Predictive Coding Is Like Self-Checkout

There are brilliant, talented people with more technology-assisted review expertise than I. There are programmers, analysts, attorneys and other professionals working diligently to formulate and improve the processes by which machines can be utilized to automate the analysis and review of electronic documents in a world where, according to IBM, we create 2.5 quintillion bytes of data daily. This topic triggers debate among thought leaders, inspires a host of studies and publications and permeates panels at industry conferences worldwide.

To all of the wise men and women working towards predictive coding nirvana: all apologies.

I am a "regular" guy and because of that I seek simple explanations of complex problems. So here is one: Predictive coding is like self-checkout.

And here is why: During a particular visit to the store you decide to take an automated adventure and use the self-checkout machine. Things go smoothly until a few items from completion you notice a price rings up differently than it was marked in the aisle... wait, that's not possible, a machine can't be wrong, can it? Can you fix this? You start pressing buttons... the screen starts scolding you... customers behind you start scoffing... red lights start flashing... pandemonium!

Before you have time to bail in an all-out sprint to your car, a calm clerk walks up, inserts a key, calms the machine and allows you to go back to scanning items. This kind soul also lets you know that your sale price will display correctly upon advancing to the payment screen. Problem solved... for now.

What I am trying to say is that from my perspective, predictive coding is like self-checkout and document reviewers are the human clerks. Both can be employed to arrive at comparable outcomes, but despite advancements, the technology does not represent a "set it and forget" innovation. The machines represent a reliable alternative, although not a miracle cure provided there is proper implementation and accurate input. This includes keeping an expert(s) involved in order to alleviate confusion and resolve issues because you still need to scan items (review docs) and monitor the process (QC) throughout.