The Future of SaaS Pricing Is Outcomes, Not Seats

For most of my career, SaaS pricing has been built around a simple commercial assumption:

More users equal more value.

That assumption powered an entire generation of enterprise software.

It shaped CPQ systems, sales compensation plans, renewal motions, discounting models, expansion playbooks, board metrics, and customer negotiations.

We counted seats. We packaged modules. We negotiated tiers. We defended price per user. We celebrated net retention when more employees logged into the system.

That model worked because the human user was the primary unit of software consumption.

AI changes that.

Not because SaaS is going away. I do not believe SaaS is dead. But the old pricing architecture of SaaS is under pressure because the way work gets done is changing.

In the AI era, software is no longer just a place where humans go to perform tasks.

Software is becoming a system that agents access, reason over, and act through.

The interface may be a chat window. The user may be an AI agent. The value may be a completed workflow, a resolved ticket, a qualified lead, a closed compliance gap, a fixed vulnerability, or a decision made faster with better context.

That is why I believe:

The future of SaaS pricing is outcomes, not seats.

And just as importantly:

Every SaaS company will become a data company.

Not because every company needs to sell data externally. But because the most valuable SaaS companies in the AI era will be the companies that own, refine, govern, and activate the proprietary data and workflow context that AI needs to deliver trusted outcomes.

A few years ago, I wrote that raw data is not the real oil. Information is.

Raw data, like crude oil, has limited value until it is refined, cleaned, structured, made fresh, and turned into something useful.

That point matters even more now.

In the AI era, context is the refinery.

Outcomes are the fuel.

SaaS is not dead. But old SaaS pricing is exposed.

There is a lot of noise right now around “the end of SaaS.”

I think that framing is too dramatic and, frankly, not very useful for operators.

SaaS will not disappear. Enterprises still need systems of record. They still need security, governance, auditability, role-based access, workflow controls, compliance, integrations, uptime, support, and accountability.

What will disappear is the assumption that customers should pay primarily because a person has access to a screen.

The seat model made sense when software was a destination.

It makes less sense when software becomes infrastructure for work done by people and agents together.

If an AI agent can do the work of five users, why would the customer want to pay for five seats?

If a support platform resolves a customer issue autonomously, the buyer does not care how many support agents logged in. The buyer cares whether the issue was resolved, whether the customer was satisfied, whether the resolution was accurate, and whether escalation was avoided.

If a sales platform qualifies leads automatically, the CRO does not want to buy more seats. The CRO wants more qualified pipeline.

If a cybersecurity platform identifies, prioritizes, and helps remediate vulnerabilities, the CISO does not want another dashboard. The CISO wants reduced exposure, faster response, and better assurance.

This is the commercial shift.

The buying question moves from:

“How many people need access?”

to:

“What work will this system complete, improve, or accelerate?”

That one shift breaks a lot of traditional SaaS pricing logic.

The market is already giving us clues.

This is not theoretical anymore.

Bain’s Technology Report 2025 was very direct: generative and agentic AI are disrupting SaaS by automating tasks and replicating workflows. Bain’s recommendation to SaaS leaders was equally direct: own the data, lead on standards, and “price for outcomes, not log-ons.”

Gartner’s May 2026 AI spending forecast shows the scale of what is happening. Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, up 47% year over year.

McKinsey’s 2025 State of AI research shows the adoption gap clearly. In its latest survey, 88% of respondents said their organizations regularly use AI in at least one business function, 62% said their organizations are at least experimenting with AI agents, and yet most organizations have not yet scaled AI deeply enough to realize enterprise-level value.

That matters for CROs.

The first wave of AI buying was experimentation.

The next wave will be accountability.

Customers will ask:

What did the AI actually do? What business process improved? What cost came out? What risk went down? What revenue went up? What should I pay for that outcome?

That is where pricing changes.

Nikesh Arora’s warning should get every SaaS executive’s attention.

Nikesh Arora, CEO of Palo Alto Networks, recently made a sharp point on the All-In Podcast.

Talking about analytical SaaS, he said, “If you’re an analytical SaaS company, it’s over.” His broader point was not that every SaaS company disappears. His point was that a company whose value proposition is simply “give us your data and we will analyze it for you” is now exposed because customers can increasingly run models against their own data.

That is a very important distinction.

Analytical SaaS without proprietary context is vulnerable.

Workflow SaaS without control points is vulnerable.

Reporting SaaS without action is vulnerable.

Horizontal SaaS without differentiated data is vulnerable.

