So what can you learn for your own monetization strategy? Anthropic is quietly running a masterclass. Look at its top plan and you'll notice something odd — it unlocks four things, and not one of them is a feature:
- 5x or 20x more usage than Pro
- Higher output limits for all tasks
- Early access to advanced Claude features
- Priority access at high traffic times
Same models, same tools as the cheaper plans. You're buying more of it, sooner, and ahead of everyone in line when things get busy. And that's the whole story of how Anthropic decides who pays what.
If you're a founder, a PM, or running growth, read closely — because these are levers most SaaS pricing pages still ignore, and they're probably sitting unused under your own product.
First — what segmentation actually is
Segmenting a market means splitting the people who might pay you into groups that want different things and will pay differently — then building a plan for each. The art is choosing which attributes you split on. A few that matter almost everywhere:
- Willingness to pay — how much value they get, and so how much they'll tolerate on the invoice.
- Sophistication — a newcomer needs hand-holding; an expert knows what they want and resents being slowed down.
- Who's paying — spending your own money is nothing like spending a company's.
- Team size — one person, a small team, or an org with procurement and security needs.
- Intensity of use — a light dabbler versus someone in the product eight hours a day.
Good segmentation reads those attributes off a customer and routes them to the right plan — ideally so smoothly the plan feels built for them. Anthropic does it with three levers: features, usage, and a new set of components AI made sellable. Watch each — and watch who it's built to catch.
Features, used as a filter
The first lever is the classic one: some capabilities aren't available until you pay — the strongest models, and Claude Code, the coding tool, which you can't touch on the free plan at all.
That single choice sorts two very different people at the door.
The developer knows exactly what Claude Code is worth. Locking it away doesn't frustrate them — it tells them what to buy. High intent, high willingness to pay (their time is expensive; a tool that saves it pays for itself fast). So you don't coddle them — you put what they came for one tier up and let their intent do the work.
The operations manager at a logistics firm is the opposite. Not an early adopter, spending the company's money, never heard of Claude Code — she just wants out of spreadsheets and supplier emails. Throttle her too early and she leaves before she feels the value. So the free tier is generous with what she will use and never mentions the tool she doesn't need.
And notice what her upgrade lever isn't: a locked feature. She'll never pay to unlock Claude Code — she doesn't know she wants it. Her lever is usage plus a moment of realization: she hits the free cap on everyday work the same week a workflow saves her a real afternoon. That's what moves her to Pro. Same product, a completely different path in.
Same lever, two jobs — an open door for the unfamiliar, a price tag for the expert. Features here aren't the strategy. They're the sorting mechanism that reads who you are and points you at the right plan.
Usage, and the business-vs-hobbyist split
Everyone past the door is paying now, holding the same tools. So how do you separate a light user from someone running Claude all day? On how much.
And "heavy user" isn't one person. Look at how Anthropic splits its Team plans into a standard and a higher business tier, and ask what behaviors they're pricing:
- The tinkerer — a junior dev or hobbyist iterating and learning. Real usage, but bursty. Needs more room, not guarantees.
- The business operator — someone running a business on Claude. Load-bearing usage. Throttled mid-afternoon means work stops and money is lost.
Design for each and you'd ask: When do they work — can they afford to be capped in those hours? A bigger bucket, or a bucket that never runs dry at the wrong moment? One person, or a team?
Anthropic's answer: make the higher tier not just "more usage" but more headroom plus team governance — shared admin, higher ceilings, room to run hard through the workday. The tinkerer stays on the lighter tier; the operator self-selects up, because being throttled now costs real money and the plan is visibly built for that fear.
Consumption isn't a new lever — metered infrastructure always charged for it. What's new is that it moved from a billing footnote to the main axis of segmentation, because for an AI product every action has real weight behind it. For the serious user, capacity becomes the product: you're not selling what it does — you're selling the guarantee it'll be there when the work is.
