Your GTM Might Be Ready for Agentic SaaS. Is Your Pricing?

Part 3 of 7 · Agentic SaaS Blueprint series

Let’s start with a scenario I keep seeing pitched, one that looks perfect on a spreadsheet: a risk-auditing platform charges customers a percentage of the risk value it identifies and prevents. The bigger the enterprise, the bigger the exposure, the bigger the value prevented, and, naturally, the bigger the fee.

Value-aligned pricing. Textbook value-based design. It’s also the kind of pricing model that can quietly open the door for a competitor to walk into your largest accounts, not by being better, but by being simpler.

We’ll come back to that. It’s one of the sharpest examples I’ve seen of a problem running through almost every pricing conversation I have with ISVs: the math can be spot-on while the underlying psychology is completely off base.

Before we get to why, there’s a more fundamental question most teams skip entirely: What are you actually charging for?

The Tier Progression: From Per-Input to Per-Outcome

In the architecture piece, I wrote about seven layers that have to be designed together. Pricing isn’t an eighth layer added afterwards; it’s a lens across all seven, and it tends to follow a progression as products mature. I think about it as five tiers, each answering the same question differently: what does the customer actually pay for?

1) Per Agent. The customer pays per deployed agent or seat served. Familiar, easy to budget, easy to sell. For the vendor, it's also the most predictable revenue line in the model; headcount is stable, contracts renew annually, and forecasting doesn't swing with usage the way activity- or outcome-based revenue does. But the value of an agent doing more autonomous work doesn’t show up in the agent count, so pricing and value begin pointing in different directions, the same disconnect I described in the GTM piece.

This is where the model most often backfires. Microsoft and Google both tried flat, per-seat pricing for their AI assistants - Copilot and Gemini at $20–30 per employee per month, and ran into real customer resistance: value varied sharply by role. For example, a developer might use it daily, an administrator barely at all, yet every seat carried the same flat fee, and a per-seat add-on for a workforce of thousands is hard to defend internally without a clear line from cost to realized value. Per-agent pricing needs either a defined scope ceiling or a visible usage signal attached to it, or it reads as a tax on headcount rather than a price on work.

For example, Claude's own consumer subscriptions work this way in practice: a flat monthly fee paired with a rolling usage window that Anthropic tightens or loosens as demand and available compute shift. It was tightened in August 2025 under heavy load, doubled again in May 2026 once new capacity came online. The ceiling isn't a flaw in the pricing; it's the mechanism that keeps a flat fee sustainable against a cost base that moves.

2) Per Input. The customer pays for what they put in: API calls, tokens, documents processed. Easiest to implement, weakest signal; it charges for effort rather than results. For the vendor, it's also the weakest pricing power in the entire progression: you're monetizing at the cost layer, easy for a competitor to beat on price alone since there's nothing to differentiate on but the rate. The token-metered version of this carries an odd penalty on top: a smarter agent that solves the same task in fewer tokens earns the vendor less, even though call volume (how often the customer actually needs the agent) doesn't shrink just because the model got more efficient. OpenAI’s API is priced this way, per million tokens processed, regardless of whether the output was useful - the classic foundation-model metric most agentic vendors build on top of, then move away from as they climb the tiers below. An agent that takes ten attempts costs the customer ten times more than one that gets it right the first time. That completely misses the point.

3) Per Activity. The customer pays per task, transaction, or workflow completed, such as a ticket resolved or a claim adjudicated. This ties price to work customers recognize as getting done, not infrastructure consumed. For the vendor, revenue now scales with actual usage instead of being capped at a flat fee, which keeps growth incentives aligned with the customers, but it requires real metering infrastructure to track reliably, and it can quietly reward maximizing the number of billable activities over the quality of each one. Document-processing vendors like Hyperscience and Rossum charge roughly $1.50 per page processed, a clean per-activity metric where the unit is the task itself, not the outcome the task eventually produces.

