AI Software Pricing Models: Why Per-Seat Is Broken

AI software pricing models are changing faster than most businesses realise — and if you are still paying per seat, you may already be overpaying.

Here is an uncomfortable question to bring to your next budget review: how many of the software seats you are paying for right now are actually being used by a human?

Industry data consistently shows that a significant share of enterprise SaaS (Software as a Service — cloud-based software you subscribe to rather than install) licences sit idle at any given time — and AI is accelerating that trend. That is a significant amount of money leaving your business every month for software that nobody is logging into.

As AI agents take on more of the knowledge work that once required a licensed human user — answering support tickets, writing code, processing documents, generating reports — the per-seat model is starting to look less like a fair exchange and more like a legacy tax. In this article, we will walk you through why per-seat pricing is structurally misaligned with an AI-powered workforce, what the new models look like, and exactly what you can do today to stop overpaying.

Why Per-Seat Pricing Made Sense — Until AI Arrived

Per-seat pricing is a SaaS billing model in which a business pays a fixed recurring fee for each individual user account provisioned on a platform, regardless of how much — or how little — that account is used.

Per-seat pricing has been the dominant model in enterprise software for decades, and for a long time, it was completely logical. The underlying assumption was simple: software creates value through human usage. One person logs in, completes tasks, and generates output. The more people you have doing that, the more value flows through the system, and the more you pay.

This model worked well in a world where every meaningful action inside a software platform required a human hand on the keyboard. Tools like Salesforce, Jira, Zendesk, and Microsoft 365 were all built on this assumption. Pricing teams designed their models to scale directly with headcount — and finance teams at the buyer end appreciated the predictability. You hire ten more customer service agents, you add ten more Zendesk seats. Clean, auditable, easy to forecast.

The problem is that AI has snapped this relationship in two.

When an AI agent can open a support ticket, research the customer’s history, draft a personalised response, and close the ticket — all without a human touching it — the question of “how many seats do you need?” becomes almost meaningless. The AI does not need a seat. It does not log in. It does not have a profile in your user management console. Yet it is doing the work that used to justify those seat licences.

Gartner has consistently highlighted that AI-augmented development workflows are becoming standard across software engineering organisations — a trend that directly challenges the rationale for paying full per-seat rates for every developer licence when AI is completing a growing share of the actual work.

Key Takeaway: Per-seat pricing assumes one human equals one unit of value. AI breaks that assumption entirely — an AI agent can produce the output of multiple human seats without holding a single licence.

The Uncomfortable Truth: You Are Already Paying for Seats AI Is Filling

According to McKinsey’s 2024 State of AI report, 65% of organisations are now regularly using generative AI — up from just 33% a year prior. That is a near-doubling in adoption in twelve months. Businesses are deploying AI tools across customer service, content creation, data analysis, legal review, and software development at a rapid pace.

At the same time, most of those same businesses are still paying for the same volume of per-seat SaaS licences they had before AI arrived. Understanding how AI software pricing models work is now a competitive requirement, not a secondary procurement concern.

That means many organisations are effectively paying twice: once for the AI tool that does the work, and again for the human-oriented seat licence on the platform the AI is interacting with. In some cases, they are also paying a surcharge to access the AI features inside their existing platform.

Microsoft’s Copilot is a striking example. It is priced at an additional USD $30 per user per month on top of existing Microsoft 365 seat costs — meaning a 100-person organisation using Microsoft 365 E3 (USD $36/user/month) that adds Copilot across the team is spending an additional USD $36,000 per year just to access AI features on a platform it already pays for. Enterprises are being asked to pay a premium to access AI capabilities that, if they work as advertised, reduce the need for as many licensed users in the first place.

Intercom’s experience illustrates the scale of potential displacement. After deploying its AI agent Fin, Intercom reported that Fin resolved over 50% of customer support queries without any human involvement — absorbing the workload of what would previously have required multiple licensed human agent seats. In Intercom’s own words: “Fin isn’t just answering questions — it’s taking on the full resolution workflow that used to require trained human agents at scale.” Would your current support platform’s per-seat contract reflect that reality? Almost certainly not.

An analysis by OpenView Partners, a leading SaaS-focused venture capital firm, found that companies using usage-based pricing models grew revenue approximately 38% faster than those on pure per-seat structures — suggesting the commercial momentum has already shifted decisively toward more flexible models.

Key Takeaway: Organisations that have adopted AI tools are frequently paying for both the AI doing the work and the per-seat licence for the platform it works within — a structural double-cost that most SaaS contracts were never designed to address.

How AI-Native Software Pricing Models Are Replacing Per-Seat

While legacy SaaS vendors are largely defending their per-seat structures, a new generation of AI-native companies is building entirely different commercial models — and they are gaining traction fast.

