strategy · hard
How should this company monetize?
How should this company monetize its product?
Monetization questions test whether you can run a real revenue decision, not whether you know the names of five business models. A weak answer lists options and picks one by category association (“I’d go freemium, like Spotify”). A strong answer runs decision logic against three constraints: what value event you can charge for, what the cost structure permits, and how your buyer budgets.
Why the standard framework fails here
Most PM prep covers a generic nine-step approach: describe product, clarify, segment users, identify pain points, evaluate revenue models, pick one, define metrics. Procedurally correct, practically useless. The real failure mode is skipping the cost constraint entirely, which in 2026 is the most disqualifying gap for AI products. For AI-native products, gross margins compress from the typical SaaS 80-90% range to 50-60% because inference is not free (Bessemer data). Heavy users under flat pricing can consume 100 to 1,000 times more resources than light users. A candidate who ignores this is not running a viable monetization analysis.
The three-axis decision
Before naming a model, map the product against three axes.
Axis 1: What is the value event? Is there a discrete, measurable outcome (ticket resolved, legal document generated, deal closed) that outcome-based pricing can latch onto? Or is value diffuse (productivity uplift, faster search), which makes per-outcome charging hard and pushes toward subscription? Intercom Fin charges $0.99 per resolved ticket. EvenUp charges per completed legal demand letter. Both have a crisp, auditable outcome definition. WhatsApp had diffuse network value with no clear per-message value event, which is why Meta kept it free and monetized through business APIs and ads in adjacent products.
Axis 2: What is the cost structure? If the product has real inference costs per query, flat per-seat pricing creates a margin problem on your heaviest users. 56% of AI company leaders now use hybrid pricing (base platform fee plus usage tiers); 38% use purely usage-based, per a 2025 Stripe survey of AI companies. GitHub Copilot and Microsoft Copilot at $30/user/month work because per-user inference cost runs below 15-20% of the subscription price and buyers want predictability. Cursor’s $7,225 overage invoice became a public crisis because users were not informed how credits translated to cost under a new model. Cost instrumentation first, pricing model second.
Axis 3: How does your buyer budget? Enterprise procurement needs line-item predictability, which favors a base platform fee with defined usage tiers. Developers and consumers tolerate consumption-based pricing because they can cap or optimize their own usage. B2B buyers whose 2024-2025 AI pilots are hitting renewal in 2026 are not approving renewal on productivity vibes. They need measurable ROI: resolved tickets, completed documents, cost savings. Outcome-based pricing wins renewal conversations because it puts the number on the table.
The three AI product archetypes
Bessemer identifies three archetypes that map directly to pricing model selection:
- Copilots (user in the loop, bounded session length): per-seat subscription works when inference cost per user is predictable. GitHub Copilot is the canonical example.
- Agents (autonomous tasks, variable compute per task): outcome-based or consumption-based pricing is required. Flat subscription on an agent product is structurally wrong unless you cap usage tightly.
- AI-enabled services (workflow-based, human plus AI): consumption or milestone-based pricing tied to workflow completion.
Name which archetype applies, then derive the model. This is the decision logic that generic frameworks omit.
Structure a strong answer
strong
"Before I name a model, three clarifying questions: Is this pre-monetization or optimizing existing revenue? Is it consumer or enterprise? And is this AI-native with real inference costs per session? Those three answers constrain what is even viable.
If it's B2B and AI-native, I'm looking at a hybrid base plus consumption model or outcome-based. If it's consumer and network-driven, ads or indirect monetization through an adjacent product are more viable than per-seat fees. I'd also ask what stage the company is at: outcome-based pricing requires years of resolution data and the ability to measure the outcome unambiguously. A pre-scale product usually cannot defend that model to a skeptical buyer yet.
For a B2B AI agent product, I'd run the three-axis test. Value event: can I define and measure a discrete outcome? If yes, outcome-based at a per-unit rate is the strongest alignment with buyer ROI. Intercom Fin at $0.99 per resolved ticket is the clearest market example. If the outcome is harder to measure, I need a hybrid: platform fee to cover the compute floor, plus consumption tiers that grow with usage.
Cost structure: if heavy users can consume 100x more inference than light users at the same seat fee, flat pricing destroys margin. I'd model cost-per-user by tier from day one before announcing a price, not discover the problem after an invoice incident.
Buyer budgeting: enterprise buyers need predictability, so I'd structure a base commitment with a defined usage envelope and overage at a known rate. That passes procurement and survives the 2026 renewal conversation where soft ROI no longer holds.
Success metrics: net revenue retention above 120% (the model grows with customer value), gross margin above 60% at scale (cost structure holds), payback period under 18 months. If NRR is below 100%, the model is not aligned with the value it actually delivers."
weak
"There are several monetization models: ads, subscriptions, freemium, and usage-based. I'd probably go with freemium, similar to Spotify, since it lets you grow the user base and then convert to paid. You could offer a free tier with basic features and a premium tier with advanced capabilities." This answer would be identical for any product brief in any category. It ignores cost structure entirely, picks a model by brand association, offers no decision logic, and has no awareness of AI-specific margin dynamics. The tell: nothing changes if you swap in a different product.
What interviewers penalize
Three gaps get candidates cut at AI-native companies specifically. First, jumping to a model before establishing the cost floor. Second, proposing outcome-based pricing without addressing how you define and measure the outcome reliably; an outcome model you cannot audit is a billing dispute waiting to happen. Third, ignoring what the monetization model does to GTM, sales motion, and team incentives. Usage-based pricing makes sales forecasting harder. Outcome-based pricing requires customer success investment to verify resolution. These downstream effects are what separate a candidate who has run a real monetization decision from one who studied vocabulary.
The 2026 renewal context
The 2024-2025 AI adoption wave ran on pilot budgets and soft productivity ROI. Those pilots are hitting renewal in 2026, and buyers are demanding hard ROI. A monetization model that worked during initial land-and-expand now has to survive a conversation where the customer asks for measurable cost savings, tickets resolved, or documents completed. Credit-based AI pricing models grew 126% year-over-year as companies tried to align revenue with demonstrated usage value. The model you propose needs to be viable at renewal, not just at first sale. In 2026, feasibility is largely solved; the interview is testing whether your monetization design is structurally sound enough to hold when a buyer demands to see the math.
See also: How would you price an AI product? for the token cost math behind these constraints, proving viability for how to frame the business case before the pricing decision, and LLM unit economics for the compute cost floor that shapes which models are even viable.