For fifteen years, the growth of a SaaS company was measured by a single metric: the number of paying users. More seats, more revenue. The model was clear for sales teams, predictable for finance departments, and easy to value for investment funds.

Artificial intelligence agents are breaking this connection. An agent executes a task across multiple software programs without any user ever logging into the interface. The customer gets the result, the software company loses the seat.

The question is no longer technological. It is about who, on a software company's executive committee, knows how to lead a business whose revenue engine is changing in nature.

Key Figures to Know

Why This Topic Is Forcing Itself Upon Us Now

On July 1, 2026, Gartner quantified for the first time what the market had been sensing. $234 billion of enterprise software spend is exposed by 2030 to what the firm calls agentic arbitrage: AI agents performing tasks across multiple systems, thereby reducing the need for users to go through traditional software interfaces.

George Brocklehurst, Vice President at Gartner, sums up the shift: agentic systems directly deliver outcomes, bypassing interface-centric applications and making software invisible.

Two false readings are circulating. The first heralds the end of SaaS. This is debunked by the figure itself: with 20% of spend exposed by 2030, the bulk of the market remains on its current foundations. The second considers this a product topic to be handed over to technical teams. This is more dangerous because it leaves intact an executive committee built for a model that is eroding.

The market, however, has already moved. According to a survey conducted in April and May 2026 by Kyle Poyar of over 230 B2B software companies, three out of four changed their pricing or packaging within the year. Hybrid pricing, which combines subscription and usage, went from 25% to 37% of software companies in twelve months.

What the Per-Seat License Allowed, and What Is Disappearing

The per-user model was not just a pricing grid. It structured the entire company.

It made growth mechanical. A hiring customer buys licenses. The software company grows with its customers without additional sales effort.

It made revenue predictable. An annual contract with a fixed number of seats provides visibility that finance departments and funds value at a multiple.

It simplified sales team compensation. A quota expressed in signed contract value is calculated without ambiguity.

It protected margins. The cost of serving an additional user is close to zero.

An agent attacks all four at once. It does not need a seat. Its consumption varies with customer activity, not headcount. It makes sales quotas difficult to define. And every task it performs consumes computing power billed by a model provider, bringing back a variable cost that SaaS had made disappear.

What the Change in Model Demands of the Executive Committee

Moving from per-seat pricing to usage-based or outcome-based pricing is not a pricing decision. It is a transformation that affects four executive roles, each in a different way.

None of these transitions happen through training. They require leaders who have already operated in a usage-based model, or who have led a revenue model shift of this scale.

The CEO: Accepting Cannibalization

The first obstacle is not technical; it is psychological. A software company deploying effective agents reduces the number of seats its customers need. It destroys a portion of its current revenue to create another, of a different nature, whose ultimate level is uncertain.

This is the classic innovator's dilemma, with a twist: the timeline is not chosen by the software company. If they do not offer agents, an AI-native player will, starting with the most standardized processes. Deloitte identifies customer support as the primary battleground for this competition.

The quality sought in a CEO in this context is rare: the ability to temporarily degrade a metric monitored by the board, while maintaining a trajectory that numbers will only confirm several quarters later.

The CRO: Selling an Outcome

This is the most exposed role, and the one where the profile mismatch is most frequent.

A sales leader trained in selling annual licenses excels at a specific exercise: negotiating volume, securing multi-year commitments, and ensuring renewals. Selling usage or outcomes is a different job. The contract is signed more easily because the customer does not commit to a volume, but the revenue builds after the signature, through actual adoption.

Three implications for the profile to recruit:

1. The line between sales and customer success disappears. Revenue depends on actual usage. The person carrying the target must master both.

2. Variable compensation for teams must be rebuilt. A quota based on signing a usage contract rewards a promise, not revenue. The sales leader must know how to design a consumption-based compensation system and get it accepted.

3. Outcome-based selling requires measuring the outcome. You must define with the customer what counts as a completed task, a processed file, or a resolved incident. This is a different negotiation skill.

The CFO: Managing the Unpredictable

Annual recurring revenue remains the reference metric for software companies and the funds that value them. Usage-based pricing makes it more volatile. A customer that slows down its business consumes less, and revenue drops without any cancellation.

Added to this is a new phenomenon for a generation of SaaS CFOs: variable cost. Every action an agent takes consumes computing power. Gross margin, which was the hallmark of the model, becomes a variable to be managed.

The ideal profile therefore combines two rarely united experiences: the culture of recurring revenue, and that of variable-cost models typically found in infrastructure services, marketplaces, or telecoms.

Product and Tech Leadership: Building for Agents

Value no longer resides in the screen. Gartner words it unambiguously: established software companies must shift from interface-based value to outcome-based value, and integrate agentic capabilities at the point of process execution.