The two AI software pricing models that are increasingly displacing per-seat billing are consumption-based and outcome-based pricing. Here is how each works:

Consumption-Based Pricing

Consumption-based pricing (also called usage-based pricing) is a billing model that charges customers based on the volume of resources or actions consumed — such as API (Application Programming Interface — the connection layer that lets software systems talk to each other) calls, tokens processed, or compute hours — rather than the number of users with access to the platform.

OpenAI’s API pricing is the most prominent example: as of 2025, GPT-4o is priced at USD $2.50 per million input tokens and USD $10.00 per million output tokens — meaning your cost scales precisely with the actual computational work done on your behalf. There are no seats, no user limits, and no minimum commitments beyond what your usage generates.

This model aligns vendor revenue directly with customer value. When you are getting more out of the product, you pay more. When usage drops, your costs drop automatically. For finance teams used to fixed headcount-linked software costs, this requires a meaningful shift in how you budget — but it eliminates the problem of paying for idle capacity.

Research consistently finds that cloud and SaaS cost optimisation ranks as one of the top priorities for enterprise technology leaders — a direct response to the cost unpredictability that consumption-based models can introduce if not managed carefully.

Outcome-Based Pricing

Outcome-based pricing is a billing model in which vendors charge for measurable, agreed-upon results — such as tickets resolved, documents reviewed, or code functions completed — rather than for access to the platform or the volume of usage it generates.

Intercom has moved in this direction, structuring parts of its pricing around AI resolution rates rather than pure seat count. Harvey AI, which provides AI-assisted legal work, charges based on the volume of work completed rather than the number of lawyers with access. Cognition’s Devin — an AI software engineering agent — charges by Agent Compute Unit (ACU), where 1 ACU equates to approximately 15 minutes of active work, rather than by the number of developer seats provisioned.

Cursor, the AI code editor, has attracted significant developer adoption for its approach: flat-rate access to AI-assisted coding without the per-seat escalation that tools like GitHub Copilot apply at enterprise scale. Leading voices in AI infrastructure have noted that the companies winning in this space are aligning their pricing with what customers actually care about — outcomes, not access.

These AI software pricing models are not fringe experiments. AI and AI-adjacent software companies have attracted enormous venture capital interest, with the broader AI category commanding more investment than the traditional SaaS market did just a few years ago.

Talk to our AI services team about which pricing model makes sense for the tools you are already using — or evaluating.

Why Your Existing SaaS Vendors Are Slow to Change

If outcome-based and consumption-based pricing better reflect the value AI delivers, why are the big SaaS vendors not rushing to adopt them?

The answer is straightforward: per-seat pricing is extraordinarily profitable for incumbent vendors, and they have little financial incentive to change until competitive pressure forces them to.

The global SaaS market is valued in the hundreds of billions of dollars and is projected to grow substantially through the early 2030s, according to Fortune Business Insights. A significant portion of that value is built on predictable, recurring per-seat revenue. Salesforce, for example, reported total revenue of USD $34.9 billion in its FY2024 earnings — revenue built substantially on seat-based enterprise contracts. Moving to outcome-based pricing would mean betting that usage grows faster than seat counts shrink: a structural risk most public SaaS companies are unwilling to take voluntarily.

What they are doing instead is bolting AI features onto existing per-seat products and charging a premium for access — Salesforce’s Einstein AI, Microsoft’s Copilot, and Atlassian’s AI features in Jira and Confluence. Research widely cited in the enterprise software sector shows strong consensus among IT leaders that AI is essential to organisational competitiveness — a perception that incumbent vendors use to justify premium AI add-ons rather than to question the underlying per-seat structure.

As Kyle Poyar, formerly Operating Partner at OpenView and a leading voice on SaaS pricing strategy, put it: “The vendors most at risk are the ones treating AI as a feature add-on to existing seat contracts. Buyers are starting to ask why they should pay more per seat for AI that reduces the number of seats they need.”

This creates a pricing arbitrage opportunity for buyers who are paying attention. If an AI-native competitor can deliver the same outcome for a fraction of the cost — and charge you only for what you actually use — the incumbent’s bundled AI surcharge starts to look like a poor deal. Most procurement teams are not yet making this comparison systematically, but those that are finding significant savings.

Key Takeaway: Legacy SaaS vendors are using AI as justification to charge more per seat — not as an opportunity to align pricing with the value AI actually delivers. Buyers who recognise this dynamic have real negotiating power.

Comparing AI Software Pricing Models: Consumption-Based vs. Outcome-Based

These two models are often discussed interchangeably, but they have meaningfully different implications for your business.