But systems that own trusted data, critical workflows, regulatory logic, and deep domain context are not automatically dead.

In many cases, they become more valuable.

Nikesh also described Palo Alto Networks testing AI against code vulnerabilities and finding in six weeks what would normally have taken five to seven years to discover.

Think about the commercial implication.

If AI can compress years of work into weeks, how do we price that?

Not by seat.

Not by login.

Not by dashboard view.

Not by number of analysts who touched the tool.

The value is in the acceleration, accuracy, risk reduction, and business outcome.

That is the new pricing conversation.

The SaaS value stack is being rebuilt.

For the last 20 years, SaaS value was often packaged around applications.

CRM. ERP. HCM. ITSM. CPQ. Customer support. Marketing automation. Cybersecurity. Analytics. Collaboration.

Each application became a destination.

A human user opened the application, performed tasks, updated records, ran reports, and moved work forward.

AI starts to rebundle that world.

Bain describes a new SaaS disruption pattern where agentic AI can automate tasks, replicate workflows, and change where control sits in the software stack.

That is a profound shift.

The customer may no longer start in the CRM.

They may start with an agent.

The employee may no longer open five applications.

They may ask for an outcome.

The agent may call multiple systems behind the scenes.

The value may accrue to whoever controls the data, workflow logic, trust layer, and final outcome.

This is where large AI labs and hyperscalers start chipping away at SaaS TAM.

OpenAI is not just offering a model. It has introduced apps inside ChatGPT and an Apps SDK, allowing users to interact with third-party apps directly inside ChatGPT. OpenAI has also introduced workspace agents in ChatGPT that can help teams execute workflows across tools, data, and approvals.

That means the AI interface can become the place where work begins.

When that happens, traditional application vendors face a hard question:

Are we still the system of engagement, or are we becoming the back-end database that an AI agent calls when needed?

That difference matters because margins often sit closest to the user experience and the decision point.

MCP is a quiet but important part of this story.

CROs do not need to become protocol experts.

But they do need to understand why MCP matters.

Anthropic introduced the Model Context Protocol in November 2024 as an open standard for connecting AI assistants to the systems where data lives, including content repositories, business tools, and development environments.

In plain English, MCP helps AI agents connect to business systems and data sources in a more standardized way.

That matters because AI needs context.

Without context, AI is a smart intern with no access.

With context, permissions, tools, and business data, AI becomes a worker that can take action.

For SaaS companies, this creates both opportunity and risk.

The opportunity: your product can become easier for agents to access and use.

The risk: if your product is only a thin workflow layer on top of data that can be accessed elsewhere, your value can be bypassed.

This is why pricing will increasingly include data access, API usage, agent actions, connector consumption, and outcome delivery.

The commercial packaging will not be just:

“Here are your seats.”

It will become:

“Here is your platform access. Here is your data entitlement. Here are the workflows agents can execute. Here are the consumption limits. Here are the outcomes we commit to. Here is how we measure them. Here is what happens when volume exceeds the plan.”

That is a very different deal desk motion.

Why the seat model breaks in the AI era.

Seat-based pricing has three major weaknesses in an AI-first world.

First, the seat model assumes humans are the primary workers.

That assumption is already changing.

AI agents can summarize, classify, draft, analyze, triage, route, reconcile, detect, enrich, and recommend.

In many workflows, a smaller number of humans will supervise more automated work.

If the vendor charges per human seat, vendor revenue can decline as the customer becomes more productive.

That creates a strange incentive: the vendor only grows when the customer adds more people, not when the customer gets more efficient.

That is misaligned.

Second, the seat model does not capture value created by automation.

Imagine an AI support agent resolving 100,000 customer issues.

If the price is based only on human support seats, the vendor may create enormous value and capture very little of it.

That is bad pricing.

Third, the seat model becomes harder to defend in procurement.

Procurement teams are getting smarter about AI.

They will ask whether the pricing metric expands with value or simply taxes access.

They will compare vendors based on cost per resolution, cost per action, cost per qualified lead, cost per workflow, cost per risk reduced, and cost per hour saved.

The old metric of “number of named users” will not disappear overnight, but it will become less central.

Seats may remain as an access control, packaging boundary, or enterprise governance layer.

But seats will not be the main value metric for AI-native SaaS.

The new model: platform + consumption + outcomes.

I believe the dominant SaaS pricing model in the AI era will be hybrid.

Not pure usage.

Not pure subscription.

Not pure outcome-based.

Not pure token pass-through.

The winning model will likely combine five layers.

1. Platform access

This is the base subscription.

