Per-Seat Pricing Breaks When the Agent's Job Is to Reduce Headcount

01 /The Concept

Paid launched its platform in March and immediately raised €10M. Less than six months later, it closed a second round of $21.6M.

The company built a monetization platform for AI agents – giving developers the tools to set pricing, manage billing, and monitor whether their margins are actually holding up.

The platform has three main blocks.

The first is pricing and billing. The standout feature here is the ability to construct complex, hybrid pricing models that layer multiple revenue mechanisms simultaneously: per-seat fees, resource-based charges, and – most importantly – outcome-based pricing. A typical configuration might look like a flat subscription base with a usage credit allowance, plus a commission for each contract the agent closes on the customer's behalf.

Charging for outcomes requires integrating the AI agent with whatever system the customer uses to track results – typically a CRM. Paid provides an API for those integrations, with pre-built connectors to follow.

A useful second feature: before going live, developers can run proposed pricing structures against historical customer activity to model what each option would have generated under real usage patterns. No guessing which price point works best.

A third feature collects and updates benchmark data on what freelancers and contractors currently get paid for various types of work. Developers can use these benchmarks to calibrate what to charge for AI agent outcomes – ensuring they're not leaving money on the table. And they can automatically include those benchmarks in customer invoices, showing what the same work would have cost with a human doing it.

The second block is margin protection. AI agents call external models – OpenAI, Anthropic, others – and those calls cost real money. The cost per task scales with complexity. A developer who isn't watching their token spend can easily end up underwater.

Paid tracks this in real time: at any given moment, the platform knows how much a task in progress will generate in revenue versus how much has already been spent on model calls. Critically, it's not just tracking whether revenue exceeds cost – it's enforcing a target margin. If the margin dips below the developer's threshold, the platform issues an alert. The developer then decides whether to adjust pricing or optimize the agent's internal logic for the underperforming task types.

The third block is a customizable analytics dashboard. Standard charts and tables are included, but the distinctive addition is an AI assistant that lets developers query the financial performance of their agents in plain language.

Pricing: free tier up to $10,000/month in processed payments (with limitations); $1,000/month for up to $100,000 in monthly payments; custom pricing above that.

Paid was [covered previously](/review/ty-malo-zarabatyvaesh-potomu-chto-dengi-ne-za-to-berjosh) when it launched and raised its first round.

02 /Why It Matters

Paid was founded by the founder of Outreach – a company he scaled to $250M in ARR and a $4.4B valuation. He stepped down as CEO last fall specifically to launch Paid.

That move signals ambition. When someone leaves a company at that scale to start something new, they're betting the new opportunity is even bigger. In this case, they spotted a structural gap in the AI agent market – both conceptual and financial.

The core problem: traditional SaaS pricing is built around seats. Charge per user connected to the service. But the whole point of an AI agent is to replace human workers – which should, logically, reduce the number of employees at the customer. And fewer employees means fewer seats, which means less revenue for the agent developer.

That's a contradiction. An effective agent literally saws off the branch it's sitting on.

Usage-based pricing produces the opposite, equally unappealing result: the agent is incentivized to burn as many compute resources as possible. Meanwhile the customer adopted the agent specifically to save money.

Paid surveyed AI product developers and found that 75% expressed concern about per-seat business models. That concern is well-founded. Industry data shows that developer revenue under per-seat models has been declining steadily in 2025, going negative for the first time mid-year – with losses widening every month after.

As a result, many AI developers have begun moving toward outcome-based pricing. It's no longer a differentiator to attract customers – it's becoming a survival requirement. Companies that stick with seat-based or usage-based models face shrinking margins; those that transition to outcome models can finally build sustainably.

Paid understands the inertia here, which is why it offers a practical on-ramp: developers can migrate to the platform while keeping their existing pricing structure intact, using it initially just to monitor margins. When they see their margins deteriorating with current pricing – which they will – switching to outcome-based tiers is a single configuration change.

But outcome-based pricing doesn't just change how revenue is collected. It becomes a strategic decision about what outcome to charge for – because different outcomes produce radically different revenue trajectories.

Paid's blog gives an illustrative example: a customer success AI agent could charge either for contract renewals or for revenue growth from existing customers. The compute cost is identical in both scenarios – but the revenue profile is completely different.

The reason: some portion of any customer base churns, reducing the cohort that could be renewed. Net Dollar Retention (NDR) – a measure of whether revenue from a stable cohort grows or shrinks over time – is the more useful metric. Companies with NDR between 110–150% see revenue grow even as some customers leave, because existing customers spend more. An agent charging for revenue growth captures that upside; one charging for renewals does not.

Naturally, a gap this obvious doesn't stay private for long. Skope ([related review](/review/dengi-teper-nuzhno-brat-vot-za-jeto)), a recent Y Combinator graduate, is building in the same space.

03 /Opportunities

The first takeaway for AI product developers: start transitioning to outcome-based pricing now – and think carefully about *which* outcome to charge for, since the choice determines the ceiling on what you can earn.

The second takeaway: if all AI developers are eventually going to move to outcome models, there's a meaningful infrastructure opportunity in building the platforms that enable that transition.

There's also room for niche platforms that collect outcome data for specific product categories. Consider a mental health app that charges based on demonstrated improvement – pulling biometric signals like heart rate and blood pressure from a connected wearable. Or a weight loss app that reads body weight from a connected smart scale, and cuts access if the user stops weighing in.

The opportunity space is enormous – both for AI developers rethinking their business models and for anyone building the measurement and payment infrastructure they'll need to make it work.

And the most compelling long-term vision: imagine a world where every service and application is genuinely aligned with delivering results for its users, because that's the only way it gets paid