0) Non‑negotiables (AI reality check)

• COGS aren’t zero anymore. Inference burns cash. Price to usage intensity and outcomes, not just seats.
• Meter or cap usage everywhere AI shows up (credits, rate limits, fair use, overages). Don’t let whales bankrupt you.
• Hybrid beats purity. Keep subscriptions for predictability; layer usage for fairness and margins.

1) Your Value Metric & Metering

• Pick a single, obvious value metric users grok in 5s (e.g., "generations", "resolutions", "documents summarized").
• Tie AI cost to that metric; map feature → cost class (light/medium/heavy) before launch.
• Offer tiers by scale/quality/automation level; gate higher‑cost models and long‑context features.
• Expose live usage counters in‑product. No "mystery burn".

2) Pricing Models That Work Now (hybrids)

A. Bundled AI usage (in plans)

• Each plan includes a fixed allowance (e.g., 1,000 generations / mo). After that: soft cap (slower, queue) or overage.
• Pros: simple to buy; predictable for finance. Cons: must explain overages clearly.

B. Credits (wallet)

• Plans ship with credits; users top‑up. Credits roll over for X months. Show unit cost equivalents.
• Use credits when different features burn compute unevenly. Avoid hidden conversion math; show "this task ≈ 24–40 credits".

C. AI as Add‑on

• "+AI" bundle per seat/org (unlocks set of AI features; sensible fair‑use). Great for existing SaaS with broad attach.

D. Outcome‑based

• Charge per resolution / lead / document shipped when outcome is clear, countable, and caps runaway usage.

E. Multi‑Model Bundling

• Instead of charging separately for each LLM/API, offer one price for access to multiple models (OpenAI, Anthropic, Mistral, etc.).
• Deliver via open‑router style aggregation or an internal orchestration layer; route requests to the optimal model for cost/quality.
• Sell it as a simplicity + savings bundle: one contract, predictable billing, best‑model routing.
• Price tiers by models included, context length, and routing intelligence.
• Great for buyers who want multi‑model optionality without the API billing chaos.

3) Free: Trial vs Tier vs New Model Hype

Reverse Trial (default)

• Start new free users on all premium features for 7–14 days (no card). After, drop to free with teasers + saved artifacts.

Classic Trial

• Short (7–14d). No CC. Auto‑save work; downgrade gracefully, never delete output.

Freemium Tier

• Purpose: learn, seed loops, enable light personal use.
• Limit via: credits, context length, queue speed, watermark/branding, no commercial use.

New Model Launches

• Free, time‑boxed spikes to drive adoption/training. Gate via waitlists, quotas, and benchmark challenges to earn more.

4) Discounts That Don’t Kill You

• Startup: time‑boxed (6–12 mo), clear usage caps, automatic review at renewal.
• Education/Student: 50–70% but non‑commercial, fixed duration; require verification.
• Never discount enterprise without removing cost drivers or scope.

5) Seat vs Usage vs Outcome — When to Use Each

• Per seat: collaboration tools where value scales with people; AI cost per seat roughly even.
• Usage/credits: heterogeneous workloads; fairness + margin control.
• Outcome: narrow, measurable impact (e.g., support per resolution). Avoid if outcomes are fuzzy or disputable.

6) API & Platform Pricing (consumption‑based)

• Sell compute‑aligned units (tokens / seconds / images / jobs). Offer bundles with volume price breaks.
• Auto‑replenish (opt‑in) with thresholds; webhooks for burn alerts; hard rate limits by key.
• Publish model matrix (quality, latency, price). Sunset old SKUs with generous migration.

7) Page Anatomy That Converts (DNA)

• Above‑the‑fold: clear value metric, simple plan grid, Most Popular flagged, monthly/annual toggle (+save %).
• Plan cards: one‑line who it’s for, 3–5 capabilities, explicit usage allowances, overage rate, fair‑use note.
• Sticky FAQ under grid: trials, overages, refunds, commercial use, data/privacy, SSO, export.
• Right rail: social proof (logos, short case metrics), and an objection killer block ("Will I blow my budget?" → explain caps).
• CTA stack: Start free + Talk to sales + Calculator.

