The Model That Built an Industry
For two decades, per-seat pricing was the gold standard in B2B software. Salesforce, Slack, Notion, Atlassian, Microsoft 365 – they all run on the same principle: you pay per user per month. More employees in the tool means more revenue for the vendor. Cleanly linear, easy to budget, predictable for both sides.
AI is breaking this model. And it's happening faster than most SaaS vendors want to admit.
Why Per-Seat Worked in the First Place
Per-seat pricing was never a true value mapping – it was a pragmatic approximation. The underlying assumption: more users means more value created. A 50-person sales team generates more business than a five-person team, so it justifies higher software costs. The assumption was never perfect, but it was good enough.
The real reason for the model's success was simplicity. Per-user pricing is transparent, predictable and fits into any budget spreadsheet. Procurement can understand it without deep product knowledge. That frictionlessness is what made SaaS the dominant distribution model.
What AI Changes
AI breaks the connection between user count and value creation in two ways:
One user can now do the work of ten. A five-person sales team using AI-powered research, outreach generation and pipeline analysis covers volumes that used to require 25 people. When software multiplies output per head by five, charging per head no longer makes sense. The customer pays for fewer seats but gets more value – and the vendor's revenue collapses.
Agents are not users. A growing share of work in modern companies is done by autonomous agents: code reviews, data preparation, reporting, first-line support. These agents have no email address and need no login – but they consume resources and produce value. Per-seat doesn't capture them.
The unavoidable consequence: per-user prices become either unaffordable (if you count every agent as a seat) or massively undervalued (if you ignore them).
The Models That Are Coming
The industry is already experimenting with alternatives, and several patterns are emerging:
Outcome-based pricing: You pay per resolved support ticket, per qualified lead, per completed workflow. Intercom was first to take this step with "Fin" – charging per resolved ticket instead of per agent. The model will spread because it directly mirrors value contribution.
Usage-based pricing: Like cloud infrastructure. You pay for what you use – API calls, tokens, storage, compute. Anthropic, OpenAI, Twilio and AWS live on it. The advantage: honest scaling. The downside: hard to plan, hard to procure.
Hybrid models: A platform fee plus consumption-based components. A base fee covers provisioning, the variable part reflects actual usage. Probably the pragmatic middle ground most B2B vendors will end up at.
Value-based contracts: Enterprise deals where vendor and customer jointly define success metrics and pricing follows them. Rare so far, but on the rise – particularly in strategic AI implementations.
The Problem No One Is Solving
All these models share one weakness: they're harder to sell.
Per-seat pricing works because a buyer can calculate the cost in two seconds. "150 employees times £25 times 12 months – that's £45,000 per year." Usage-based models require estimates of future consumption; outcome models require agreement on which outcomes to measure. Both increase friction in the sales process.
That's exactly why many SaaS vendors cling to per-seat – not because the model still fits, but because their sales team is trained on it and the pricing page has looked the same for five years. That inertia will cost them market share.
What This Means for Buyers
If you're buying SaaS today, two things matter:
Prefer short contract terms. The next two years will reshape pricing models. Locking in a three-year deal for 200 seats now means sitting on a licence structure in 2027 that no longer matches reality.
Pay attention to consumption metrics. Even if a vendor charges per user today, ask for the underlying usage data. API calls, AI operations, storage volume. These metrics will matter in future negotiations.
Rethink build-vs-buy. When an AI-powered SaaS tool costs £200 per user, the make case deserves another look. Building your own workflows directly on foundation models has become cheaper – and avoids the pricing risk of external vendors.
What This Means for Vendors
Anyone selling SaaS needs to ask honestly: what value is actually created per user – versus per operation, per workflow, per outcome? If the answer is "more per outcome than per user", per-seat is a ticking time bomb. Competitors will attack exactly where the model distorts value.
At nh labs, we see this shift in every other consulting conversation. Customers who used to stack SaaS licences without question are now asking about consumption transparency, contract flexibility and in-house build options. This isn't a passing trend – it's the new normal.
Conclusion
Per-seat pricing won't disappear overnight. It will erode slowly, contract by contract, use case by use case. Vendors that switch to hybrid or outcome-based models in time will secure the next generation of customer relationships. The rest will discover, over the next three to five years, that their most loyal customers are suddenly cancelling seats – not because they no longer need the product, but because they need fewer people to do the same work. That's the real AI shock for the SaaS industry. And it has only just begun.