Neither consumption-based nor outcome-based pricing is universally superior — the right choice depends on how predictable your usage is and how clearly you can define “success” for a given AI task.

For most Australian SMEs and mid-market businesses, consumption-based pricing is the more accessible entry point among these AI software pricing models. It is transparent, directly measurable, and easy to compare against the cost of the seats it might be replacing. Outcome-based pricing makes more sense for high-value, well-defined workflows — like AI customer support resolution or AI-assisted document review — where the output can be clearly measured and priced.

Research from enterprise software buyers consistently shows strong preference for usage-based or outcome-based pricing models, even as the majority of incumbent SaaS vendors have yet to offer them at scale — highlighting the gap between buyer preference and current market reality.

What AI Software Pricing Models Mean for Your Budget

For CFOs and finance teams, the shift in AI software pricing models is not just a procurement question — it is a budgeting and forecasting challenge.

Per-seat contracts are easy to model: you know your headcount, your seat count, and you pay a fixed amount each month. Consumption-based pricing introduces genuine variability — your software costs now fluctuate with your AI usage, which can shift significantly based on seasonal demand, project load, or how aggressively your team adopts new workflows.

Organisations that have transitioned major SaaS platforms to usage-based pricing often report meaningful reductions in software spend — though many also report periods of unexpected cost spikes during the transition, making proactive spend monitoring essential.

Here is how to manage this transition:

1. Set usage budgets, not seat budgets. Work with your AI tool vendors to establish monthly consumption caps or alerts that trigger a review before spend escalates.
2. Monitor cost-per-outcome, not cost-per-seat. Calculate what each resolved support ticket, generated document, or code review costs under AI pricing — and compare it against your previous per-seat equivalent.
3. Build a 12-month rolling forecast model that accounts for usage growth as AI adoption increases internally. Do not assume your AI spend will stay flat as your team’s comfort with the tools grows.
4. Negotiate hybrid structures where possible. Some vendors offer a base commitment combined with consumption pricing above that floor — giving you cost certainty without paying for unused capacity.
5. Benchmark against industry data. Flexera’s annual State of Tech Spend report and Vendr’s SaaS Benchmarks provide median per-seat costs by software category — invaluable reference points when entering renewal negotiations.

Real-World Examples: Companies Already Breaking the Per-Seat Model

These examples illustrate how AI software pricing models are already reshaping costs across industries:

• Harvey AI (legal): Charges law firms based on work volume processed, not attorney seats provisioned. Early adopters include Allen & Overy (now A&O Shearman), which deployed Harvey across its global legal teams rather than scaling per-seat licences for document review tools.
• Cognition’s Devin: An AI software engineering agent priced per Agent Compute Unit (ACU), where 1 ACU equates to approximately 15 minutes of active work — directly analogous to billing for a contractor’s time rather than a software seat. Cognition has positioned Devin as a replacement for entry-level engineering licence costs, not an add-on to them.
• Cursor: Flat-rate AI-assisted coding that is attracting developers away from per-seat alternatives. The company has reported rapid growth in its user base, reflecting strong developer appetite for alternatives to traditional per-seat code tooling.
• Intercom + Fin: Restructured pricing to reflect AI resolution volume, so customers whose AI handles more queries pay in a way that reflects actual usage. Intercom has reported that customers using Fin at scale achieve substantial reductions in per-resolution costs compared to human-agent handling.

The pattern is consistent: AI-native vendors are designing pricing that reflects what AI actually does, rather than maintaining the fiction that every interaction requires a human with a licence.

Your Action Plan: Audit, Negotiate, and Future-Proof Your SaaS Stack

You do not need to wait for the market to sort itself out. Here is a practical framework you can start using this week to align your SaaS costs with the realities of today’s AI software pricing models.

Step 1: Run a Seat Utilisation Audit

Pull your SaaS licence data and compare provisioned seats against monthly active users (MAU) for each tool. Most SaaS platforms expose this in admin dashboards. According to Productiv’s 2024 SaaS data, only around 45% of users are actively engaging with a given SaaS application on average — meaning more than half of licences sit unused. Any tool where MAU is consistently below 70% of provisioned seats is an immediate target for renegotiation or reduction.

Step 2: Map AI Displacement Risk

For each tool, ask: Is an AI tool we already have or are evaluating capable of performing the tasks this seat was provisioned for? Create a simple spreadsheet:

• Tool name
• Number of seats currently paid for
• Monthly cost (AUD)
• Tasks performed per seat
• AI alternative available? (Y/N)
• Estimated seat reduction potential (%)

Step 3: Prioritise High-Cost, High-Displacement Tools

Focus your renegotiation energy on tools where the combination of high per-seat cost and high AI displacement potential is greatest. Based on Vendr’s 2024 SaaS Benchmarks data, the categories with the highest median per-seat cost and fastest-growing AI alternative penetration are: customer support platforms, sales intelligence tools, document management software, and basic project management applications.

