Must Read
AI can do both, but the workflows you need to put in place around them can be different .
difference between automating tasks and automating decisions with AI
When you automate a task, it's usually pretty easy to see if the task was done correctly because there is only one clear outcome. If you automate moving data from point A to point B, you check point B and if there is data there, the task was done.
But when you use AI to make a decision, it isn't always clear the right decision was made. And the more AI improves and climbs the decision making hierarchy to higher and higher order decisions, the more difficult it is to tell if the right decision was made.
Let's take an example.
Using AI to auto-scrape a bunch of reports and industry data and automatically suggest new lines of business for your company might be cool. Using AI to analyze and generate more specific projections based on those initial suggestions might be harder to trust.
Using AI to suggest the best of all those options it originally suggested and generate a plan (once we get to that stage of AI) might be even harder to trust.
It's sort of like what you do with new employees.
You initially cross check a lot of their work, but as they prove themselves and you develop trust, you do so less and less.
We've learned to trust algorithms in many areas because their outputs have usually been crips and clear. But as AI moves into more complex decision domains, this won't always be the case.
We will need to audit decisions of algorithms the way we audit decisions from humans. Algorithms will need to build trust with the humans who work with them.
This is an important consideration when setting up good AI workflows, compared to the task based workflows where it was easy to judge whether or not the automation worked. Just something I've been thinking about as I look at what AI is automating next.
Thanks for reading.