Why Sales Pay Trails How Software Is Priced

In 2026, 54 percent of Anthropic's new enterprise logos closed through a self-serve funnel, with real contract value and no account executive involved. The figure is Anthropic's own, presented by its Head of Industries at SaaStr AI 2026, so read it as vendor-reported. Now put a second fact next to it: crowd-sourced compensation data on RepVue shows Anthropic's enterprise AEs on a roughly 60/40 base-to-variable split, with the variable side tied primarily to bookings.

The company running the most AI-native sales motion in software prices and closes like 2026 while paying its sellers like 2019.

I have spent 27 years in enterprise SaaS, literally there for the beginning where I remember explaining "annual recurring revenue" in meetings. The Anthropic pairing shows a pattern I now find everywhere which I am calling: the Comp Lag. Organizations redesign their go-to-market in months and their compensation plans in years, because comp is where an organization's real beliefs live. You can announce a new strategy in a keynote, but the comp plan tells your sellers what you actually want, and right now most comp plans are telling sellers to want something customers have stopped paying for. (Also read: Enterprise AI Is an Organizational Design Problem in Disguise)

Pricing and pay are one ladder

Let's look at this standard industry framing that OpenView, Bessemer, and every pricing consultancy has drawn some version of.Think of software monetization as a ladder with three rungs. Each step up transfers risk from the customer to the vendor, and each rung has a matching way to pay the people who sell it.

Pricing and compensation have to sit on the same rung, because a rep paid one rung below the pricing model is optimizing a number the customer no longer cares about. A seller comped on bookings in a consumption business gets paid in full the day a contract signs, while the vendor only gets paid as usage ramps, which means the seller's best-paid moment and the customer's least valuable moment are the same moment. That mismatch, not AI, is the force actually breaking the AE job. AI just made it impossible to ignore, because once machines absorbed the prospecting and the follow-up, the only thing left to examine was what the human is paid to do.

The rungs are being tested in production right now

Snowflake ran the second rung to its logical end years before the current AI cycle. CFO Mike Scarpelli described the design plainly in a 2021 Lightspeed interview: "Every one of our reps has a big consumption quota." Snowflake went further and dissolved its customer success function entirely, because a rep paid on usage should own the customer's success personally. The account's growth became the rep's growth.

MongoDB learned the darker lesson of rung two. Meghan Gill, its SVP of Sales Ops, tells a fable about a village that paid a bounty per dead cobra and ended up with cobra farms, because compensation drives behavior, including behavior nobody wanted. MongoDB now blends ARR and unit-based metrics because paying on any single number teaches sellers to farm that number.

The third rung is where the newest AI vendors are trying to live. Intercom's Fin charges 99 cents per resolved conversation, HubSpot cut its Customer Agent to 50 cents per resolution in April 2026, and Sierra sells enterprise contracts priced on outcomes. Underneath those clean numbers sits a fight over definitions. Intercom has cited a resolution rate around 71 percent while independent reports place it closer to 42 to 50 percent, a gap that comes down to what counts as resolved, including conversations where the customer simply stops replying. When the unit of payment is contested, every invoice is a negotiation. (Also read: The Proxy Collapse: When Web Traffic Stops Being Human)

Whaling crews were paid on outcomes because barrels are countable

Economic history already ran this experiment. Nineteenth-century whaling crews were paid in "lays," fractional shares of the voyage's catch, because pay had to arrive the way revenue arrived. It worked because a barrel of whale oil is indisputable. Both sides count it on the dock and the share pays out.

Software outcomes are not barrels. A "resolution" or an "adoption milestone" requires a referee, and the vendor employing the rep is not a neutral one. Which leads to a claim I did not expect to reach: consumption, not outcomes, is the honest rung for compensation. Usage is measurable, fast, and hard to fake, while outcomes are contested and slow. I believe the future of sales compensation is deliberately less radical than the future of pricing. Vendors can absorb outcome risk on the pricing side because they hold the portfolio, while an individual rep with a mortgage cannot wait eighteen months for a disputed outcome to clear.

