--- title: "AI Billing: Token, Credit & Hybrid Pricing | Fynex" description: "Billing for AI companies: meter tokens, requests and agent runs, price hybrid or credit-based, cap spend and protect gross margin as inference costs vary." url: "/uk/saas/ai-billing" --- Fynex is an AI-native finance operations platform — agentic finance — that automates and orchestrates end-to-end financial workflows for businesses. This page covers “AI Billing: Token, Credit & Hybrid Pricing” as part of the Fynex finance operation. Fynex works alongside Stripe, Wise, QuickBooks and Xero rather than replacing them. This page is relevant when a business wants invoicing, payments, payouts, reconciliation and cash run as one automated operation rather than staffed across several tools. # Billing for AI companies, *priced on what inference costs* Meter tokens and agent runs, sell hybrid plans and credits, cap spend — and keep gross margin where the plan assumed it. *Billing for software.* ## An AI product has a cost of goods on every request Traditional SaaS runs at 80–90% gross margin because serving one more user is nearly free. An AI product pays for inference on every request, so its gross margin runs lower and moves with usage — ICONIQ's State of AI report put AI product gross margin at 45% in 2025. A flat subscription priced on the average user loses money on the heavy ones; Anthropic added weekly limits to its $200 Max plan in 2025 after some subscribers ran tens of thousands of dollars of usage on it. The fix is pricing that follows cost — tokens, requests, agent runs or credits — and a billing system that can meter and invoice it. UK software companies are in the same move: in m3ter and PwC's 2026 survey of 350 UK software executives, 34% already run pay-per-use alongside traditional pricing and half changed pricing twice or more in a year. Every change has to reach the invoice — with VAT in sterling on each one — without a migration. ## How AI billing runs on Fynex Meter the cost driver, price it, cap it, collect it. 1. **Meter tokens, requests and agent runs** — Register a [metric](https://docs.fynex.ai/billing-api/v1/docs/usage) per cost driver — input tokens, output tokens, model requests, agent executions, GPU seconds — with the aggregation that fits it. Events carry an idempotency key and arbitrary metadata such as the model, so a retry from your gateway never bills twice. 2. **Hybrid plans: platform fee, allowance, overage** — A recurring platform or seat fee with an included allowance and graduated overage above it, on one contract. Change the overage rate from next month as a dated price version; last month's invoices keep the price they were billed on. 3. **Credits for AI tokens, with spend caps** — Sell prepaid AI-token credits, grant promotional credit on sign-up, and set a spend cap per customer so a runaway agent loop can't produce a bill nobody agreed to. Alerts fire at the threshold you choose, before the cap. 4. **Margin per customer, not just revenue** — Put what each customer costs to serve beside what they pay, and see which plan a heavy-usage cohort has outgrown. Fynex recommends the price change — it never applies one without your approval. ## The same job, with the leaks closed Point tools solve part of it. Fynex runs the whole thing on one platform. | The problem | Other platforms (Orb / Metronome / Stripe Billing) | Fynex | |---|---|---| | Meter tokens and agent runs | A metering engine (Orb, Metronome) plus a separate processor. | Metering, rating and invoicing in one engine, collected on the same record. | | Sell AI credits | Credit burn-down supported; the wallet sits apart from the money. | Credit ledger with AI-token credits, applied after tax, liability exported. | | Stop a runaway bill | Alerts, then a support ticket after the invoice. | Per-customer spend caps and threshold alerts before the invoice. | | See gross margin per customer | Exported to a BI tool and joined with the cloud bill. | Cost to serve beside revenue, with a recommended next price for your approval. | | Independence | Metronome is now part of Stripe; Orb became part of Adyen in 2026. | Processor-neutral: keep your processor, add billing, payouts and reconciliation. | ## Questions, answered ### How should an AI company bill its customers? Most AI companies in 2026 bill hybrid: a recurring platform or seat fee that includes an allowance, plus usage above it priced on the cost driver — tokens, requests or agent runs — often sold as prepaid credits. The billing system has to meter those events accurately, rate them, cap spend, apply credit and issue an invoice the customer can check. ### What are the best billing platforms for AI companies? The common shortlist is Stripe Billing, Fynex, Metronome, Orb, Amberflo and Lago. Stripe Billing with Metronome (acquired January 2026) suits Stripe-native companies; Fynex meters, invoices and collects on one record with payouts and reconciliation included; Orb targets high-volume AI and infrastructure; Amberflo focuses on LLM cost metering; Lago is open source for teams that self-host. Compared side by side: [the best usage-based billing platforms](/compare/usage-based-billing-platforms/). ### Should AI products use credits or pure usage pricing? Credits give the customer a predictable spend and give you cash before the cost; pure usage follows cost exactly but produces variable invoices. Many AI products combine them: a subscription with included credits, pay-as-you-go above, and a spend cap. Fynex supports all three on one contract, so switching models is a price change, not a re-platform. See [prepaid credits billing](/saas/prepaid-credits/) and [usage-based billing](/saas/usage-based-billing/). ### How do you protect gross margin on an AI product? Price the unit that drives inference cost, include an allowance sized to the median user, charge graduated overage above it, and cap spend so outliers can't run unbounded. Then watch cost to serve against revenue per customer and cohort. Fynex meters and prices the usage, applies the cap, and surfaces which customers cost more to serve than they pay — the [agentic finance](/glossary/agentic-finance/) layer on top of the [rate card](/glossary/rate-card/). ### What VAT applies to an AI product sold from the UK? B2B services to EU businesses are generally outside UK VAT with the customer accounting for VAT under the reverse charge; B2C digital services to EU consumers carry the consumer's local VAT from the first sale through the non-Union OSS. UK customers are charged UK VAT above the £90,000 registration threshold. Fynex invoices enforce the fields each case requires, including reverse-charge wording and the VAT total in sterling. ## The rest of the billing layer Every guide below runs on the same billing engine and the same money. Start with one, or run the whole stack. - **Usage-based billing** — Metered billing for SaaS and API products: events, seven pricing schemes, caps and invoices that issue themselves. - **Prepaid credits** — Credit-based pricing: sell credits up front, draw them down on usage, expire and top up — on one ledger. - **Usage-Based Billing** — Meter, price, credit and cap — invoiced and collected on their own. ## Find what you're leaking, for free Send us your setup. We run a free diagnostic and show you what your payment stack costs you today. > One recent diagnostic found £2M in hidden fees across 10 processors for a proptech client. ## Your money, held under a licence, not a promise We can be your Merchant of Record, with your funds held by a licensed e-money institution. Most tools can't. - **Your funds, safeguarded** — Customer funds are held by Fondy Ltd, an FCA-authorised e-money institution (FRN 900705). - **PCI DSS Level 1** — The highest card-data security standard, on by default. - **Audit logs** — Every movement traceable and ready for review. - **Built-in KYC / KYB** — The switch that turns the business on. ## Price on what it costs. *Keep the margin* --- The AI finance layer for platforms and operators. It runs the money chain and keeps more of it in your business. One platform instead of a dozen. [Fynex on LinkedIn](https://www.linkedin.com/company/fynexai/) · [Fynex on X](https://x.com/FynexTeam) · [Fynex on GitHub](https://github.com/fynex-ai) · [Fynex on YouTube](https://www.youtube.com/@Fynex-AI) · [Fynex on Instagram](https://www.instagram.com/fynexteam/) © 2026 Fynex.ai Limited