Agentic go-to-market will consume usage at a scale that breaks current pricing models

The number Harry Stebbings puts on agentic go-to-market is stark. When agents run these motions, he argues, they will consume 10 to 100 times more usage than humans ever could. That is not an incremental adjustment to growth assumptions. It is a claim that the total addressable market most software vendors are currently pricing against is understated by an order of magnitude.

The logic behind the call is worth unpacking. Human-led go-to-market is capacity-constrained in ways that agent-led motions are not. A sales rep works a finite number of hours, touches a finite number of accounts, and runs a finite number of sequences per day. An agent has none of those ceilings. When the bottleneck shifts from human attention to compute and data access, the volume of activity a platform must support scales by a factor that pricing models built around per-seat or usage-cap structures were never designed to absorb.

This creates an immediate pricing problem for software vendors. Products sold on unlimited usage tiers, or on the promise of unlimited async client access, tend to look attractive at the point of sale precisely because buyers expect light consumption. The perceived value is high; the actual draw on infrastructure is low. That arithmetic inverts the moment agents replace humans as the primary users. A tier priced for a sales team of 20 humans running manual sequences becomes structurally different when 20 agents each run sequences at a pace no human could match.

When agents can run these GTM motions, they will consume 10 to 100 times more usage than humans ever could. Harry Stebbings

What makes Stebbings’s call worth watching is its falsifiability. The claim is not that AI will eventually change software pricing. It is that the usage differential between agentic and human go-to-market motions is large enough, right now, to require vendors to rethink their total addressable market estimates. That is a claim software companies can test against their own telemetry as agent adoption accelerates. If the usage multiples in live deployments land anywhere in the 10-to-100 range, the vendors who built their market-size projections on human-consumption baselines will find themselves undersupplied in infrastructure and underpriced in their contracts at the same moment demand spikes.

The harder question is whether the TAM expansion Stebbings describes accrues evenly or concentrates. A market that grows 10 to 100 times in raw usage volume does not automatically produce 10 to 100 times more revenue for incumbents. It depends on whether vendors reprice before the expansion is obvious, whether buyers accept usage-based models that reflect actual agent consumption, and whether the agents themselves are running on the vendor’s platform or on a competitor’s. The call about market size is directionally clear. The call about who captures that market is not in the evidence, and Stebbings does not make it.

The bet can be checked. As enterprise software companies report cohort data on agent-enabled versus human-led accounts, the usage differential will become visible. If the gap lands closer to two or three times rather than 10 to 100, the market-size thesis softens considerably. If it lands at the high end of Stebbings’s range, the repricing pressure on the industry will be significant and rapid. Either outcome arrives with observable data attached, which is what makes this a genuine prediction rather than a general observation about AI’s potential. Readers who track software-as-a-service metrics will have an answer within a few years.