8) Price Presentation & Nudges

• Default annual (show monthly equivalent and save 20–30%).
• Anchor with a premium tier; use decoy to steer to Most Popular.
• Usage calculator: sliders for team size & workload → shows monthly estimate + confidence range.
• Localize currency; show tax/VAT handling; transparent per‑add‑on pricing.

9) Guardrails & Abuse Control

• Per‑org rate limits; per‑user caps; high‑cost actions require confirmations.
• Quality‑tier throttling: high‑end models limited on lower plans.
• Kill switches: auto‑pause on anomalous spikes; instant key rotation; export usage audit.

10) Metrics to Run the System

• Trial → paid, Reverse‑trial → paid, Free → paid, Plan mix, Attach rate for AI add‑on.
• Unit economics: gross margin by plan, by workload cohort; overage revenue %; credits breakage.
• Abuse/variance: 95th/99th %ile cost per active; anomaly count.
• Page: pricing → checkout CTR, checkout → pay, annual vs monthly mix.

11) Enterprise & Security

• Public enterprise card with: SSO/SCIM, RBAC, data retention, private model routing, on‑prem/VPC, DPA, SOC2/ISO.
• Clear starting at price + contact. Enterprise calculator for committed‑use deals (prepaid credits + floors).

12) Opinionated Playbooks (use as is)

Startup ≤$1M ARR

• 3 tiers + AI add‑on. Reverse trial 14d. Credits in all paid tiers. Annual default (25% off). Student & startup programs.

$1–10M ARR

• Add outcome SKU if you have a crisp outcome. Ship calculator. Turn on overages with guardrails. Introduce enterprise commitments.

10M+ ARR

• Segment by vertical; private pricing for high‑COGS workloads; prepaid commit with drawdown; launch model‑specific bundles.

13) Copy Blocks You Can Paste

• Fair Use: "Premium AI features include a monthly allowance. We’ll warn you at 80%, then throttle or bill overage at $X per Y. Your work is always saved."
• Budget Safety: "Set a hard monthly cap. We’ll pause AI actions and notify admins when you hit it."
• Why Credits: "Different AI tasks burn different amounts of compute. Credits keep pricing fair—pay only for what you use."

Appendix

Free New Models

• Time‑boxed launches
• earn more usage via contributions
• Quotas + waitlist.
• Increase adoption
• Train the model

Free trial

• Prefer reverse trial.
• 7–14 days.
• No credit card required
• Strong upgrade moment when value is fresh.

Free tier

• Offer a taste of the product
• Keep forever‑free
• restrict credits, models, latency, commercial use
• watermark output
• upsell with saved work

Startup discounts

• 70-90% off
• Deep but time‑boxed
• Pay when you scale (6months to 1 year)
• verification
• cap usage

Student Discounts

• 50-70% off for students
• Limited period of time
• auto‑review.

Pay per month

• No commitment
• More usage
• Extensions or Integrations

Per seat per Month

• Keep seats where collaboration drives value
• add AI Add‑on or bundled credits
• publish overage

“Most Popular”

• or “recommended”
• Use it as the value anchor
• Promote a specific offer
• Price‑position the decoy just above it.

APIs (Consumption)

• Credits/top‑ups
• roll‑over window
• auto‑renew thresholds
• explicit unit prices
• model matrix

Yearly Plan

• 20–30% savings
• default selected
• show monthly equivalent
• easy upgrade/downgrade

Collaboration

• Highlight collaboration features per tier
• Features like: Software Systems, roles, etc…
• Centralized admin tools

Enterprise

• Custom pricing, custom sales process
• Custom support
• Advanced security
• Advanced RBAC
• Analytics