Great breakdown on margins here. The biggest illusion I see with AI business models right now is treating API calls like traditional SaaS hosting costs when It’s a totally different beast.
I run my own stack (Hermes/Obsidian) on Hetzner specifically to avoid getting bled dry by the market behemoths. That way I can just switch models on the fly and still have my whole business logic in a single place, basically a single source of truth.
The Anjeanette Carter line is the whole thesis in one sentence and its the insight most AI commentary misses entirely. "If you dont have expertise, AI isnt going to make it better. Its only going to amplify your limited knowledge." thats the clearest articulation of why AI expands the gap between good operators and average ones rather than closing it.
The margin expansion story is the one that matters most for valuation. A services business doing $500K in revenue with a 15% margin is worth a very different multiple than the same business doing $500K at 45% margin. AI didnt change the top line in that example. It changed what the business keeps, which changes the enterprise value, the reinvestment capacity, and how long the founder can grow before needing outside capital.
The second order effect is the one Im most curious about though. when every expert operator can run a 5-person output on a 1-person cost base, the number of viable competitors in any given niche increases. more people can afford to compete because the fixed cost floor dropped. margins expand in year one as early adopters reduce their cost base. margins compress in year three as the new cost structure becomes the baseline expectation and pricing adjusts downward.
The call centre example from Ben Tasker is the cautionary tale that belongs in every board presentation. fired the staff, hired them back weeks later. thats what happens when you optimise for cost reduction without understanding which parts of the process required human judgment that the model cant replicate. the savings look real on the spreadsheet right up until the customer complaints arrive.
The founders who win this transition are the ones in your interviews who already knew what good looked like before AI arrived. they just couldnt produce enough of it at the old cost structure. now they can. thats a genuine unlock. the founders who didnt know what good looked like are just producing mediocrity faster and more cheaply, which is a different kind of competitive disadvantage.
That Anjeanette line did a lot of the heavy lifting for me too.
AI seems to widen the gap more than close it. If you already know what good looks like, you can do more of it. If you don’t, you just get faster at average.
Her entire episode is definitely worth a watch/listen.
The margin story is always more durable than the revenue story.
Top line can look exciting for a while.
Margin is what tells you whether the business is actually getting stronger.
Great breakdown on margins here. The biggest illusion I see with AI business models right now is treating API calls like traditional SaaS hosting costs when It’s a totally different beast.
I run my own stack (Hermes/Obsidian) on Hetzner specifically to avoid getting bled dry by the market behemoths. That way I can just switch models on the fly and still have my whole business logic in a single place, basically a single source of truth.
That’s a very real point, Dan.
API spend can look harmless at the start, then turn into the thing quietly eating the business once usage grows.
Keeping your own stack flexible feels a lot healthier than building around someone else’s pricing mood swings.
The Anjeanette Carter line is the whole thesis in one sentence and its the insight most AI commentary misses entirely. "If you dont have expertise, AI isnt going to make it better. Its only going to amplify your limited knowledge." thats the clearest articulation of why AI expands the gap between good operators and average ones rather than closing it.
The margin expansion story is the one that matters most for valuation. A services business doing $500K in revenue with a 15% margin is worth a very different multiple than the same business doing $500K at 45% margin. AI didnt change the top line in that example. It changed what the business keeps, which changes the enterprise value, the reinvestment capacity, and how long the founder can grow before needing outside capital.
The second order effect is the one Im most curious about though. when every expert operator can run a 5-person output on a 1-person cost base, the number of viable competitors in any given niche increases. more people can afford to compete because the fixed cost floor dropped. margins expand in year one as early adopters reduce their cost base. margins compress in year three as the new cost structure becomes the baseline expectation and pricing adjusts downward.
The call centre example from Ben Tasker is the cautionary tale that belongs in every board presentation. fired the staff, hired them back weeks later. thats what happens when you optimise for cost reduction without understanding which parts of the process required human judgment that the model cant replicate. the savings look real on the spreadsheet right up until the customer complaints arrive.
The founders who win this transition are the ones in your interviews who already knew what good looked like before AI arrived. they just couldnt produce enough of it at the old cost structure. now they can. thats a genuine unlock. the founders who didnt know what good looked like are just producing mediocrity faster and more cheaply, which is a different kind of competitive disadvantage.
That Anjeanette line did a lot of the heavy lifting for me too.
AI seems to widen the gap more than close it. If you already know what good looks like, you can do more of it. If you don’t, you just get faster at average.
Her entire episode is definitely worth a watch/listen.