Shooting sparrows with a cannon
Today I have stumbled upon a blog post on dev.to in which the author is talking about how he built a team of AI agents to mimic a company structure: CEO, CFO, Marketing, Legal etc.
I got interested in this because I have also been dabbling with this idea. After all, I am using AI in my work (mainly software engineering).
What sticked out to me was the fact that even with numerous agents and automated marketing the projects barely got any traction. And this seems to highlight a bigger underlying issue with this approach.
You see, AI can help automate finely formalised tasks, the ones where you can describe it as an algorithm. Otherwise you need human in the loop. That’s why chat interfaces for AI has worked so well: there is literally a human in the loop.
The hardest thing about building a business is not building a product, it is finding product-market fit. It is to verify whether people actually want your product and are ready to pay for it. And even after that then need to become aware of it.
There is a saying: “shooting sparrows with a cannon”, which is really applicable to how such AI agents work. We must not forget that these are stochastic algorithms, they essentially predict the likely hood of the outcome.
Problem is, they lack intuition and this significantly undermines their ability to predict human behaviour. And that is what the market is all about.
I do not feel like adding more agents to your workflow will significantly increase productivity. It might, in fact, increase the burden. Instead of 10x engineer you can become 0.1x engineer because now you have introduced more moving parts which you have to control.
All of this is valid, of course, for now. Maybe in the future the agents will be able to do the work more reliably. They probably will.
This reminds me Data from Star Trek and his search for the human equation. He is literally superior to humans in many aspects, but not in every aspect.