The new components AI created
Here's the lever that couldn't have existed a decade ago.
Back to that top-plan list: early access. Priority access at high traffic times. Not features, not quite "more usage." A new category of value — sellable only because the resource underneath got scarce and live.
Priority in the queue targets the person who can't afford to wait. When demand spikes, someone gets throttled — paying more means it isn't you. Classic SaaS had no busy-hour queue to auction; nobody paid Salesforce for "priority when the servers are full," because the servers were never full in that sense. Scarcity created a thing to sell.
Early access targets the person who needs to be at the edge. Everyone gets the features eventually — you're paying to be first, not different. It sorts out the segment for whom being a month ahead is worth money.
Neither is "more." They're certainty and position under scarcity — and they exist only because compute is scarce and live. Take the scarcity away and both evaporate. That's what makes them genuinely new, and genuinely AI-created.
Put the three levers together and the map appears
Read the ladder through the three levers and a different person is caught by each:
- Free ($0) — the newcomer and evaluator. Sorted by the feature filter.
- Pro ($20) — the individual professional, past the door, full toolset.
- Max ($100 / $200) — the power user and edge-seeker. Sold on usage and new components: 5x/20x, higher output, early access, priority.
- Team ($25 / $125 per seat) — the small org, split into tinkerer-grade and business-grade by usage and governance.
- Enterprise (~$20/seat + metered usage) — the large regulated buyer. Adds compliance and identity, and meters usage on top — quietly turning the subscription into a meter.
Features sort exactly one boundary: paid or not. Everyone above that line is segmented by usage and the new components. The feature filter sorts; usage and the new levers monetize.
AI didn't just hand us a usage meter. It created new kinds of value — access, priority, timing under scarcity — and those, alongside usage, now do the segmentation work features used to do alone.
This isn't only an Anthropic move
Once you see the pattern, it's everywhere the resource underneath is scarce or metered.
Lovable barely segments on features at all. Its plans are defined by credits — a consumption unit that burns faster on complex requests. The tiers mostly differ in how many credits you get and how much governance wraps around them. The value metric is the segmentation.
Base44 goes further and splits the meter itself: message credits you spend while building, and integration credits your live app's users spend when the app runs. It separates "you building the thing" from "your users running the thing," and prices each — a split there was no reason, or mechanism, to make a decade ago.
Same lesson: when a real, usage-linked cost sits under your product, you get value dimensions to segment on that pure-software SaaS never had.
A caution — don't add a lever just because you can
These components are seductive because they're new. It's tempting to bolt on a "priority tier" or a credit system because it looks modern. Don't — not unless it maps to something the customer genuinely values and experiences.
A lever only works if it reflects real value and helps the customer see which tier they belong in. A credit system where usage barely varies just breeds anxiety and billing confusion. A "priority" tier with no real scarcity is a fee for nothing, and customers feel it. If a lever doesn't make the choice clearer, it's noise — and noise erodes trust faster than a plain plan ever would.
Add the dimension because it clarifies who belongs where. Never because it's fashionable.
Key takeaways for your product
- AI unlocked new levers to segment on — priority in the queue, early access, output ceilings, credits, even a meter split in two (build vs. run).
- Segmenting on features alone is probably a mistake — usage and the new levers often separate your customers better.
- Build each tier around a real segment, not a price point — and if you only have one segment, you only need one tier.
- A new lever must make the choice clearer, or it's just noise.
- One rule never changed, and it still validates your whole offer: if a customer can't tell which tier is theirs in ten seconds, your segmentation is wrong — not just your page.
- This isn't only for AI companies. Any product where each use costs you something real — compute, data, live capacity — has these levers, whether or not anyone's pulled them.
So, whatever you build:
What new value dimensions is your product already sitting on — that you haven't thought to price?
I teach how to find and price these dimensions in Product Strategy in the AI Era — going deep on segmenting your market and choosing the right value metric when features and seats no longer tell the whole story.