4) Per Output. The customer pays for the deliverable produced: a report, a design, a drafted contract, regardless of how many attempts it took. For the vendor, this shifts technical efficiency risk onto you as you absorb the cost of every failed attempt, not just the one that ships, so your margin depends on how efficiently your agent gets to a usable result. For the customer, it’s a stronger value signal: they are paying for what they keep, not what they tried. AdCreative.ai prices this way for ad creatives: generation itself is unlimited, and a credit is only spent when the customer downloads a finished creative.

5) Per Outcome. The customer pays for the business result itself: revenue protected, costs avoided, risk prevented. Some of the most attractive economics in agentic SaaS live here, because price scales with value delivered rather than effort or volume. Chargeflow, an AI chargeback-recovery platform, charges 25% of each disputed payment it wins back and nothing if it loses, a clean, real example of the model: no recovery, no fee. For the vendor, it's also the tier that demands the most maturity: you can't charge for an outcome you can't reliably measure and attribute, and unlike every tier before it, the burden of proving causality, not just correlation, sits entirely on your side of the table

Choosing Your Metric: The COMPASS Framework

The five-tier progression tells you what the options are. It doesn’t tell you which fits your product. For that, I’d point to COMPASS (Choice of Optimal Metrics for Pricing Agentic Systems & Solutions), developed by Michael Mansard and published in a joint AWS, Zuora, and Simon-Kucher whitepaper on agentic AI pricing.

COMPASS asks two questions. First, what’s the scope of work: narrow tasks like a factory-line worker, a multi-step process like a service, or an open-ended goal like a partner? Second, how attributable is the impact? Can you trace a result directly to what the agent did, or is it mixed with other factors? Plot your product on those two dimensions, and you get a strong signal for where you can credibly operate on the tier progression.

A narrow-scope agent with diffuse attribution belongs closer to per-input or per-activity; claiming per outcome pricing there is hard to defend to a sophisticated buyer and hard to measure yourself. Broadscope agents pursuing measurable goals sit at the other end - where outcome pricing becomes credible, economically attractive, and, by extension, where premium pricing becomes possible. Broad scope combined with high attribution creates a metric competitors struggle to replicate convincingly.

This is where the example from the architecture piece becomes relevant again. Intercom’s Fin support agent moved to outcome-based pricing by charging per resolution and reported net revenue retention climbing from 112% to 146%. That’s not a coincidence; resolution is a high-attribution, well-scoped outcome, exactly the COMPASS sweet spot.

The framework’s own framing makes this point directly; pricing metrics are often the first thing teams focus on, but they’re only one piece of the strategy. In agentic AI, pricing and packaging aren’t downstream of go-to-market; they’re part of the foundation the rest of the motion gets built on.

Value Measurement and Aligned Reporting

Choosing a high-attribution metric only holds up if the measurement behind it holds up too, and this is where most of the gap actually lives. A Gartner survey found 31% of chief sales officers cite proving the ROI of AI-driven tools as a top challenge to their 2026 objectives, and IBM’s Institute for Business Value found only 25% of AI initiatives have delivered their expected ROI. The gap isn’t uniform: Microsoft-sponsored IDC research puts the average return at $3.70 per dollar invested, with top-performing, well-measured organizations reaching $10.30; measurement rigor is a large part of what separates the two.

Three practices close that gap.

1. Capture a pre-agent baseline before go-live; you simply can’t prove lift retroactively.

2. Build customer read-back access into the contract rather than retrofitting it at renewal, when the ask looks defensive.

3. Publish the methodology itself, not just the number, so the customer can audit it rather than take it on faith.

Reporting cadence matters as much as methodology. A recurring business review that shows the customer the same number you’re billing against turns the invoice into a confirmation of something they already saw coming, not a sudden surprise.

Designing the Commercial Model: Packaging, Segments, and the Margin Question

Choosing your metric is the foundation. Building the commercial model around it raises a second set of decisions, starting with understanding what any single metric necessarily gives up.