It gives the customer access to the product, security model, admin controls, integrations, workflow configuration, audit logs, reporting, support, and governance.

This layer creates predictability for the vendor and the customer.

For enterprise buyers, predictability still matters.

No CFO wants a completely open-ended AI bill.

No CRO wants a renewal conversation where the customer feels surprised by usage.

No deal desk wants to explain uncontrolled overages after the fact.

So the platform fee remains important.

But it should not be the whole model.

2. Data access and data entitlements

This becomes a major monetization lever.

Which data can the customer access?

How fresh is it?

How complete is it?

How trusted is it?

Can it be used by agents?

Can it be exported?

Can it train customer-specific models?

Can it be used through APIs or MCP connectors?

Does it include enrichment, normalization, scoring, lineage, and confidence levels?

This is where every SaaS company becomes a data company.

The value is not raw database rows.

The value is curated, permissioned, governed, contextual information that helps AI produce better decisions and actions.

In my view, the next generation of SaaS pricing will increasingly distinguish between application access and data value.

Application access gets you into the product.

Data access determines what intelligence you can unlock.

3. Agent or workflow consumption

This is where pricing starts to follow work.

How many actions did the agent perform?

How many workflows did it execute?

How many cases did it triage?

How many documents did it review?

How many alerts did it investigate?

How many invoices did it reconcile?

How many quotes did it generate?

How many supplier risks did it score?

Salesforce’s Agentforce pricing already shows the direction. Salesforce introduced Flex Credits, priced at $500 per 100,000 credits, with one Agentforce action consuming 20 Flex Credits, or $0.10 per action.

That is not traditional SaaS seat pricing.

That is a move toward pricing digital labor.

4. Outcome-based pricing

This is the most important layer and the hardest to operationalize.

Intercom’s Fin AI Agent is publicly priced from $0.99 per outcome.

HubSpot moved Breeze Customer Agent and Breeze Prospecting Agent to outcome-based pricing in April 2026, charging $0.50 per resolved conversation and $1 per lead recommended for outreach. HubSpot’s own language is clear: AI should be measured in outcomes, not output.

Zendesk has also moved toward outcome-based pricing for AI agents, focusing on automated resolutions and tying cost directly to value received.

This is the signal.

Outcome pricing is no longer a theory buried in pricing strategy decks.

It is entering production in real categories.

The challenge is measurement.

What counts as a resolved case?

What counts as a qualified lead?

What counts as a remediated risk?

What counts as a successful workflow?

Who validates it?

What happens when the customer disagrees?

How do you prevent gaming?

How do you handle partial success?

How do you price high-complexity outcomes versus simple ones?

This is where commercial operations, product telemetry, legal terms, customer success, finance, and deal desk all need to work together.

Outcome pricing is not just a SKU change.

It is an operating model change.

5. Premium governance, security, and assurance

As AI takes more actions, customers will pay for trust.

They will need controls around data access, explainability, audit trails, approval workflows, model governance, risk scoring, human-in-the-loop policies, and compliance evidence.

This is especially true in regulated industries, cybersecurity, finance, healthcare, energy, critical infrastructure, and public sector environments.

The more AI does, the more governance matters.

A vendor that can prove not only that the AI completed the work, but that it completed the work safely, securely, and auditable, will be able to defend premium pricing.

Tokens are not the value metric. They are the cost structure.

There is a lot of discussion around AI tokens.

Tokens matter.

They drive cost.

They affect gross margin.

They need to be tracked, governed, optimized, and forecasted.

But customers do not want to buy tokens.

Customers want work done.

In the cloud era, customers did not really want to buy CPU cycles.

They wanted applications, uptime, elasticity, and business capability.

Over time, sophisticated buyers learned to manage cloud consumption, but the business buyer still cared about the outcome.

AI will follow a similar pattern.

Tokens will matter to finance, product, engineering, and AI FinOps.

Tokens will matter to margin management.

Tokens will matter to packaging design.

But tokens are not the best customer-facing value metric for most SaaS categories.

The better value metrics will be closer to the business result:

Resolved issues. Qualified opportunities. Generated quotes. Completed reconciliations. Reduced risk. Prevented incidents. Hours saved. Revenue recovered. Compliance gaps closed.

This is where CROs need to be careful.

Passing through token costs may protect margin in the short term, but it does not create a strong value story.

It makes the customer feel like they are buying infrastructure.

The better move is to understand token economics internally, then package externally around value.

What this means for CROs and revenue leaders.

This shift will hit revenue leaders faster than many product teams expect.

