I've tried doing a little bit of that before but didn't seem work very well.
I could hardly spot a good "pattern" in the jobs posted.
Most of them seemed ad-hoc and custom enough that after looking at a 100 you will still struggle to find a common denominator.
https://blog.parsehub.com/how-we-got-our-first-10-paying-saa...
I believe this post was the genesis of the odesk idea.
SaaS tends to be inexpensive compared to paying a human to do something, even if that human is underpaid, since costs are amortized across many customers. You're paying for something more similar to (CPU time+profit margin) instead of human time. I'd also hasten to add that there are two sided marketplaces where average costs are higher.
The upside is that you can find a probably that people are provably interested in paying money to solve. The downside is that most ideas you'll find are going to be a real grind to solve and are unlikely to be passion projects which is often helpful for success.