AI Product Pricing Strategy Template: From Value Metric to Tier Design

How to Use This Template

Work through the five sections in order. Each section builds on the previous one: you cannot set tier architecture until you have defined your value metric, and you cannot set launch pricing until you understand unit economics. Expect to spend 2 to 4 hours filling this out the first time; revisit it quarterly as your cost structure, competitive position, and customer data evolve.

What this template does NOT cover

Enterprise custom pricing negotiations (those require your AE and a separate deal-specific analysis)
International pricing and currency localization
Channel pricing (resellers, OEM, API licensing to partners)
Marketplace pricing on third-party platforms

The template uses fill-in prompts in [brackets]. Replace each one with your actual data. Where you do not have data yet, note it explicitly: gaps in the template are decisions you need to make before you price, not after.

Section 1: Value Metric and Unit Economics

The value metric is the unit by which customers pay for your product. It should correlate with the value they receive, not with your cost to serve. Common AI product value metrics:

Per seat / per user

Use when: When all users get roughly equal value and usage is broad. Works well for productivity tools and copilots where every user has the assistant embedded in their workflow.

Risk: Incentivizes customers to buy the minimum number of seats and share credentials.

Per output unit

Use when: When customers have a clear unit of work: documents generated, calls analyzed, emails processed. The value per output is measurable and consistent across customers.

Risk: Customers optimize for output count, not quality. They may churn if output count plateaus even if value is high.

Per token consumed

Use when: API products and developer-facing tools where usage varies enormously between customers. Matches cost structure directly.

Risk: Unpredictable costs for customers lead to anxiety and underuse. Rarely works as a primary pricing lever for non-technical buyers.

Percentage of value created

Use when: Outcome-based pricing where the AI clearly drives a measurable result: calls converted, revenue recovered, time saved. Requires reliable attribution.

Risk: Hard to implement without deep instrumentation. Customers dispute attribution. Builds strong relationships when it works.

Hybrid (platform fee + usage)

Use when: The most common production pattern: a fixed monthly fee for access, plus a variable component that scales with usage. Gives customers cost predictability while preserving upside.

Risk: More complex to communicate; requires more sales education.

Template: Value Metric Decision

Our primary value metric is: [seat / output unit / token / % of value / hybrid]

We chose this because customers measure value in: [describe how customers quantify the ROI of your product]

Our variable unit is: [if hybrid: what triggers the variable charge and how it is measured]

Average monthly variable usage per account at current pricing: [$X per account]

One metric we considered and rejected: [alternative metric] because [reason it creates misalignment]

Unit Economics Inputs

Average AI inference cost per unit

[$ per document / per call / per 1K tokens — pull from your API invoice]

Gross margin target

[your finance team's target, typically 60-80% for SaaS]

Cost floor (minimum viable price)

[$X per unit = inference cost / (1 - target margin)]

Current average revenue per account (ARPA)

[$X/month from CRM or billing system]

Cost per account at current usage

[$X/month from API invoices divided by active accounts]

Gross margin per account today

[(ARPA - cost per account) / ARPA * 100 = X%]

Section 2: Tier Architecture

Tiers serve two purposes: natural upsell paths and customer self-selection. A good tier architecture means customers are never buying more than they need and are always one tier below where they want to be as they grow.

Most AI products run three tiers. A fourth enterprise tier may be appropriate if your enterprise sales motion is distinct from your self-serve motion.

Free or Trial

Acquisition. Not monetization.

Limit principle: Enough to validate the core value proposition, not enough to run real workloads.

Upgrade trigger: The conversion trigger should be a usage limit the customer hits doing something valuable, not a paywall they hit before seeing value.

Our free tier allows: [X actions per month / X seats / X tokens]. Customers convert to paid when they: [describe the natural friction point].

Starter / Pro

Self-serve revenue from individual users or small teams.

Limit principle: Volume and feature gates that self-serve customers stay under but teams quickly hit.

Upgrade trigger: The conversion trigger to Growth/Business is collaboration features, admin controls, higher limits, or integrations that teams need.

Our Starter tier is $[X]/month for [Y seats / Z units / W tokens/month]. Growth trigger: [what causes upgrade to next tier].

Growth / Business

SMB and mid-market self-serve or low-touch sales.

Limit principle: Enough for a growing team, with missing enterprise controls (SSO, audit logs, SLA) as the conversion lever to enterprise.

Upgrade trigger: The conversion trigger to Enterprise is security requirements, custom contract needs, or usage that exceeds self-serve caps.

