2017-07-13 HRExaminer limits of ai by john sumser photo img cc0 via pexels photo 199215 abstract penalty penalties 544x331px.jpg
If you listen ever so closely, you’ll hear the rumble of an ongoing argument. Are machines smarter than people? Can they ever be smarter than people? Will that happen soon? What about my job?

Machines are great at a very important thing: the precise execution of repeatable tasks with uninterrupted consistency over infinite periods of time. Handling task variation is something that humans are better at. Humans handle abrupt variation exquisitely. Machines must be retrained in slow, painful increments. Every time the organization pivots, the machine’s effectiveness will decline until it learns the new parameters unless a human intervenes.

If the change is gradual and happens in an even flow over time, machines are okay at it. If it happens abruptly, they must be retrained. Machines are better at the very things that people are bad at and vice versa. 

The differences are amplified in B2B settings. When ML is applied to a consumer problem, millions of transactions per day shape the machine’s adjustment. While the incremental pace of adjustment is the same on a per transaction basis, the sheer volume speeds up the elapsed response time. When it’s 10s of thousands of employees responding over the course of a month or two, the quality of the response will appear to be much lower. 

One of the biggest complaints executives have about their complex workforce is that it is so slow to change. The leadership dictates a change in strategy but the organization itself moves like an aircraft carrier or cargo ship. The adjustment is painfully slow. While individuals may embrace the change at light speed, the organization preserves the status quo.

So do machines. To the extent that they’ve ‘learned’ something, it takes a nearly equal set of transactions to fully unlearn or relearn the topic.

This is the actual cost of ownership of machine led decision making. There will always be some level of variance between the reality of the organization and what the machine thinks is going on. The machine learns from evidence. The organization produces the evidence. 

When there is a change, the lag time means that decisions driven by the machine will be less effective.


Read previous post:
HRExaminer Radio Executive Conversations Badge Podcast Logo
HRExaminer Radio – Executive Conversations: Episode #245: Jill Witty

John Sumser talks with Jill Witty, the Vice President of Talent & Operations at Entelo. Witty is a speaker at...