When Anthropic announced its enterprise pricing changes earlier this month, most observers missed what was actually happening. They saw a price hike. I saw a confession.

I noticed it firsthand. My own Claude Pro subscription reached its usage limits faster than I expected, forcing me to increase my monthly spending cap for token consumption. It’s not painful, and still a great deal for the value delivered, but the friction got me thinking about something I wrote about a few months ago: that we’ve been operating under a pricing model that simply doesn’t match the economics of what we’re actually building.

Back in November 2025, I published a piece on how AI and consumption-based pricing are reshaping HR tech. The framing was clear: SaaS is not dead, but the traditional per-employee-per-month model doesn’t capture what’s happening when AI agents start doing the work instead of just augmenting it. I argued that vendors and buyers were both groping toward outcome-based and consumption-based models, but hadn’t yet figured out the mechanics.

Anthropic just gave us the answer to why they haven’t figured it out. Because the subsidy has to end somewhere.

To be clear: this isn’t a criticism of Anthropic. Every major AI lab: OpenAI, Google, Microsoft, and Meta, has been pricing below true cost to accelerate adoption and capture market share. That’s not a secret, and it’s not unique to any one company. What’s notable about Anthropic is that they’re the first to adjust publicly and transparently. That takes more courage than quietly absorbing the loss or burying the change in a support doc update. The rest of the industry is watching and will follow.

The Subsidy Was Always About Their Growth, Not Your Budget

Let’s start with the obvious part: Twenty dollars a month for Claude Pro, free API credits for new users, enterprise deals bundled with token allowances that made no sense economically; all of it was predicated on the assumption that compute costs would decline faster than adoption would grow.

That assumption held for a while. The per-token cost of running Claude dropped significantly. But then two things happened at once.

First, actual usage exploded. Claude Code weekly active users doubled between January 1 and mid-February 2026, according to Boris Cherny, who leads the product at Anthropic, a figure he cited directly in the context of the company’s $30 billion Series G announcement. Business subscriptions quadrupled in the same period.

Second, those workloads are expensive. Agentic AI, systems that break tasks into steps, iterate, and handle reasoning across multiple rounds, consume anywhere from 5 to 30 times more tokens per task than a standard chat interaction. For complex coding and multi-step workflows, recent research puts that figure at orders of magnitude higher. So even if your per-token cost drops 70 percent, the volume increase more than offsets the savings. The math doesn’t work at the old price.

The math broke. And Anthropic made a choice: be honest about it.

Starting in November 2025, the company began transitioning enterprise customers to usage-based billing. As reported by The Register, Anthropic introduced a single $20-per-employee monthly fee covering Claude chat, Claude Code, and Cowork on a usage-based plan, but customers must now commit to and pre-pay estimated token consumption on top of that seat fee. The bundled token allowance is gone.

That’s true. It’s also incomplete. What they were actually saying is: we can’t afford to guess at how much compute you’ll consume anymore. You need to forecast it, commit to it, and pay for it.

For enterprises, that’s a structural shift. You’re no longer buying a software seat. You’re budgeting for compute consumption, just as you do for cloud infrastructure.

This Is Happening Everywhere in B2B

Anthropic is not alone. This is the canary in the coal mine for the entire enterprise software market.

That shift accelerated sharply through 2025 and 2026, with usage-based and hybrid models now adopted by 85 percent of SaaS leaders.ICONIQ’s 2026 State of Go-to-Market survey found that 48 percent of companies now report hybrid pricing as their primary model.SAP shifted from per-user and subscription-based pricing to a consumption-based model in direct response to AI agents automating workflows. Microsoft layered Copilot consumption fees on top of existing subscriptions. OpenAI’s Nick Turley, head of ChatGPT, acknowledged publicly that “it’s possible that in the current era, having an unlimited plan is like having an unlimited electricity plan. It just doesn’t make sense.” The trigger is the same everywhere: AI broke the per-seat economics.

When value came from access to software, from a seat at a screen, pricing per seat made sense. You paid for the tool, and the tool’s productivity scaled with how well you used it. The vendor’s cost was largely fixed: infrastructure, support, and development.

AI inverted that. Value now comes from outputs: decisions made, documents processed, code written, candidates screened, etc. And output scales with compute, which scales with usage, which scales with token consumption or API calls. The vendor’s cost is no longer fixed. It’s variable, and it’s substantial.

For HR and Work Tech buyers, this isn’t an abstraction. Most of your vendors are building on top of these frontier models. When compute costs go up at Anthropic, OpenAI, or Google, those costs don’t stop there. It flows downstream into every AI-powered recruiting, engagement, HRIS, Payroll, and workforce management tool you’re paying for. 

That’s why the industry is migrating toward hybrid models: a base subscription fee for stability, plus metered consumption for growth. It’s not elegant, but it works. It aligns incentives without making buyers’ finance teams panic about unpredictable bills.

But Here’s Where Most Vendors Are Bluffing

The research shows something interesting. 85% of SaaS leaders say they use usage-based or hybrid pricing. But when you dig into the data, a different picture emerges.

