This is a great point, and I've been thinking about this post and just haven't had time to respond.
I think using retailers is a good example for the point you're making, but I think you'd probably need an individualized analysis across industries to really look at the potential impacts of AI adoption.
Many small caps have good gross margins, but do not have the economies of scale to convert those into great net profit margins. I feel like a massive increase in productivity could certainly flow to the bottom line in these types of scenarios.
Looking even smaller, non-publicly traded mom and pop type businesses are at the bottom of the food chain. They have the least access to good employees, and higher administrative costs to spread over a limited revenue base. These types of businesses have huge tailwinds from tools like AI (just like they did with the internet/cloud computing/ credit card software / the advent the smartphone / etc.) as it can help level the playing field.
As a general matter, employee productivity has been increasing for decades along with higher profit margins. This coincides with the adoption of technology which has assisted the labor force with improved productivity gains, and overall, this has indeed improved profit margins. Just from my own experience dabbling with AI applications, I believe we will probably see a significant increase in employee productivity in a surprising number of fields.
I think it's a reasonable probability that service businesses sprout up to offer small/mid cap/private companies access to information, technology, legal, HR, market research, etc. services that have historically been available to Fortune 500 type companies.
There will certainly be winners and losers, and retailer seems like the great example of an industry that will stop at nothing to compete its profits away. I think Buffett said something along the lines that the only way to make money in a commodity business is to have the structurally lowest cost of production, and that probably applies to retail as well. I would say for industrial or services type business that already have good unit economics, improved productivity is likely to flow at least in part to improving the bottom line.