Pricing the AI Workforce: From Pilots to Real Revenue

AI breaks software’s near-zero marginal cost assumption for the first time.
- Unit economics now depend on pricing strategy, not just growth.
New framework: Autonomy × Attribution matrix determines right pricing model.
- Low autonomy + low attribution → seat-based (Figma, Slack).
- High attribution → outcome-based pricing unlocked regardless of autonomy.
Only 5% of AI pilots deliver value — bad design, not bad AI.
- Free pilots attract uncommitted buyers; charge from day one.
- Co-create ROI metrics in customer’s language before pilot starts.
Outcome-based pricing = holy grail, but not a moat.
- Competitors replicate pricing fast; moats come from network effects + embedding.
5% of companies at high autonomy + high attribution now; 25-30% within years.
- This quadrant is where AI agents fully replace human workflows.
Pricing innovation is an early wedge, not durable defensibility.
- Build network effects, scale, brand — pricing is table stakes.
Founders who define category pricing now own the narrative long-term.

X discourse

@JasonrShuman: “The shift is from selling tools (per-seat) to selling work (per-output). $2K/month AI agent replacing an $80K/year agenc” (704 likes)
@levie: “Compute budgets for workers wielding AI agents will monotonically go up. Companies will budget for this owned by busines” (496 likes)
@BoringBiz_: “AI will replace a portion of the labor force. Amazon, Softbank and NVIDIA voting with dollars.” (508 likes)
@GalvinAlmanza: “Creates generational resume gap, overworked staff. Some expect 100% robot replacement.” (1306 likes)
@emollick: “Token costs are not trivial, humans are much cheaper than AI for many tasks. Efficiency & compute will alter those calcu” (380 likes)
@MindTheGapMTG: “Budget AI as fractional hire, not software license. $300/month vs $8K/month salary: rounding error.” (4 likes)

Anna Piñol and Pete Flint — NFX (Pete is founding partner, ex-Trulia founder) · 2025-09-03 · Read on nfx.com