Bain & Company analyzed more than 30 established SaaS vendors introducing generative AI capabilities and found that none had fully shifted to AI usage- or outcome-based pricing. Not one. The talk track is universal; the execution is rare. In HR tech and work tech, that gap is especially acute. Vendors can measure what happened: a hire was made, a risk was mitigated, a process was automated, but proving that the software caused it is brutally hard. Was the candidate hired because the recruiting tool was better, or because the economy improved? Did retention go up because the engagement platform worked, or because leadership got better?

Without clean attribution, outcome-based pricing collapses into finger-pointing and contract disputes.

So what most vendors are actually doing is simpler: they’re adopting consumption language while staying operationally hybrid. They charge a base fee for core functionality, then layer on usage fees for the parts that scale. It’s consumption-based in name and often in pricing structure, but it’s not truly results-based or outcome-based. The vendor hasn’t actually taken on the risk that their tool will deliver the outcome. The buyer is still on the hook.

In HR tech and work tech specifically, I see this play out in three maturity levels:

Level One: Still figuring it out. These vendors are retrofitting their existing SaaS infrastructure to accept usage metrics, but they haven’t thought through what metric actually captures value. They’re charging by API calls or events, or worse, by the number of times a feature is clicked. It’s consumption-based in the most literal, least useful sense.

Level Two: Genuine traction. These vendors have invested in operational infrastructure to deliver on outcome- or consumption-based pricing. They can track the metrics that matter, they’ve built the billing system that supports them, and, crucially, they’ve structured their customer success and implementation to ensure the outcome. They own part of the outcome. The vendor is genuinely committed.

Level Three: Positioning ahead of capability. These vendors are talking about outcome-based pricing because it’s what buyers want to hear. But operationally, they’re still running the old model. They haven’t restructured how they implement, measure success, or support customers. They’re waiting to see if the market will force the issue before they invest.

The uncomfortable truth: most vendors are at Level Three.

(This three-level maturity framework is original to WorkTech’s analysis and reflects patterns observed across the vendor landscape over eight-plus years of tracking HR and work tech funding and go-to-market activity.)

What This Actually Means for Your Next Move

If you’re a vendor, Anthropic’s shift should make you uncomfortable in a good way. It’s a signal that the market is calling your bluff. Buyers are getting smarter about the difference between pricing language and pricing mechanics. They’re asking harder questions: Does the vendor actually own part of the outcome, or am I buying a tool and hoping for the best? Can they prove what metric they’re optimizing for? Will they restructure their implementation and support to ensure I hit my goals, or are they just billing me for using the product a lot?

If you’re a buyer, the signal is different but just as important. For nearly three decades, I’ve told HR tech buyers: “You’re not shopping for software; you’re shopping for partners.” It was good advice then. It’s now a critical success factor. The difference is structural: in a consumption or outcome-based model, the vendor’s revenue is directly tied to whether you get results. If they don’t invest in your success, in implementation, in adoption, in measurement,  they don’t get paid. That alignment changes the nature of the relationship from transactional to genuinely collaborative. Choosing the wrong vendor isn’t just a bad purchase; it’s a mistake. It’s a failed partnership with measurable business consequences.

That means asking vendors the questions that separate real from rhetoric: What happens to your revenue if we don’t hit the outcome? How exactly do you measure success, and who owns the data? Are you restructuring your implementation and customer success model to support this, or are you just relabeling your pricing page?

The hard part,  the part vendors and buyers are both still avoiding, is that real outcome-based and consumption-based pricing requires operational transformation, not just accounting changes. It requires vendors to invest in implementation rigor. You can’t claim outcomes if you’re not willing to spend the time getting customers to adoption and usage maturity. It requires building transparent measurement systems: the metrics must be agreed upon upfront, tracked in real time, and auditable. No guessing. No disputes. And it requires aligning your success with theirs. If the outcome doesn’t happen, your revenue takes a hit. That’s the only way buyers will actually believe you’re committed.

Most vendors aren’t there yet. And frankly, most buyers aren’t ready to hold them accountable to that standard either.

The Reckoning Is Just Beginning

The question now is how the market will respond. Will vendors use this as an excuse to shift the cost burden to buyers, or as an opportunity to genuinely restructure around customer outcomes?

I lean toward the former happening first, and the latter emerging only when buyers demand it loudly enough.

For Work Tech and HR Tech, the stakes are higher than most. Hiring, retention, engagement, and productivity are outcome-focused by definition. There’s no hiding behind “we provide tools and you decide how to use them.” The vendors who figure this out first will have an enormous competitive advantage.

The rest will find themselves explaining why their “outcome-based” pricing model is just consumption-based pricing with better marketing.

What Comes Next

This transformation in how Work Tech is priced and delivered isn’t just a billing question. It’s a fundamental shift in what vendors are actually selling and how they’re structured to deliver it. The economics of AI have forced the issue, but the real opportunity lies in a bigger structural change: how software is actually built, implemented, and operated to create measurable outcomes.

I’m currently working on a follow-up piece that dives into exactly this: where vendors and buyers are succeeding together on outcome-based and consumption-based models, and what separates those wins from the vendors who are still just talking about it. You’ll see concrete examples of vendors who’ve fundamentally restructured how they operate: how they implement, how they measure success, how they partner with buyers to hit specific outcomes. And you’ll see what happens when that alignment is real.

Because the vendors who figure this out, who genuinely restructure their operations and success metrics around outcomes, aren’t just winning on pricing strategy. They’re winning on the market itself. That’s the story worth telling.