A pricing change is not just a price book update.

It changes how the company sells, forecasts, negotiates, renews, compensates, and reports.

Here are the commercial questions every CRO should be asking now.

What is our real value metric?

Not the metric we inherited.

Not the metric that makes billing easy.

Not the metric that investors are used to seeing.

What metric expands when the customer receives more value?

If the product reduces manual labor, a seat may be the wrong metric.

If the product automates workflows, completed tasks may be better.

If the product improves decision quality, verified decisions or risk reductions may be better.

If the product enriches AI agents, data calls or context usage may be better.

If the product resolves issues, outcomes may be better.

The value metric is the heart of pricing.

Get it wrong, and everything downstream becomes harder.

How will we define and prove outcomes?

Outcome pricing sounds attractive until the first customer disputes the invoice.

This is where instrumentation matters.

Revenue leaders will need product telemetry that can prove what happened.

They will need customer-facing dashboards.

They will need definitions in contracts.

They will need dispute workflows.

They will need auditability.

An outcome is not a marketing slogan.

It is a billable event.

The deal desk needs to know exactly how it is measured.

How do we protect customers from bill shock?

Consumption-based pricing can create expansion, but it can also create fear.

Customers want flexibility, but they also want predictability.

This means vendors need usage caps, alerts, prepaid pools, rollover rules, volume tiers, rate cards, ramp schedules, and clear overage terms.

AI pricing without usage transparency will create renewal friction.

How do we compensate sales teams?

This is a big one.

Sales compensation built around upfront ARR does not always map cleanly to usage or outcome-based pricing.

Do reps get paid on committed consumption?

Actual usage?

Outcome volume?

Expansion?

Gross margin?

Customer value delivered?

Renewal quality?

If compensation rewards the wrong behavior, sales teams will sell the wrong structure.

AI pricing will force companies to rethink not only packaging, but sales incentives.

How do we forecast revenue?

Traditional SaaS forecasting is built around committed recurring revenue.

Hybrid AI pricing introduces variability.

Platform ARR may be predictable.

Consumption and outcomes may ramp over time.

Usage may spike seasonally.

Customers may adopt slowly and then scale quickly.

Some use cases may have great gross retention but uneven monthly consumption.

Finance, RevOps, and Sales will need new forecasting models.

The question will not just be:

“What is the renewal ARR?”

It will be:

What is the committed platform fee?

What is the expected consumption curve?

What is the outcome volume?

What is the attach rate of AI workflows?

What is the margin profile by use case?

What is the expansion trigger?

This is a different revenue operating system.

Who wins in this new SaaS world?

I see four groups of potential winners.

1. Incumbents with trusted data and workflow depth

Not every incumbent is doomed.

In fact, many incumbents have major advantages: installed base, trust, procurement relationships, compliance posture, domain workflows, historical data, and system-of-record status.

But those advantages only matter if incumbents move fast.

If they simply add AI features to old seat-based packaging, they will under-monetize the value and leave room for AI-native challengers.

If they turn proprietary data and workflow control into agentic outcomes, they can defend and expand.

2. AI-native startups that pick narrow, painful workflows

The most dangerous new players will not be generic AI wrappers.

They will be companies that pick a specific workflow, automate it deeply, price it around outcomes, and prove ROI quickly.

Bessemer’s State of AI 2025 argues that what constituted a great startup in the SaaS era no longer fully applies, with some AI companies growing at a pace software history has rarely seen. Bessemer also makes a blunt point: “There is no cloud without AI anymore.”

Speed matters.

An AI-native company can start with the outcome and build backward.

Incumbents often start with the product and try to retrofit AI into it.

That difference creates disruption.

3. Foundation model providers and AI platforms

The big labs will not stay neatly in the infrastructure layer.

They are already moving up the stack.

ChatGPT is becoming more connected to enterprise data, apps, agents, research, coding, and workflow execution. OpenAI’s apps in ChatGPT and workspace agents show how AI platforms can become the starting point for work, not just a back-end model provider.

This does not mean every SaaS category gets replaced by a lab.

But it does mean every SaaS company has to ask:

What happens when the AI interface owns the user relationship?

If the lab owns the interface, the SaaS vendor must own something harder to replace: proprietary data, workflow authority, trust, compliance, or execution.

4. Companies that master commercial architecture

This is the part I think is under-discussed.

Winning in AI SaaS will not only be about having the best model or the best demo.

It will be about building the right commercial architecture.