Step 4: Research AI-Native Alternatives

Before your next contract renewal, identify whether an AI-native competitor with outcome-based or consumption-based pricing exists. Compare total cost of ownership — not just the headline per-seat price, but what you actually spend versus what you actually use. Evaluating AI tools on a “cost per completed outcome” basis rather than a feature comparison is increasingly recommended by analysts and procurement specialists.

Step 5: Renegotiate on the Back of Utilisation Data

Go into renewal conversations armed with your utilisation data. If 35% of your seats are underused, you have real power to negotiate a seat reduction, a price concession, or a move to a more flexible pricing structure. Vendors prefer to renegotiate on your terms than to lose the contract entirely — and buyers who enter renewal negotiations with documented utilisation data consistently achieve better pricing outcomes than those who do not.

Step 6: Evaluate New AI Tools on Commercial Terms, Not Just Features

When assessing any new AI tool, treat the pricing model as a primary evaluation criterion — not an afterthought. Ask vendors: How does your pricing scale as we deploy AI agents? What happens to our cost structure if AI handles 60% of the use cases we currently have humans managing?

Get an independent assessment of your AI tool options from our team — we help businesses cut through the noise and choose solutions that make commercial sense, not just technical sense.

FAQs About AI Software Pricing Models

What is wrong with per-seat pricing for AI software?

Per-seat pricing was designed for a world where every meaningful software action required a licensed human user. When AI agents can perform those same actions without occupying a seat, the model breaks down — you end up paying for user licences even though the work is being done autonomously. Industry research consistently shows that a large share of enterprise SaaS seats are already sitting idle, and AI displacement is accelerating that trend.

What are the main AI software pricing models replacing per-seat billing?

The two main alternatives are consumption-based pricing (you pay based on actual usage, such as API calls or tokens processed) and outcome-based pricing (you pay per completed task or measurable result). AI-native companies like OpenAI, Intercom, Harvey AI, and Cursor are leading the shift toward these models, with the AI software sector attracting significant venture capital investment in recent years.

How do I know if I am overpaying for SaaS seats because of AI?

Start by comparing your provisioned seat count against monthly active users for each tool. According to Productiv’s 2024 SaaS data, only around 45% of users actively engage with a given SaaS application on average — a significant signal of idle spend. If you are also deploying AI tools that absorb work previously done by human seat holders, you are almost certainly overpaying.

Will Microsoft and Salesforce ever move away from per-seat pricing?

Not voluntarily, and not quickly. Salesforce reported total revenue of USD $34.9 billion in FY2024 — substantially from seat-based enterprise contracts. The more likely path is competitive pressure from AI-native alternatives forcing incremental changes: more flexible tiers, consumption add-ons, or outcome-linked discounts rather than a clean break from per-seat billing.

What is outcome-based pricing and how does it work in practice?

Outcome-based pricing is a billing model in which vendors charge for measurable, agreed-upon results rather than for access or usage volume. For example, Intercom’s AI agent Fin charges based on the volume of support queries resolved autonomously — so customers pay proportionally to the value delivered, not to the number of agent seats provisioned.

How should my finance team budget for consumption-based AI pricing?

Build a rolling 12-month usage forecast based on your current AI adoption rate and expected growth. Set monthly spending caps with your vendors and track cost-per-outcome as your primary metric. Organisations transitioning to usage-based pricing commonly report meaningful software cost reductions in year one, though unexpected cost spikes during the transition are also common — making proactive spend monitoring essential.

The Bottom Line

Per-seat pricing is not going to disappear overnight. But it is increasingly misaligned with how AI-powered organisations actually work — and the businesses that understand AI software pricing models early will have a meaningful commercial advantage over those still paying legacy rates for idle seats.

The shift is already happening. McKinsey reports that 65% of organisations now regularly use generative AI (2024). AI-native vendors have attracted unprecedented levels of venture capital investment. Intercom’s AI is resolving more than half of all support queries without human involvement. These are not projections — they are current market realities that most SaaS contracts have not caught up with yet.

The businesses that treat software procurement as a strategic function — benchmarking utilisation, comparing AI-native alternatives, and entering renewals with data — are finding that the SaaS cost line most organisations treat as fixed is, in fact, highly negotiable.

Are you confident your current SaaS spend actually reflects the value you are getting — or are you funding seats that AI has already made redundant? Book a free consultation with our team to talk through how AI tools can reshape your digital operations, and what that means for the software costs that support them.