The Impossible Triangle. Every pricing model balances three forces pulling against each other: cost recovery (does price track inference and compute cost as usage scales?), value capture (does price reflect the outcome delivered?), and predictability (can the customer forecast next month’s bill?). No model optimizes all three. Per-seat protects predictability and sacrifices the other two. Pure usage-based pricing protects cost recovery and value capture but breaks forecasting, which is usually where finance-side resistance kicks in. Knowing which two forces your model protects, and naming the one it sacrifices, is a sharper starting question than asking “is this the right metric” in the abstract.

1) Packaging tiers vs. metric tiers. Don’t confuse the tier progression above - what you charge for - with packaging tiers like Starter, Growth, and Enterprise, how much access a customer gets. Most products need a metric and a packaging structure, not one instead of the other. The mistake I see most often is treating the two as a single decision.

2) Segments change the calculus. A single pricing metric rarely works for every customer segment, because different buyers reject the same model for different reasons. A mid-market buyer often picks peragent pricing over per-outcome, even knowing per-outcome would cost less; however, a flat, predictable bill matters more to them than the savings. An enterprise buyer resists per-outcome pricing too, but for a different reason: the fee grows precisely as they succeed, which can start to feel like a penalty for growing. Two different objections, two different fixes: swap the metric for mid-market, redesign the fee structure for enterprise.

3) Overage design is where a surprising amount of goodwill gets lost. With usage-based agentic pricing, overages often correlate with the agent doing more valuable work, not with customer carelessness; punitive overages on a value-scaling metric send the wrong signal. Better: generous committed-use bands, overage priced close to marginal rate, and proactive “approaching your limit” communication rather than a surprise invoice.

Amazon Quick is a public example of this structure: the per-seat price covers access, while a separate bundled allowance of agent and research hours covers usage. Professional tier includes 4 hours a month, Enterprise 8 hours with overage metered to the second at $3 per agent hour and $6 per research hour. The seat price and the compute cost are deliberately two different numbers, priced and communicated separately, rather than one flat fee quietly capping what 'unlimited' actually means.

4) Unit economics and initial gross margin bring the architecture conversation back into view. Every agent action has an inference cost that doesn’t disappear because the pricing model looks elegant on paper. Inference now accounts for roughly 70% of a model’s lifetime cost, and agentic workloads compound that further; multi-step reasoning loops can consume five to thirty times more tokens per task than a standard chatbot interaction (Gartner, 2026). The real question isn’t “what’s our margin at pilot scale,” it’s “what’s our margin at the usage level we’re actually selling toward.” Model your economics at target scale, not current scale.

It’s worth separating the cost conversation from the quality conversation. Klarna’s widely covered pullback from an AI-first customer service model is instructive here, not as a verdict on AI capability, but as a pricing lesson. By the company’s own account, cost became too dominant a factor in how the system was evaluated, and quality slipped. A pricing model built entirely around cost savings, without a matching commitment to the quality bar those savings are supposed to fund, is a fragile foundation for both the margin story and the customer relationship.

What Lets a Company Charge a Premium

Every framework above assumes you’ve earned the right to price at the top of your category. That right isn’t generic; it’s built from specific, benchmarkable characteristics, and it’s worth checking your own product against them honestly, not just against competitors.

Attribution rigor is the biggest lever. A vendor who can show, with instrumented telemetry, exactly what the agent did and what it caused can charge more than one making the same claim on a slide. Autonomy depth matters because customers pay more for judgment than execution. Switching cost compounds pricing power over time: an agent embedded deep in a customer’s workflow is harder to rip out than one bolted on at the edge. Domain specificity: a vertical-trained agent that understands claims adjudication or clinical coding commands more than a horizontal tool doing the same task generically, because the buyer is paying for judgment, not the model underneath it. Governance maturity, meaning audit trails, humanin-the-loop controls, explainability, is what lets regulated buyers say yes at all. And a published, jointly auditable measurement methodology turns a contested number into an agreed one. That’s the difference between a premium a customer resents and one they defend internally.

None of this is static. Benchmark yourself against these characteristics on a recurring cadence, not just at launch, and benchmark against named competitors on the same axes, not just their list price, but which of these characteristics their pricing model is actually resting on.

