The Labor Substitution Mirage
June 23, 2025
In one of my previous posts I argued why moving to outcome-based pricing is difficult despite being very prominent in the zeitgeist. But there's an even deeper issue lurking beneath the surface: the long-term premise behind outcome-based pricing might be a mirage.
The logic seems bulletproof at first glance. AI companies pitch massive TAMs by targeting labor spend: why charge $100/month when you can replace a $100 K employee? Position your product as a head-count replacement instead of just another tool, and suddenly you can justify pricing higher than you ever imagined.
But some of this logic falls apart when everyone follows this same playbook.
Why Labor Substitution Feels Like a No-Brainer
So many companies these days are pitching their TAM in terms of % of total labor spend in a certain space, and for good reason. Dollars spent on labor are much higher than dollars spent on software. Positioning your product as a replacement for 1-5 people instead of just another SaaS tool sets you up to charge a lot more. Replacing five head-count could justify $200 K+ in spend on your product vs. $100 per month if it's just a tool for five people.
Moving to outcomes-based pricing and positioning your software as a labor replacement vs. a productivity tool seems like a no-brainer then, right?
The Current State of Affairs
While the idea of outcome-based pricing is more or less “consensus” today (people on Twitter can’t stop talking about it), the implementation of it is still “non-consensus.” Only ~5 % of SaaS companies actually do outcome-based pricing today (2024 SaaS Benchmarks, 2025 State of B2B Monetization).
Most of the biggest names in AI haven't migrated over just yet.
ChatGPT is subscription-based (although a bad example since it doesn't try to solve specific problems).
Cursor is usage-based, and Claude Code—which markets itself as a tool to “turn issues into PRs” (feels perfect for outcomes-based pricing)—also is usage-based.
Some AI SDRs charge per outcome (like 11x), but others (like Regie) haven't moved to that pricing model yet.
You primarily see AI CS/CX platforms (Decagon, Sierra, Intercom) charging for outcomes, and there's a smattering of other companies experimenting with new pricing models, but companies are still overall using existing pricing methods.
What Happens in 10 Years?
Right now, companies are able to position their TAMs as massive since they are being compared to teams of humans who cost a lot more. But what happens when, in five years, you're being compared to another AI-native company who can charge $0.90 per outcome instead of $0.99 per outcome? And what happens in 10 years when you're being compared against an AI-native solution who can charge $0.10 or $0.01 per outcome?
Look at what happened to cloud storage: it went from $1/GB to $0.023/GB in a decade as competition intensified. Or consider how video conferencing went from expensive enterprise solutions to free Zoom calls. The same commoditization cycle that hits every technology market will eventually hit AI-powered labor replacement.
The timeline typically follows a predictable pattern: proprietary advantage → competitive market → commoditized utility. Early AI companies enjoy the advantages of being early today, but as the underlying models become more accessible and competition intensifies, most labor-replacement tasks will slide toward commodity pricing.
Some markets will consolidate into winner-take-all scenarios where network effects or switching costs create moats. But others will fragment and devolve into price wars.
To be clear, this is great news for people and companies who will eventually see massive productivity gains at a fraction of today's costs. But it should also be a wake-up call for some founders and VCs betting on new entrants disrupting $100 B+ TAM markets. In the long run, those markets likely won't reflect the total current labor spend; they'll instead shrink to just a fraction of what they are today. As competition drives prices down to the marginal cost of delivering outcomes (which will continue to fall as compute gets cheaper), today's $100 B market might look more like $10 B in a decade (assuming the same level of demand).
Jevons to the Rescue
What you really have to hope for is Jevons' Paradox to arrive like a knight in shining armor to save the day. Lower costs need to drive up usage enough to offset the shrinking unit economics.
- If creating software costs 1/10 of what it does today, will people create 10× the software? (probably)
- If responding to CS/CX tickets costs 1/10 of what it does today, will you be able to serve 10× more customers? (possibly)
- If doing your taxes cost 1/10 of what it does today, would you file 10× the amount of taxes? (probably not)
The tax example perfectly illustrates the trap many founders fall into. You can pitch the spend currently allocated to those salaries as their TAM, but in the long run, this doesn't work unless people actually file more taxes (which they won't, because tax filing is a compliance requirement, not a discretionary activity that people do more of when it gets cheaper).
So What Should You Do?
If you're going after labor spend, you need to pressure-test your assumptions with brutal honesty. Ask yourself: If your cost per outcome drops 90 %, will demand increase 10× to maintain the same market size? Map out the adjacent workflows you could expand into if your core market commoditizes. Identify what would have to be true about human behavior for Jevons' Paradox to save your market.
Without that expansion story (either through massive demand increases or adjacent workflow capture) the $100 B vision you're selling to investors might vanish like a shimmer on hot desert sand. The labor substitution mirage looks real from a distance, but gets hazier the closer you get to long-term market dynamics.