “later” list.

How Long is a Typical Call (aka how much time will a caller work in a self-service mode):
This one gets a bit trickier, but the philosophy is that shorter, repetitive calls provide the best opportunities for successful self service. This includes not only time but number of steps. This is not to say you couldn’t create a 15 step 10 minute self service app, but we usually start with shorter repetitive calls. Case in point, “what is my balance”?  Very short, very repetitive – almost everyone automates this if they get balance calls (they even provide balance as an automatically provided message after identifying the caller).  But you see the point: short, repetitive.  Lot’s of transactions can fit in this category like making payments, payment arrangements, cancel or reschedule an appointment, and the list goes on and on.

Again, good data is important to have. If the transaction typically takes an agent over 6-8 minutes, you may want to hold off until a second round of self-service development.

Stay tuned for Part 2 of this post, where we'll take a look at two additional “rules” we use to identify good automation opportunities...