The rep that survives is a general manager with fewer colleagues

So what does the rung-two seller actually do? I believe the job consolidates rather than shrinks: territory marketing, net-new selling, standing the tools up, shaping configured solutions alongside product, and growing the account's usage. Snowflake already showed where the old customer success duties land, which is squarely on the rep. Anthropic's Head of Industries compressed the shift on the SaaStr stage: "Sales leaders are rapidly becoming systems thinkers over deal strategists." The seller becomes a general manager of a territory, keeping every responsibility while machines and self-serve funnels absorb the colleagues who used to share them. (Also read: What Survives When AI Becomes a Commodity)

Fewer colleagues is a gentle phrase for a hard fact. If half of enterprise deals close without a rep, the industry needs fewer and more senior reps, and the entry-level pipeline underneath them, the BDR seats where careers like mine began, thins out first. A generation of sellers may never get the apprenticeship deal my generation got.

Where does that leave pricing designs that gate good terms behind mandatory rep engagement? Anthropic answered that one by publishing identical pricing for self-serve and sales-assisted enterprise plans. When rep involvement is price-neutral, the rep has to be worth engaging on merit rather than protected by a toll booth.

Nobody has finished this transition, and Anthropic's own hiring proves it, because the company is recruiting incentive compensation designers with consumption-model experience while its AEs sit on bookings plans. The org chart moved in four months and the comp plan is still in the interview stage. If the company furthest ahead is mid-climb, so is everyone. The diagnostic takes ten minutes. Write down the rung your pricing sits on and the rung your comp plan sits on, and if they differ you have found your Comp Lag, and your best sellers found it first.

Key Takeaways

• Pricing and sales compensation form one ladder with three rungs, seats, consumption, and outcomes, and incentives break whenever the two sit on different rungs.
• Anthropic reported that 54 percent of its 2026 new enterprise logos closed self-serve while crowd-sourced data shows its AEs still paid primarily on bookings, making the Comp Lag visible at the industry's leading edge.
• Consumption is the honest rung for rep pay because usage is measurable and fast, while outcomes require a referee and the vendor is not a neutral one.
• Snowflake's decision to dissolve customer success and put consumption quotas on every rep is the working prototype of the consolidated seller role.
• Price-neutral rep involvement, the model Anthropic published for its enterprise plans, forces the rep to be worth engaging on merit rather than protected by pricing design.
• The consolidated rep keeps every responsibility while the entry-level pipeline beneath the role thins out, a cost the industry should name rather than euphemize.

FAQ

What is the Comp Lag? The Comp Lag is the gap between how a software company prices its product and how it pays the people who sell it. Companies move pricing to consumption or outcome models in months, while compensation plans stay anchored to bookings for years, leaving sellers optimizing a number customers no longer pay for.

Why not just pay reps on outcomes if customers pay for outcomes? Outcomes in software are contested measurements rather than countable units, as the dispute over what counts as a "resolution" in AI agent pricing shows. A vendor can absorb outcome risk across a portfolio of customers, but an individual rep cannot wait out an eighteen-month attribution argument for a paycheck.

Does the Comp Lag mean fewer sales jobs? It points to fewer, more senior, better-paid sellers who operate as general managers of territories, with self-serve funnels and AI absorbing transactional work. The entry-level pipeline shrinks first, which is the human cost of the transition and worth planning for directly.

About the Author

Martin Goetzinger has spent his career in enterprise software sales, helping large organizations such as Apple, Microsoft, and Verizon connect data, insight, and action. His work focuses on transforming how businesses measure success and create customer value through technology.

Outside the enterprise world, he writes about the five forces he believes are reshaping everything: AI, blockchain, energy, personalized health, and robotics. Not from a purely technical lens, but from a human one as to how these technologies will redefine work, wealth, and well-being.

He is based in the U.S. and publishes at www.MartinGoetzinger.com.

Disclaimer

The views expressed in this article are the personal opinions of the author and are provided for informational and educational purposes only. Nothing in this article constitutes investment advice, financial advice, legal advice, or any other form of professional advice. Do not make investment or financial decisions based on the content of this article. Always consult a qualified professional before making decisions that affect your finances, business, or livelihood.