The most common question we hear from early-stage founders today is deceptively simple: "How should I price my AI product?"
Getting this wrong can cripple your company. Founders are wrestling with fundamental questions about value capture in the AI era, including:
- Should I use seat, usage, or outcome-based pricing? Or a hybrid?
- How do I price when my product replaces human labor?
- How should falling LLM costs factor into my pricing strategy?
To help answer these questions, Jake Saper, GP at Emergence Capital, hosted a lively discussion at the Emergence offices with two of the best minds in pricing: Madhavan Ramanujam and Josh Bloom—both GPs at 49 Palms Ventures and former Managing Partners at Simon-Kucher, where they advised 500+ companies on pricing. Madhavan also wrote "Monetizing Innovation", the most widely read book on pricing strategy.
Here are the key insights:
What Has Changed About Pricing in the Age of AI
- Three fundamental shifts are reshaping software pricing: AI enables true outcome-based pricing. Unlike traditional SaaS, AI can be tied directly to business results—though attribution challenges remain.
- AI taps labor budgets, not just software budgets. But as AI reduces headcount, per-seat models become self-defeating.
- LLM usage creates new cost dynamics. While inference costs are declining, they remain a significant factor in unit economics.
Determining How Much to Charge for Your AI Product
%202-min.jpg)
Despite all that's changed, one principle remains constant: have "willingness to pay" conversations early—ideally during product development.
1. Bring pricing into sales conversations immediately
Technical founders often avoid early pricing discussions, thinking it feels "salesy." This is a critical mistake.
Early pricing conversations accomplish several objectives:
- Qualify buyers with real intent and budget
- Anchor discussions around value, not features
- Co-develop ROI models with customers
- Prioritize your roadmap based on what customers value
If prospects resist ROI discussions, propose a value audit—a post-pilot session where you jointly calculate actual value delivered. This shifts pricing conversations to when real usage data exists.
Value audits are particularly powerful for AI products, which often deliver increasing value as models improve. Use them to justify initial discounts: "I'll offer this discount, but in six months we'll conduct a value audit to reassess pricing."
Never wait until procurement to discuss pricing. Procurement's job is cost reduction, not value assessment. They'll resist outcome-based models and force you into standard SaaS buckets. Business users who experience the benefits firsthand are far more receptive to ROI-based pricing.
Even if customers don't choose outcome-based pricing, offering it signals confidence and willingness to share risk. Present it alongside fixed-fee options—the comparison makes your fixed fee appear fairer and lower-risk. When customers push back on price, you can point to the outcome-based alternative as a lower-risk option.
2. Calculate ROI holistically
Most ROI analyses focus on hard savings (labor/vendor reduction) or revenue uplift. While attribution can be challenging, establishing an agreed framework upfront simplifies post-deployment calculations.
Critical insight: Work with buyers to agree on ROI model inputs before deployment. Once inputs are locked, outputs become harder to challenge.
As you identify ROI inputs, don't overlook:
- Time efficiency gains and their opportunity cost
- Implementation cost advantages over traditional SaaS
With proper ROI frameworks, AI products are capturing 25-50% of created value—significantly higher than traditional SaaS's 10-20%.
3. Use the “acceptable, expensive, prohibitively expensive” framework
To gauge willingness to pay, ask three questions:
- "What would be an acceptable price?"
- "What would be an expensive price?"
- "What would be a prohibitively expensive price?"
Willingness to pay typically lands near the "expensive" point. Rahul Vohra used this simple technique from "Monetizing Innovation" to price Superhuman.
Evaluating AI Pricing Models
Software pricing has evolved dramatically over 30 years, with category-defining companies pioneering new models:
- Salesforce (Emergence’s first investment) —> seat-based pricing
- AWS —> consumption-based pricing (pay for what you use)
- Uber —> dynamic pricing
The category-defining companies in the AI era will also likely popularize new pricing models. We’re already seeing a lot of experimentation on this front, though it’s still early days.
To help make sense of the various options, we’ve put together a chart laying out the major pricing models we see AI companies adopt today and a grading rubric for each model.
Some notable points:
- Per seat:
- Example: Notion (AI features bundled with per seat).
- Pros:
- Simple, predictable, familiar to buyers
- Cons
- Misaligned with value—reducing headcount reduces your addressable seats. Power users can destroy margins through heavy compute usage.
- Per action:
- Examples: Voice AI companies like Bland and Regal (charging per minute).
- Pros
- Simple, margin-aligned
- Cons
- Penalizes efficiency (faster calls = less revenue).
- Invites commoditization when competitors use identical units. To mitigate this risk, consider adding another pricing dimension to create more of a hybrid model.
