Skip to content
Back to skills

Ai Billing Models

ASecurity

AI vendor billing-model fluency — auto-activates when normalizing, comparing, or forecasting AI spend across seats, usage, credits, and reserved capacity

  • 29 stars
  • 0 votes
  • 0 copies
  • 7 views
  • Added September 6, 2026
ai-agentsgoapi

Works with

  • api

Security analysis

A100/100

Scanned September 6, 2026

npx -y skills add alexclowe/awesome-claude-cowork-plugins --skill ai-billing-models --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ai Billing Models?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Ai Billing Models
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/alexclowe-ai-billing-models/badge)](https://www.skillsdirectory.com/skills/alexclowe-ai-billing-models)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: ai-billing-models
description: AI vendor billing-model fluency — auto-activates when normalizing, comparing, or forecasting AI spend across seats, usage, credits, and reserved capacity
---

You have deep fluency in how AI products bill. When the user is working on AI spend tasks, apply this knowledge automatically.

## The billing shapes

- **Per-seat subscription** — flat monthly/annual per user (chat assistants, copilots, embedded AI tiers). The waste mode is inactive seats; the control is seat-to-active-user reconciliation on a cadence
- **Usage-based API** — metered per token, request, image, or minute, usually with different input/output rates and per-model pricing. The waste modes are retries, oversized prompts/contexts, and defaulting to a bigger model than the task needs
- **Prepaid credits** — money in, credits out, often with expiry and sometimes with bonus tiers. The waste modes are expiry (breakage) and the false comfort of "already paid" masking a rising run rate
- **Reserved / provisioned capacity** — committed throughput (e.g., provisioned units) billed whether used or not. The waste mode is committing to peak instead of base load; the control is utilization tracking against the commitment
- **Hybrid** — seat + usage overage, platform fee + credits, or an enterprise minimum with drawdown. Read the invoice's fixed and variable parts separately or the normalization will be wrong
- **Embedded AI** — AI features priced inside a non-AI product (CRM tiers, office suites, support desks). Frequently invisible in an "AI spend" search because the line item doesn't say AI

## Normalization rules

- Amortize annual prepay to monthly (÷12) and SAY it's amortized — cash timing and run rate are different questions
- Convert credit purchases to run rate from actual burn (credits used per month × price per credit), not from the purchase amount
- Count reserved capacity at commitment, with a utilization percentage alongside — 40% utilization of a committed pool is the finding, not a lower cost
- Keep one currency and one period; label estimates as estimates every time

## Contract mechanics worth flagging

- Auto-renewal windows and notice periods (the negotiation clock starts at notice-minus-30, not at renewal)
- True-up clauses on seat growth; whether unused seats true DOWN
- Credit rollover and expiry terms
- Rate-lock or price-protection language, and what happens to grandfathered pricing at renewal

Always distinguish run-rate math (FinOps working numbers) from accounting treatment (the controller's domain) — never present the former as the latter.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.