Is your SaaS business being left behind in the pricing race? | Pricing in the age of AI

Over half of SaaS companies have integrated AI features into their products as of 2024, according to Tenet. Yet, most of the companies have their pricing stuck in 2018. The per-seat model, the one that built the SaaS industry, is starting to crack under the weight of what AI actually does to how products get used.

This isn't a fringe opinion. It's showing up in sales cycles, in renewal conversations, and in the margin compression that catches finance teams off guard when they finally do the math on AI-delivered features.

The seat model's problem? It no longer fits

Per-seat pricing made sense when value tracked to individual users. One person, one license, clear value attribution. But AI doesn't map that cleanly onto a human. When one person with an agent does the work of three, the seat count goes down and if your revenue is tied to seat count, so does your ARR. Customers aren't doing this to be difficult. They're responding to what the product now lets them do.

There's also the margin side. Traditional SaaS targets 70–80% gross margins. AI product builders in 2026 are expecting closer to 52%, as stated by The SaaS CFO. That gap matters for how you bundle, what you charge, and whether revenue growth is actually business growth. Get the packaging wrong and you can be adding customers while quietly eroding unit economics.

The pricing models emerging to fill the gap—token consumption, per-resolution billing, hybrid seat plus usage, digital worker pricing that sits somewhere between a software license and an FTE salary—aren't entirely new. They're the evolution of models that have been on the market for years. What's new is that they're now necessary for a much larger share of SaaS businesses than ever before.

During our webinar Pricing SaaS in the age of AI, Ajit Ghuman, CEO of Monetizely and the author of the book Price to Scale, explored the ground reality and surfaced a split worth naming. Established SaaS companies are largely in a pressure zone right now: seats are shrinking, margins are compressing, and the move to consumption or agentic pricing is happening under duress rather than by design. Some established companies had to make uncomfortable structural pricing shifts mid-flight. New entrants can build agentic and outcome-based pricing from the ground up. They don't have a legacy seat model to protect, so they can price for what they actually are. That asymmetry is worth keeping in mind; the playbook for a native agentic company looks meaningfully different from the one for an established SaaS business retrofitting AI onto an existing model.

How are companies thinking of pricing revisions?

A quick pulse check that we did during our webinar, SaaS business leaders told an interesting story. When asked when they last seriously discussed revisiting their pricing, the largest group said last month (68%), but 26% said they hadn't revisited in over a year, and a huge gap left in between (no revisions in the last quarter or in the last year). A follow-up asked what triggered the conversation. The most common driver was a new product launch (28%), followed by an internal decision to capture market opportunity (20%), customer pushback on pricing (16%), and competitive pricing changes (8%).

That pattern matters. The healthiest trigger, tied to a product moment, is the most common one. But a meaningful chunk of decisions are still reactive. And by the time customers are pushing back or competitors have already moved, you're playing catch-up.
Ajit Ghuman is direct on the competitor-following instinct: it accounts for maybe 5–10% of the right decision, at most. A competitor's pricing reflects their cost structure, their customer mix, their strategic position, and the stage they're at. Using it as a primary input means you're solving their problem, not yours.

Factors, not frameworks

Before you can choose a pricing model, there are three lenses worth working through:

1. Customers
2. Market
3. Value

On the customer side: Which features are getting the most traction outside your core product? That tells you what's worth gating or charging more for. And when customers churn, is it the price that's wrong, or the model? Those are different diagnoses with very different remedies. One you fix by adjusting a number, the other by rethinking the structure entirely.

On the market side: Are buyers walking into deals expecting something different from what you're currently offering? If pricing is consistently surfacing in stalled or lost deals, that's a structural misalignment, not a sales execution problem.
The value question is the most fundamental: What does it actually cost you to deliver one unit of output, and what outcome is the buyer paying for? That cost floor is your floor, so price below it and you may see early traction, but scalability tends to unwind over time.

Three traps that show up repeatedly

Companies get this wrong in fairly predictable ways. Changing pricing before customers have experienced the new value first triggers resentment, you're asking them to pay more for something they haven't yet proven to themselves. Copying a competitor's model without understanding their cost structure can quietly destroy margins; what works at their scale or unit economics may be actively harmful at yours. Adding an AI surcharge on top of an existing package, without asking whether customers actually value that feature enough to pay extra, often inflates perceived price without adding perceived value, accelerating voluntary churn rather than reducing it.

That last trap is increasingly common. A lot of companies added AI as a paid SKU and are now discovering that the market has shifted: Every major platform has a co-pilot, inference costs have fallen, and customers aren't willing to pay a premium for something that now reads as table stakes. When asked how they'd package a new co-pilot, none of the webinar attendees voted for keeping AI as a paid add-on, yet that's still the default move many teams reach for.

Pricing agents is a different problem

When the product is an agent rather than a tool, the pricing question changes meaningfully. Ghuman calls this the "Agentic Monetization Spectrum," and it runs along three variables.

The first is zero human ability. How much can the agent perform work fully without human help or oversight? The greater that potential, the more the pricing metric detaches from users entirely. You're no longer selling seats because there's no human to attach a seat to.

The second is operational domain. Is the agent a copywriter, a marketer, or effectively a COO? The broader the role, the more budget it's displacing, and the more you can price against that displacement rather than against a software line item.

The third is the output-to-cost ratio. Is what the agent delivers relative to what it costs to run? Healthcare and legal applications tend to produce large, high-value outputs. Customer service tends toward lower-to-mid outputs at higher volume. Where you fall on that ratio shapes whether outcome-based pricing is even viable, or whether you're still anchored to consumption and credits.

Watch the webinar here to learn the pricing and value nuances that many popular companies of today follow

What the "live pricing scenario analysis" revealed

During our webinar, three live scenarios put the theory to work, and the audience responses were instructive.

Scenario 1

In the first scenario, most people gravitated toward the hybrid model. Keep some seat structure, layer consumption on top. It's a reasonable instinct, and for most mid-market SaaS companies it's probably correct. But the more interesting insight is that the hybrid answer only holds if you have time. Google Workspace can raise seat prices and hold because its distribution makes switching too painful. Intercom couldn't and it was staring at category-level disruption, forcing them to accept a significant revenue reset to rebuild around agents. The hybrid vote was right for the middle, but the middle is a narrower band than most companies assume.

Scenario 2

The second scenario produced a telling gap between what people voted for and where the market is actually heading. A majority went with selling the AI co-pilot as a paid add-on, which was the instinct two years ago. The ground has shifted. When every major platform ships a co-pilot and inference costs keep dropping, charging a premium for it only works if yours does something no one else's does. Otherwise you're adding friction to adoption and inviting customers to compare line items. Treating AI as table stakes and using it to deepen retention is increasingly the harder but more durable call.

Scenario 3

The third scenario was deliberately underspecified, and that was the point. Audience votes split fairly evenly across consumption, seat, and outcome-based options, which is exactly what you'd expect when the variables that determine the answer (autonomy level, cost structure) haven't been defined. The right metric for an agentic HR product depends entirely on how autonomous it actually is and what the unit economics look like. If inference costs are high, consumption pricing protects the margin. If costs are negligible, relative to the value of a successful hire, outcome-based pricing is both viable and more aligned with what the buyer actually cares about. The scenario was a reminder that there is no default answer, only the answer that fits your specific product's economics.

In conclusion, pricing SaaS when AI is changing the way everything operates is indeed complicated. But there is a process to it, and the right infrastructure makes following that process a lot less painful. If the billing infrastructure to execute on a hybrid, consumption, or outcome-based model is the constraint, Zoho Billing is built to handle that complexity without requiring custom engineering work.