Our Business tier is $[X]/month for [Y seats / Z units / W tokens/month]. Enterprise trigger: [specific features or requirements not in this tier].

Enterprise (custom)

Large-account revenue with white-glove support.

Limit principle: Custom volume, dedicated support SLA, custom security and compliance features.

Upgrade trigger: Customers are identified by AE and routed out of self-serve by sales team. Pricing by negotiation from a minimum ACV floor.

Our Enterprise minimum ACV is $[X]/year. Custom features include: [SSO, SCIM, audit logs, custom DPA, dedicated CSM, etc.].

The AI-specific tier design mistake

The most common mistake in AI product tier design is gating the best model behind the highest tier. This creates a quality cliff: free and starter users see a meaningfully worse product. They churn before converting because the product they trialed is not the product they are being sold. Gate on volume, features, and support; not on model quality. Give every tier access to your best model in the limits appropriate for that tier.

Build AI Products That Grow Revenue

The AI PM Masterclass covers pricing strategy, growth levers, and how to build AI products that convert. Taught live by a Salesforce Sr. Director PM.

Section 3: Competitive Positioning

Pricing is relative. Customers make buy decisions by comparing your price to alternatives, not by evaluating your unit economics. Competitive positioning in the template means two things: knowing where you sit relative to alternatives, and choosing deliberately where you want to sit.

Template: Competitive Pricing Map

Primary competitor 1: [name] charges [$X/month or $Y/unit] for [comparable tier/plan]. Key pricing difference: [how their model differs from yours].

Primary competitor 2: [name] charges [$X/month or $Y/unit] for [comparable tier/plan]. Key pricing difference: [how their model differs from yours].

The customer's best alternative if they do not buy us: [internal tool / spreadsheet / hiring someone / doing nothing / another category]. Cost of that alternative: [$X in time, money, or risk].

Our positioning choice: [premium at X% above competitors / at parity / below market to drive share]. We are choosing this because: [strategic rationale].

The price ceiling (what the market will not pay above): [$X, based on customer interviews / willingness to pay research / lost deals at price point $Y].

Price on value, not cost

Your inference costs may be $0.02 per document. If that document used to require 2 hours of analyst time at $75/hour, the value is $150. Pricing at $0.50 leaves 99.7% of value on the table. The cost floor sets a minimum; willingness to pay sets the target.

Anchor up before anchoring down

Show your enterprise tier price before your starter price. The high number anchors the customer's reference frame, making the starter tier feel accessible rather than cheap.

Name tiers after outcomes, not sizes

'Starter / Growth / Enterprise' is weaker than 'Individuals / Teams / Organizations' or 'Explore / Build / Scale.' Tier names set customer expectations about who the tier is for.

Section 4: Launch Pricing and Migration Plan

Launch pricing for AI products has two phases: beta pricing (discounted or free to gather data) and GA pricing (the real price). Most teams fail to define the transition between them, which creates two problems: customers treat beta pricing as permanent, and the team has no data to justify the GA price.

Template: Launch Pricing Decision

Beta pricing structure: [free / heavy discount / pay-what-you-want]. Rationale: [what data you are trading for the discount].

Beta customer commitment: [X interviews / Y usage hours / Z reference calls]. This is what beta pricing buys you, not goodwill.

GA pricing transition date: [specific date, not "when we're ready"]. Beta customers receive [X weeks] advance notice.

Grandfathering policy: [beta customers grandfathered for X months at beta pricing / move to GA immediately / custom case-by-case]. State this in writing during beta.

Price increase mechanism: [how future price increases are communicated, how much advance notice, whether existing customers are locked in at current pricing for contract term].

For products repricing (not launching fresh), the migration plan matters more than the new price itself. Customers accept price increases when they understand the value delivered since last pricing. They churn when a price increase feels arbitrary.

Value delivered recap

Document specifically what has improved since the last pricing decision: new features, model upgrades, performance gains, reliability improvements. This is the narrative that makes a price increase rational.

Segmented impact analysis

Which customer segments are most affected by the new pricing? High-usage free-tier users, customers near tier limits, long-term grandfathered customers. Segment the impact before communicating.

Migration path per segment

Low-usage customers can absorb tier changes easily. High-usage customers need a plan: a longer grace period, a custom plan, or a direct call from their CSM. Do not apply one policy to all segments.

Rollout sequencing

New customers see new pricing immediately. Existing customers migrate on their renewal date. Customers with annual contracts are held at current pricing for the contract term. This sequencing minimizes churn spike.