Building an AI-Resilient Vertical SaaS


Premise: If you believe AI will materially improve your core product over time, you should assume productivity per worker will compound. In markets with a hard ceiling on total output, per-seat pricing quietly compresses TAM as your product gets better. Outcome/usage pricing keeps you aligned with value as AI scales.


The Mechanism

  • Let the per worker productivity boost of your product be $p$. As AI optimists, we believe believe $p$ will continue to increase over time.

  • Total work output $ D(p)$ rises with $p$ but saturates at a demand ceiling.

  • Seats required $≈D(p)/p$. As $p$ increases, headcount falls faster than output rises.

  • Per-seat revenue $∝D(p)/p$ - it declines with AI progress.

  • Usage revenue $∝D(p)$ - it tracks adoption and then plateaus at the natural market ceiling.

Buyers rarely accept seat prices that scale inversely with productivity (“you’re 10× faster, so pay 10× per seat”). So per-seat monetization decouples from the value your AI creates.


!Figure 1 - Output rises with productivity, then caps at demand ceiling (illustrative, not to scale)


!Figure 2 - Workers required falls as productivity increases (illustrative, not to scale)


!Figure 3 - TAM under different pricing models (illustrative, not to scale)


What To Do Instead


Price the unit of outcome you actually create. Projects completed, designs approved, simulations run, filings submitted. That aligns revenue with realized value and preserves TAM as AI lifts throughput.


Practical design

  • Define the atomic work unit (not seats)

  • Make collaboration cheap or free; gate value on output

  • Pair usage pricing with a platform minimum (governance, SSO, audit)

  • Add transparent meters (e.g., solver minutes, render frames) for heavy workloads

  • Revisit meters quarterly as AI improves throughput

Where per-seat can still work

  • Uncapped or expanding demand (AI creates new categories of work)

  • Strong network effects tied to identity/roles (but still consider hybrid: platform + usage)

  • Early product market fit, where productivity boost $p$ is still low (<4x)

Bottom line for HVAKR


We can stay seat-based near-term as our $p$ is estimated between 2 and 4. But the strategic direction of HVAKR should be outcome-indexed pricing. As our AI lifts engineer throughput, usage-based monetization will better reflect value delivery, protect TAM, and make the business structurally stronger.