THE SHORT ANSWER
AI weakens per-seat pricing because fewer human users can now produce more work while providers incur real costs with each AI interaction. Hybrid models combine a predictable base fee with usage-based components. SaaS leaders should evaluate these models against both the value created and the cost of successful adoption.
At SaaS Kitchen, Dr. Sebastian Voigt from hy presented a shift that changes both sides of the software market. AI providers are increasingly testing pricing models beyond fixed seat licenses because their products now perform work and create real variable costs. Buyers, at the same time, expect AI to reduce the amount of human work they need.
Both can be true. But together they create a new question for every SaaS founder, CRO and CFO:
If we reduce team costs but pay more for AI usage, credits and outcomes, are we actually becoming more efficient?
Why AI is weakening per-seat SaaS pricing
Traditional SaaS pricing had a convenient logic. Customers paid for access, usually per user. The provider’s marginal cost stayed low as usage grew. Generative AI breaks that relationship. Every interaction can create compute costs, while agents may complete tasks without another human user ever logging in.
More users create more revenue. Usage is largely covered by a fixed subscription.
More automated work creates more revenue and often more cost for the provider.
The report shows how quickly the market expects this shift to happen:
plan or evaluate a pricing change within the next 12 months
expect hybrid pricing to be most relevant for their business in two years
expected change in the relevance of user or seat pricing
expected change in the relevance of outcome or result pricing
Source: SaaS & AI Pricing Report 2027. Survey results reflect the participating companies and their expectations.
Why AI pushes software vendors toward usage-based pricing
The report identifies four structural changes. AI changes the product, the provider’s cost structure, the unit of customer value and the role software plays in the workflow.
From feature to agent
Software no longer only helps a person complete a task. It increasingly performs the task itself.
From marginal cost to real usage cost
More prompts, inference, memory and orchestration can create additional cost for every customer interaction.
From access to delivered work
Pricing increasingly reflects completed tasks, workflows or results rather than access alone.
From interface to infrastructure
AI systems and agents may access software through APIs or the Model Context Protocol (MCP) without a person using the interface.
“Product, packaging and pricing must be understood as one monetization system.”
DR. SEBASTIAN VOIGT · HY CONSULTING GROUP
How usage-based and hybrid pricing change budget risk
Seat pricing made budgets relatively predictable. A company knew how many people needed access and could estimate annual cost. Usage models make the bill follow activity. Outcome models make it follow delivered value. Hybrid models combine a fixed base with a variable component.
That can be fairer. It can also make successful adoption more expensive.
Why outcome pricing will remain the exception for now
Outcome pricing sounds like the cleanest model because payment follows value. In practice, it only works when both sides can define the result, measure it consistently and agree how much of it the software actually caused.
The report therefore expects task, workflow and agent pricing to establish itself faster than pure outcome pricing. Work completed by an agent is often easier to count than revenue created, cost avoided or customer success influenced. The market is still working this out: 25 percent of surveyed companies have not yet defined their future pricing metric.
The next step after seats is not automatically outcome pricing. For many products, a hybrid model with measurable tasks or workflows is the more credible transition.
How SaaS providers should choose an AI pricing metric
The report shows that 50 percent of surveyed providers currently include AI in the core product, while 22 percent do not monetize it separately at all. Only 15 percent use a paid add-on and 5 percent use credits or consumption.
That may support adoption today, but it is risky when usage cost grows faster than revenue. The answer is not automatically a credit system. Credits can help providers connect revenue to variable cost, but they can also make the customer’s business case harder to understand.
The better metric is the one that meets three tests:
- Value: it grows when the customer receives more benefit.
- Cost: it tracks the provider’s variable economics closely enough to protect margin.
- Clarity: a buyer can forecast and verify the bill without an internal pricing analyst.
Faster delivery can justify a higher price
The same shift matters in professional services. When price is anchored to hours, faster delivery appears less valuable. When price is anchored to the result and time to value, the opposite can be true.
A six week project that prevents months of delay can be worth more than a twelve week project with the same slide count. The relevant question is not how long the provider worked, but what changed for the customer and how quickly.
Rewards effort
Rewards deliverables
Rewards business impact
Prices time to value
The same logic applies to software. Buyers should not expect AI to be cheaper merely because fewer human hours are involved. Providers should not expect customers to accept higher prices merely because AI is used. Both sides need a credible link between price and value.
Dominic’s take: calculate the system, not the payroll
The seductive equation is simple: fewer employees plus more AI equals lower cost. But it ignores five expenses that sit outside payroll:
- Software: subscriptions, credits, API calls and agent runs
- Implementation: integration, data preparation and workflow design
- Control: human review, compliance and exception handling
- Maintenance: prompts, models, routing logic and changing tools
- Failure: rework, bad decisions, lost trust and missed revenue
A company can reduce headcount and still end up with the same total cost, less internal knowledge and greater supplier dependency. That may still be a good decision if speed, scalability or output quality improves. But it is not a cost saving. It is a different operating model.
First decide what you are actually buying
Saved time is not automatically a saving. Before building an ROI model, classify the investment. Four business cases cover most AI investments.
Cost saving
A payroll position, planned hire, agency retainer or other budget line disappears or is avoided. If cash cost does not fall, do not book the saved hours as savings.
Capacity gain
The same team can handle more work or deliver faster. Total cost may rise because AI is added to the existing cost base. Judge it by throughput, cycle time and stable quality.
Growth investment
AI creates additional pipeline, conversions or gross profit that would not otherwise exist. The relevant proof is incremental commercial output, not activity volume.
Risk or quality improvement
AI reduces errors, compliance exposure or quality variance. Measure the incidents avoided, review effort saved or cost of failure reduced rather than inventing a productivity saving.
If you cannot name the budget line that shrinks, the business output that rises or the risk that falls, you do not have an AI business case yet. You have a productivity hypothesis.
Five questions SaaS leaders should ask before scaling AI
What increases the bill? Seats, prompts, credits, tasks, workflow runs, tokens, storage or outcomes?
Does cost scale with customer value? Or can usage rise without a meaningful business result?
What happens under successful adoption? Model the bill at three, ten and thirty times current usage.
Who carries error and rework cost? Clarify what counts as a completed task or valid outcome.
How reversible is the operating model? Understand data portability, workflow ownership and supplier dependency.
A practical decision rule
- the work is frequent and measurable
- quality can be checked efficiently
- usage cost grows slower than delivered value
- the workflow can survive a vendor change
- context and trust determine the result
- exceptions are more common than the standard case
- errors have high commercial or legal cost
- the work creates strategic learning you need to retain
The conclusion
AI is changing software from a tool people use into a system that increasingly performs work. Pricing therefore expands beyond access to include consumption, tasks, workflows and, where value can be attributed clearly, outcomes.
For providers, this creates new revenue models. For buyers, it creates a new responsibility: calculate total operating leverage before treating AI as a headcount strategy.
The winning model is not fewer people or more AI. It is the combination that produces the best result after every cost is counted.
ABOUT THE SOURCE
Dr. Sebastian Voigt & hy
Sebastian Voigt is Partner and Managing Director at hy Consulting Group. The SaaS & AI Pricing Report 2027 was created by hy, OMR Reviews and Appinio.