Most SaaS tools charge per seat. One user, one fee — clean, predictable, easy to justify in a budget meeting. It worked because software was a multiplier on human effort: one person using it still meant one person’s hours of work. That assumption is now breaking.
The Logic That Built Per-Seat Pricing
When Salesforce launched in 1999, the premise was simple: your sales team uses this, your sales team pays for it. The value was proportional to the number of humans who had access. Per-seat pricing reflected that perfectly — it was a proxy for usage because usage was tied to people.
Even as SaaS matured, the model held. Slack charges per active user. Notion charges per member. Figma charges per editor. The logic is consistent: more humans in the seat, more value extracted, the higher the bill.
But AI has introduced a third party into that equation.
What AI Does to the Math
When an AI agent can do the research, write the first draft, analyze the data, and send the follow-up — all without a human “using” the software in the traditional sense — per-seat pricing starts to feel arbitrary.
Consider a concrete case. A startup replaces three junior analysts with a single AI pipeline. They now need one Notion seat, one Slack seat, one data platform account. But they’re extracting more value than ever. The SaaS vendors are getting less money for more output delivered. That math does not hold long-term.
The vendors know this. That’s why the models are shifting.
Usage-Based Pricing
Usage-based pricing (UBP) charges you for what you actually consume — API calls, tokens processed, rows queried, emails sent. OpenAI does this. Snowflake has always done this. Twilio built its entire business on it.
The appeal: you pay proportionally to how hard the system works, not how many people access it. For AI-heavy workflows where one engineer running 10,000 AI calls a day generates significant value, this is fairer than paying for a headcount that no longer reflects the work.
The trap: unpredictability. A misconfigured loop, an unexpected traffic spike, a product launch going viral — and your bill triples. Finance hates variance. For teams that need budget predictability, UBP requires spend monitoring from day one, not as an afterthought.
Value-Based Pricing
Value-based pricing is the purest form of the idea: you pay for outcomes, not inputs. If the software saves you $50,000 a year in labor, it charges a percentage of that. If it generates revenue, it takes a cut.
This is already happening in pockets. Some AI legal tools are experimenting with outcome pricing for contract review. Sales AI platforms are starting to tie pricing to pipeline generated. It’s messy to implement, but directionally it’s where things are heading as the gap between “access” and “output” widens.
The problem: verifying outcomes is hard, contested, and creates adversarial dynamics. How do you prove the tool generated the deal and not the rep? How do you audit the savings? Until measurement becomes more reliable, value-based pricing will remain niche — used in enterprise contracts more than product-led growth.
What This Means If You’re Buying SaaS
If you’re evaluating tools right now, ask a question most buyers skip: does this pricing model reflect how we’ll actually use it?
If you’re adopting an AI tool to automate work that used to require headcount, per-seat pricing might mean you’re temporarily under-paying — until vendors reprice at renewal. Usage-based might fit better, but forecast your usage before you commit.
Three practical checks:
- Set spend alerts immediately. If the tool is usage-based, cap your exposure from day one. Most platforms offer this natively. Most buyers ignore it until the first shock bill.
- Audit idle seats quarterly. If you’re on per-seat, run a usage report. SaaS seat creep is real — you’re almost certainly paying for people who barely log in.
- Ask about flat-rate tiers. Many SaaS vendors will negotiate flat-rate or hybrid contracts for higher-spend customers. It is not always listed on the pricing page, but it is almost always on the table.
What This Means If You’re Building SaaS
If you’re building a product, your pricing model is a product decision, not a finance decision. It shapes retention, expansion, and how customers perceive your value over time.
The shift toward AI-delivered outcomes makes this harder. If your product does ten times more per user because of AI, per-seat pricing lets customers under-pay you as their usage grows. But usage-based pricing shifts unpredictability to them — and some will churn rather than absorb variable bills.
The emerging answer most SaaS builders are landing on: a hybrid. A base subscription for predictability, plus usage overage or outcome tiers for upside. It’s messier to explain on a pricing page, but it aligns incentives better than either model alone.
One thing to avoid: pricing your AI features identically to your human-workflow features. If your AI can do in two minutes what used to take a junior analyst two days, charge for the two-day value, not the two-minute compute cost. The buyer doesn’t care about your inference bill.
What to Do Next
If you’re a founder or product owner evaluating your SaaS stack:
- Map your current spend to actual usage. Run the report — you’ll find seats you’re not using.
- Before renewing any AI-heavy tool, ask the vendor what their pricing roadmap looks like. Expect it to change within the next contract cycle.
- If you’re building a product with AI in it: price against the value you deliver, not the compute you consume.
Per-seat pricing will not disappear overnight — large enterprises run on multi-year contracts, and inertia is stubborn. But the direction is clear. AI does not sit in a seat. It runs in the background, produces outputs at scale, and does not log in nine to five. Pricing models will follow.
If you’re rethinking how AI fits into your product or your stack, let’s talk.