Pricing to Win the Market: Entry, Competition, and Willingness to Pay

Choosing the right metric and commercial model still leaves one question: what number do you actually put in front of a customer? Three considerations shape it.

1) Entry pricing is a different decision from steady-state pricing. Charging below what the value justifies early on can be the right tradeoff; it builds the case studies and usage data that eventually make premium pricing credible. The risk is anchoring. Salesforce’s Agentforce is a useful public illustration of what happens without a plan for that transition: it launched in late 2024 at a flat $2 per conversation, simple on paper, but the flat rate priced out smaller organizations while giving larger customers no way to reconcile cost against value at their scale. Within about a year, Salesforce had introduced two more pricing models running alongside the original; three live models for one product is what happens when a metric launches without a tested path off it. You need to design the exit before you need it: a usage threshold, a renewal event, a tier transition, and pressure-test it against your smallest and largest likely customers before it ships.

2) Competitor pricing. Agentic products increasingly compete with the cost of an internal team building something similar with the foundation models now available; that comparison is about whether your price reflects something customers can’t easily replicate themselves.

3) Willingness to pay. The category is new enough that many customers are comparing your price to the cost of their problem, not a competitor’s number. The most useful signal comes from the pilot itself: structure it so the customer has to quantify the problem to secure budget.

Migrating an Existing Base

Most ISVs aren't pricing a new product for a new customer; they're moving a seat-based base onto agentic pricing. The direction of the market isn't ambiguous: IDC expects 70% of software vendors to move away from pure per-seat pricing by 2028, and Gartner projects more than 40% of enterprise SaaS spend shifting to usage-, agent-, or outcome-based pricing by 2030. Companies that hold onto pure per-seat pricing after the product has genuinely shifted to agentic work see meaningfully higher churn than peers who've already moved; delaying the repricing event to protect a comfortable renewal usually backfires.

Sequence communication to lead with why the value equation changed, not just the new number. Set a dated sunset on grandfathered pricing rather than an indefinite exception, and expect to run legacy and new billing in parallel longer than feels comfortable.

When Outcome-Based Pricing Backfires: The Psychology Problem

Back to the risk-auditing example. By every framework above, charging a percentage of risk value prevented looks excellent: Per Outcome, highly attributable, exactly the COMPASS sweet spot. It can still fail with your largest accounts not because the math is wrong, but because the psychology is.

It’s worth knowing how common the underlying attribution problem is before getting to the psychology. Research from MIT’s NANDA initiative found 95% of enterprise generative AI pilots deliver no measurable P&L impact, and McKinsey’s 2025 State of AI survey found no more than 10% of organizations are scaling agents in any single business function. The common thread is attribution, not capability. The risk-auditing example is a case where attribution actually works and yet, the psychology still breaks it.

As the enterprise gets larger, exposure grows, value prevented grows, and so does the fee. From the vendor’s side, that’s value-aligned pricing working as designed. From the buyer’s side, it can feel like a tax on success, meaning the better we get at managing risk and the bigger we become, the more we pay for the privilege.

Large enterprises expect better economics as they grow; a model that reverses that expectation is hard to justify internally, and it’s exactly the opening a competitor needs. A rival offering flat-rate pricing suddenly looks attractive because it restores what buyers intuitively expect: predictability, and the sense that bigger customers get better economics. You can lose your best accounts not because your product is worse, but because your pricing fights how enterprise buyers expect growth to feel.

None of this means abandoning outcome-based pricing. The economics are often too attractive to walk away from. It means designing around psychology, not just mathematics.

1) Cap the fee. An absolute ceiling per period stops the “success tax” from growing indefinitely while still capturing most of the upside.

2) Tier the percentage down. As the dollar amount you’re protecting gets larger, charge a smaller percentage of it, building a volume discount directly into the structure, and answering ‘where’s my enterprise discount’ before the customer has to ask.

3) Blend the model. A smaller base subscription plus a smaller outcome-based component spreads risk on both sides.