Pricing. Packaging. Contracts. Usage governance. Outcome definitions. Sales compensation. Deal desk policy. Customer success playbooks. Gross margin controls. Renewal analytics.

The companies that master this will capture value.

The companies that do not may create value for customers while giving too much of it away.

The deal desk will become more strategic.

I have spent enough time around CPQ, commercial operations, SaaS pricing strategy, and global deal desk motions to know that pricing strategy often sounds elegant until it hits a real enterprise negotiation.

That is where the truth shows up.

The customer wants flexibility.

Finance wants predictability.

Sales wants simplicity.

Product wants adoption.

Legal wants protection.

Customer success wants expansion.

The buyer wants proof.

Procurement wants leverage.

AI pricing will make this more complex, not less.

The deal desk will need to evolve from a discount-control function into a commercial design function.

It will need to help answer:

Should this customer buy platform access, prepaid consumption, or outcome bundles?

Should unused credits expire or roll over?

Should we allow ramped commitments?

Should the customer have hard caps or soft alerts?

What is the floor price for high-cost AI workflows?

What usage assumptions are embedded in this deal?

Are we discounting the platform, the consumption, or the outcome?

What happens if the AI performs better than expected?

What happens if it performs worse?

Do we have the contractual right to measure and bill the outcome?

These are not back-office questions.

These are growth questions.

In the AI era, pricing operations becomes strategy.

My prediction: SaaS will be redesigned around work, not users.

The first SaaS era moved software from on-premise to cloud.

The second SaaS era optimized adoption, product-led growth, usage analytics, and ecosystem integrations.

The AI era will redesign SaaS around work itself.

Not screens.

Not modules.

Not seats.

Work.

A customer does not wake up wanting more software.

A customer wants a business process to run better.

For years, SaaS vendors sold tools that helped humans do the work.

Now the best vendors will sell systems that do more of the work.

That requires a different product philosophy.

Design around workflows, not features.

Design around agents, not just users.

Design around data context, not just UI.

Design around outcomes, not activity.

Design around trust, not just automation.

This is the rebuild.

And yes, many application categories will be rebuilt from scratch.

The biggest disruption since cloud computing will not come because SaaS disappears.

It will come because the boundaries of software categories change.

CRM, support, analytics, finance, HR, security, procurement, and operations will not remain neatly separated when agents can move across systems.

The winning products will not be the ones with the most dashboards.

They will be the ones that sit closest to trusted data and completed work.

What should SaaS leaders do now?

For CROs and revenue leaders, I would start with five moves.

First, audit your pricing metric.

Ask whether your current metric grows with customer value or fights against it.

If your product saves labor but you price by labor count, you have a problem coming.

Second, identify your proprietary data advantage.

What data do you own or generate that a model cannot easily get elsewhere?

Is it fresh, accurate, permissioned, governed, and tied to workflow?

If not, fix that before obsessing over more AI features.

Third, map your workflows by AI exposure.

Which workflows can agents automate?

Which can they bypass?

Which require your system of record?

Which depend on regulated or proprietary context?

Which workflows are enhanced by AI?

Which are compressed by AI?

Which could be replaced by AI?

Which are defensible because of your data, trust, and execution layer?

That is the strategic work.

Fourth, experiment with hybrid pricing now.

Do not wait for the perfect answer.

Test platform plus usage.

Test prepaid credits.

Test per-action pricing.

Test outcome bundles.

Test premium governance tiers.

Test customer dashboards for usage transparency.

The market is still forming, which means customers are also learning.

The companies that test early will build pricing muscle before the model hardens.

Fifth, bring deal desk, finance, product, and legal into the conversation early.

AI pricing cannot be solved by product alone or sales alone.

It touches margin, risk, revenue recognition, contracting, compensation, forecasting, and customer trust.

The companies that coordinate early will move faster.

Final thought

The SaaS winners of the next decade will not be the companies that simply add AI to their existing products and keep the same commercial model.

The winners will rethink what they sell.

They will stop selling access and start selling work completed.

They will stop measuring value by logins and start measuring value by outcomes.

They will stop treating data as exhaust and start treating it as a strategic asset.

They will stop thinking of AI tokens as the product and start treating them as the cost structure behind the product.

And they will accept a hard truth:

Every SaaS company will become a data company.

The future of SaaS pricing is outcomes, not seats.

The companies that understand this early will shape the next market.

The companies that do not will spend the next few years defending old pricing models against customers, competitors, AI-native startups, and foundation model platforms that are already moving up the stack.

SaaS is not dead.

But SaaS as we priced it is being rewritten.