Concretely, the product must become accessible to agents—its own and those of others—via robust, documented, and billable APIs. The orchestration layer, which decides which agent does what, becomes the place where value is captured.

This is a shift in culture as much as architecture. A product team that used to measure its success by time spent in the application must learn to measure it by the work completed without the user ever logging in.

The Case of Private Equity-Owned Software Companies

For a software company owned by an investment fund, the question is even more acute. The value creation plan was often built on existing account expansion—meaning growth in the number of seats sold to each customer.

If agents reduce this number, the main lever of the plan weakens even before exit. The fund then has two options: adjust the plan, or adjust the leadership team that must execute it. In most cases, both must be done, and the second task conditions the first.

The timing of this hiring matters. A change in revenue model takes several quarters to show its effects in the financials. A leader appointed eighteen months before a sale will not have time to demonstrate what an acquirer wants to see.

Common Mistakes

Hiring an AI profile with no P&L responsibility. Appointing a Head of AI reporting to product, with no authority over pricing or sales, produces agents that no one knows how to sell. The transformation happens at the P&L level.

Keeping the same sales leadership while changing the pricing grid. New pricing carried by a sales team compensated the old way will be sold like the old way. Teams will prioritize what pays their commission.

Confusing AI adoption with model change. Adding AI features to a per-seat license is a product enhancement. Charging for AI by usage or outcome is a model change. The two require different profiles.

Looking for the ideal profile solely within SaaS. Leaders who have already managed usage-based revenue can also be found in cloud infrastructure, payments, telecoms, or marketplaces. Restricting the search to traditional software companies means looking for experience in the new model where it exists least.

How to Evaluate a Leader for This Transition

1. Have they ever managed a business where revenue varied with usage? Ask for the specific case, the volatility observed, and what was put in place to manage it.

2. Have they made a decision that degraded a short-term metric? And how did they defend it to a board or shareholder?

3. How do they design variable compensation in a usage-based model? The answer immediately distinguishes real-world experience from theoretical knowledge.

4. Do they know how to read a gross margin with variable computing costs? For finance and general management roles, this is a dealbreaker.

5. What is their stance on cannibalization? A leader looking to protect per-seat revenue for as long as possible is not the right profile for this phase.

Frequently Asked Questions

Will AI agents kill SaaS? No. Gartner estimates that around 20% of enterprise SaaS spend, or $234 billion, will be exposed to agentic arbitrage by 2030. This is a profound transformation of part of the market, not its demise. Value is shifting from the interface to the outcome and to the orchestration layer.

What is agentic arbitrage? According to Gartner's definition, it occurs when AI agents perform tasks across multiple systems, reducing the need for users to interact with traditional software interfaces. The software continues to run, but no one logs in to use it anymore.

Why is the per-user license threatened? Because an agent doesn't take up a seat. It does the work that several users used to do, and its consumption depends on the customer's volume of activity, not their headcount. A software company selling seats sees its revenue decline as its customers adopt effective agents.

Which leadership roles are most affected? Sales leadership first, because selling usage or outcomes is a different job than selling licenses. Then finance leadership, which must manage variable revenue and computing costs. General management carries the central trade-off: accepting the cannibalization of current revenue before a competitor does.

Should we hire a Chief AI Officer? Not necessarily, and not first. A Head of AI with no authority over pricing and sales produces features that no one knows how to monetize. The first useful hire is often a sales or finance leader who has already operated in a usage-based model.

Where can we find leaders who understand usage-based pricing? In cloud infrastructure, payments, telecoms, and marketplaces, where revenue has varied with volume for a long time. These profiles are more common outside of traditional SaaS than within it.

Key Takeaways

$234 billion in software spend, about 20% of the enterprise SaaS market, is exposed to AI agents by 2030.

The market is already moving: three out of four software companies have changed their pricing within the year, and hybrid pricing has risen from 25% to 37%.

The per-seat license organized the entire company, from growth to margins. Its disruption affects four leadership roles simultaneously.

Sales leadership is the most exposed, because selling an outcome is not the same as selling volume.

For PE-backed software companies, value creation plans built on seat expansion must be revised, along with the team executing them.

The Laroze Partners Perspective

Most software companies treat the arrival of agents as a product roadmap item. This is understandable: it is where the technology is first visible.

But the product is the easy part. Technical teams know how to build agents. What is missing are leaders capable of selling them, billing for them, managing their margin, and explaining to a board why per-seat revenue is down while the company is growing stronger.

The software companies that navigate this shift will not be the ones with the best agents. They will be the ones that, early on, placed the right people in the roles that decide how value is captured.