4) Make the calculation theirs, not just yours. A number only you calculate is a number customers can challenge. Build it jointly and make it auditable once customers agree with the assumptions; the fee stops feeling imposed.

The math matters, but in pricing, psychology often determines whether the math gets a chance to work.

The Full Pricing System

Everything above concentrates on the highest-leverage decisions: what you charge for, how you build the commercial model around it, what earns premium pricing, and the psychology of getting outcome pricing wrong. It’s deliberately not a complete map.

I think about pricing as a system of interlocking elements - metering infrastructure, discounting discipline, pricing governance, regional strategy, bundling, freemium design, channel pricing, experimentation cadence, and more. This isn’t a fixed or universal checklist; add, merge, or drop elements as your own product dictates. Some elements are foundational prerequisites that everything else inherits if broken, some are structural once the foundation holds, and some are optimizations that compound once both are in place. Treat the diagram below as a starting checklist, not a spec.

The Questions Worth Asking

I’m not going to tell you which pricing model to adopt; that depends on your product, your customers, and how mature your agentic capabilities are. But these are the questions I’d ask.

1) On Value Metric. If a sophisticated customer asked you to prove the link between what your agent did and what you’re charging, could you show them today or would you be explaining a methodology you haven’t built yet?

2) On Tier Progression and Unit Economics. Could you move one level up the value chain with the observability you have today? Do you know your gross margin at the usage level you’re actually selling toward, or only at pilot scale?

3) On Premium Positioning. Which benchmarkable characteristic: attribution rigor, autonomy depth, switching cost, domain specificity, governance maturity, auditable methodology does your pricing actually rest on, and which are you assuming?

4) On Transition. If entry pricing sits below what your value justifies, or you’re migrating an existing base, what’s the mechanism and the dated sunset or has “temporary” quietly become indefinite?

5) On Psychology. If your best, fastest-growing customers pay the most in absolute terms, have you designed for the conversation where they ask why success costs more, or will that happen for the first time when a competitor’s simpler pricing is on the table?

Where This Connects Back

If you’ve followed this series, the throughline should be familiar: architecture, go-to-market, and pricing aren’t separate workstreams that happen to influence each other. They’re three views of the same underlying decisions. The tier you can credibly price at is constrained by what your observability layer can prove, and the GTM motion that can sell per-outcome pricing is the same motion that needs that observability to make the claim credible in a sales conversation. As the COMPASS framing puts it, pricing and packaging are the foundation the rest of go-to-market gets built on.

One pattern runs underneath nearly every fix in this piece, even though I never named it until now: layeredcake pricing. The overage structures earlier in this piece - a seat fee for access, a separate usage allowance metered on top, and storage and infrastructure priced independently again - are one version of it. The blend of a base subscription and a smaller outcome-based fee that defuses the psychology problem in outcome pricing is another. The Impossible Triangle itself is the reason why: no single metric resolves cost recovery, value capture, and predictability at once, but two or three metrics, stacked, usually can. None of these is one clever metric solving pricing by itself; they're simple layers, each absorbing a different piece of cost, value, or risk, stacked instead of forced into a single number that can't hold all three. If there's one practical takeaway from this whole piece, it's that: stop looking for the metric that does everything, and start deciding which layer handles which job.

For deeper reading, AWS’s and Zuora’s joint resource on why agentic AI requires SaaS pricing to transform and Zuora’s walkthrough of the COMPASS framework are both worth the time.

There’s one more piece to this series - the one that ties architecture, GTM, and pricing together from a different angle. When you build an agent for one role inside a customer’s organization, how many other roles end up with a stake in how it’s priced, governed, and deployed? RevOps cares about the data the pricing model depends on. Finance cares about the unit economics. Legal cares about how the outcome gets defined in the contract.

If you’re working through your pricing model - metric selection, packaging design, or benchmarking your premium against the market - that’s a conversation my team at AWS SaaS & Agentic Apps works through with ISVs regularly. If you’re an AWS customer, your account team can connect you with us. If not, I’m happy to connect here on LinkedIn.