We have reached a point in terms of raw processing power and access to tools where automation is within the grasp of more and more organizations.You no longer need a huge budget to be able to automate certain parts of your business. Just take a look at this video of Kiva robots being used by Amazon.com to improve their fulfillment process. Pretty cool, huh?
You can apply the same concept to other industries, not just order fulfillment, or warehousing. You can apply this to contact centers (be it live chat or via the telephone), invoicing, client services and so much more.
For example, a large number of contact centers are setup where supervisors go over chat transcripts and proactively monitor conversations to make sure quality targets are achieved. In principle, this is a great thing to do. Trainees get a supervisor that is close by, and clients get better support. Unfortunately, this type of model starts to breakdown the larger a support organization becomes. For example, if you have a ratio of 5 representatives to 1 supervisor, then it essentially means you are paying somebody to monitor client interactions from five different sources. The supervisor will basically jump from one person to another, listening in, giving advice, and so on. The supervisor will also be spending a large amount of time monitoring interactions that do not need to be monitored. This is dead time. It is unproductive. And you are paying for it. So let's look at the alternative.
Instead of hiring a supervisor for every X number of team members, it would be more effective to come up with an algorithm that proactively monitors client interactions, and flags them to a supervisor.
This algorithm can flag interactions based on the following pieces of information:
- response times (both for client and the team member)
- linguistic cues
- detecting special keywords
- the duration of the interaction
The advantages of this method are as follows:
- better quality through prevention - by being proactive the algorithm can alert supervisors when a client interaction might be be heading the wrong way. This will allow the supervisor to step in at the right time and prevent a situation from getting worse.
- cost savings- you can have less supervisors monitoring a larger number of team members.
- scalability - you can increase the number of team members handling client interactions. The algorithm will work just fine monitoring one team member or 100 team members. Just throw some more processing power at it!
- always on - the algo does not sleep, and does not eat. It will keep doing its job and do it without mistakes.
If you have not dabbled in automation then my question to you is: what are you waiting for? Processing power is cheap (just get a couple of EC2 instances), there are plenty of free and relatively cheap libraries for searching, text processing, OCR (optical character recognition) and so on. If you really do not know where to start, then you can take a small step and look at implementing a leaderboard. In case you need a couple of good reasons to use a leaderboard then just read my article called Top Three Reasons To Use A Leaderboard In Your Organization.
What are your thoughts on using algorithms to automatically parse client interactions and flag them to supervisors? Let me know by commenting in the box below!