Software pricing has always been a mirror of its era. From perpetual licenses that reflected the economics of physical distribution to per-seat SaaS that mapped cleanly onto org charts and budget cycles, each model made sense for its moment in time. But the arrival of AI agents may represent something more disruptive than a new pricing trend: a fundamental shift in the assumptions that have guided commercial models for decades. 

In a recent episode of the Mostly Growth podcast, Metronome CEO Scott Woody shared his thoughts on what happens to pricing in a future where the buyer is no longer human. In this blog we’ll focus on those points, and others, that Scott raised, the first being: the shift in pricing becomes architectural, where companies must account for the purchasing processes of AI agents, and legacy infrastructure begins to break. As we sit in the Value Era, where usage-based structures align cost with realized outcomes, it’s worth examining what the future might look like in a new era of software.

Which pricing era are we in now? Get oriented with the Monetization Operating Model whitepaper.

The shift from seats to work

Before exploring what agent-driven pricing might look like, it's worth understanding why the commercial model question is opening up at all. Scott's framing is useful here: the value proposition of software is undergoing a fundamental shift.

In the Access Era, software was valued by how many people had access to a shared tool or source of truth. This seat model worked because humans were the ones doing the work. In the episode, Scott touched on how pricing was introduced within SaaS pricing as a human construct, where decisions have been made on a wealth of research into how humans make buying decisions. Tiered pricing, high anchors, good/better/best packaging—all examples of how pricing has been structured for human processing. 

In the era of agents, he noted, it’s unlikely that the buyers will function within this mental construct or constraints. This era is likely better suited for outcomes and price optimizations, where agents can evaluate thousands of variables simultaneously. 

Scott also calls out how this inverts a core assumption: rather than wanting fewer levers, as humans do, agents may actually benefit from more of them. One example he uses is the pricing page, which was built for human comprehension. In the Agentic Era, pricing pages could become less relevant over time as agent-to-software interactions become more common.

From plan selection to budget allocation

Another useful mental model Scott covered is what the end-state of agent-driven purchasing might look like. Rather than a human buyer evaluating plans and selecting the one that fits, the interaction might look more like a human setting a budget and a set of priorities, then delegating the actual spending decisions to an agent.

The analogy he used is a marketing budget. A CMO isn't evaluated on which line items they spent against—they're measured on outcomes. The negotiation isn't "Which plan did you buy?" but "How much did you allocate, and what did you get for it?" Applied to software, this suggests the commercial conversation may migrate away from packaging and toward goal-setting and resource allocation.

That dynamic has an implication for how pricing is structured. A human buyer makes one decision and rarely revisits it. An agent may be continuously calculating against cost, which means bundled or opaque pricing structures could create friction where granular, queryable pricing might not.It's worth noting this is still largely a forward-looking frame. Most software purchasing today still runs through human evaluation. But understanding where the logic is heading can help teams build infrastructure and commercial models that won't require rebuilding from scratch when the transition accelerates. 

Credits as a tool, not a metric

The use of credits in a pricing model has been generating a lot of debate in monetization circles recently. They continue to emerge and work well in specific contexts and tend to break down in others.

In the episode, Scott noted how in sales-led motions, credits serve a useful function by adding abstraction, creating negotiation surface, and allowing for last-mile discounting without changing list prices. A sales process can absorb that complexity because there's a human on both sides to explain and interpret the terms. In product-led or self-serve motions, complex credit models tend to struggle. Without a sales process to translate value into credits and back again, the abstraction becomes a barrier rather than a feature. 

‍But when your buyer is an agent, and no sales conversation even needs to be had, that complexity can be absorbed , and the agent can make the purchasing decision at any level of complexity. Scott's framing is that the category error to watch for is deploying a complex credit model in a context where there's no mechanism to help the buyer make sense of it.

What teams should be thinking about now

‍Scott was deliberate about not overstating how far along this transition is. Most purchasing decisions still run through humans, and teams that try to force new pricing models before their market is ready presents risks.

‍But he pointed to a couple of companies he sees are thinking about this in a good way. Clay is treating pricing as a continuous variable—actively thinking about what the right model is for this moment in their market. Lagora is staying on a seat model for now, but appears clear-eyed about the fact that their value is increasingly the work done on a customer's behalf. Scott's read is that the right posture is knowing where the logic is heading and building infrastructure that won't require rebuilding from scratch when the transition accelerates.

‍The useful question to pressure-test today: will your pricing and billing infrastructure hold up when the buyer is an agent?