- Per agent:
- Per completed workflow:
- Examples: Salesforce’s Agentforce which uses credits for successfully completed tasks (e.g., populating a database)
- Pros:
- Balances consumption and outcomes, making it ideal for complex but standardized processes.
- Cons:
- More complex implementation
- Per business outcome:
- Examples: Popular in support companies like Assembled and Intercom’s Fin (per successful resolution)
- Pros:
- Strongest value capture, clear differentiation
- Cons:
- Attribution challenges—requires clear evidence of causation
Determining the Right AI Pricing Model for You
Madhavan and Josh described a 2x2 to help founders decide on a pricing model based on two key dimensions, as covered in the upcoming book Scaling Innovation:
- Autonomy: How independently does the AI operate? Does it augment a human or replace them entirely?
- Attribution: How clearly can the AI’s actions be linked to measurable outcomes?
There are four strategic AI pricing zones:
1. Low Autonomy + Low Attribution → Seat-Based Pricing (Bottom Left)
AI assists users in co-pilot fashion, but lacks measurable impact. Common in productivity tools (e.g., email summarizers), this model prices per seat. It’s easy to implement but limits upside—boosting attribution is key to unlocking more value.
2. Low Autonomy + High Attribution → Hybrid Pricing (Bottom Right)
AI assists users in co-pilot fashion and its value is provable. Combine seat-based pricing with usage metrics (e.g., credits or tiers). You could also combine seat/usage model with outcomes-based. Useful in domains like finance or legal where time savings are quantifiable. Gradually deepen attribution and autonomy to transition fully toward outcome-based models.
3. Autonomous + Low Attribution → Usage-Based Pricing (Top Left)
AI works independently but its business impact is unclear. Use volume- or task-based pricing, often seen in backend automation (e.g., document processing). Usage becomes the closest stand-in for value delivered.
4. Autonomous + High Attribution → Outcome-Based Pricing (Top Right)
The AI replaces human effort and delivers measurable results (e.g., cost savings, increased sales). This is the ideal scenario—charge based on outcomes and capture a share of the upside. It requires strong attribution, trust, and consistent value proof.
While the authors all collectively believe that outcome-based models are the future of AI monetization, we are also sanguine about the fact that many companies haven’t built out the product autonomy and value attribution necessary to tap into the model yet. We would prefer companies start out in whichever quadrant you find yourself in today, yet set a goal to evolve over time.
The State of AI Pricing Today
AI is rapidly disrupting traditional pricing models. A year ago, half of software used seat-based or flat-fee pricing. Today, hybrid models dominate, combining elements of seat, usage, per agent, and outcome pricing.
This shift reflects the need for experimentation. Hybrid models offer flexibility while founders learn what works in this new landscape.
The Future of AI Pricing
While it will be a long road with lots of detours and bumps, we think the arc of AI history will bend towards paying for results.
Buyers never wanted to buy software- they wanted solutions. As AI autonomously delivers those solutions, pricing will evolve to charge for outcomes.
Some may raise objections around the challenges of attribution, buyer preference, and more. We’d argue that, over time, those will be solved. Indeed, buyers will eventually demand outcome-based pricing.
Consider a future where AI makes code a commodity. Any feature ships instantly. Value shifts from feature-building to execution guarantees. You're not selling software—you're selling outcome insurance.
It's a humbling reminder: customers aren’t buying tech, they’re buying trust that something critical will work. To be clear, we don’t think this world of total code commoditization is just around the corner (there’s a lot of value in a vendor focused on solving a specific problem crafting the experience for the user). But this thought exercise helps clarify the direction we need to head with our pricing models.
Survey data from Growth Unhinged confirms this direction: while only 5% of companies currently use outcome-based pricing, 25% expect to by 2028.
Today, perhaps the most straightforward business model to charge on an outcomes basis is AI-enabled services. Service businesses naturally own end-to-end execution with clear attribution. Mechanical Orchard is a great example; they use AI to move mainframe workloads into the cloud. They take ownership of the entire process, and thus are well suited to charge for success.
Emergence has written extensively about building AI-enabled services businesses - refer here and here for more.
Key to Outcome-based Model Success: Getting Proof of Concept (POC) Motions Right
Why POCs Matter More Than You Think
- POCs shape your commercial motion: These interactions teach you how buyers evaluate value and make decisions. Treat them as commercial experiments, not just technical validations.
- Choose design wins carefully: Early adopters set expectations for your market. Select partners with real pain, co-building willingness, and industry influence—not just friendly champions.
- Structure prevents drift: Unstructured pilots become costly distractions with vague results and no path to revenue.
Should You Charge for POCs? Yes—But Smartly
Avoid the "free forever" trap. Endless free pilots drain resources without revenue.
Price signals seriousness. Even nominal fees filter serious buyers from tire-kickers.
Frame pilots as fixed-fee engagements: Say "We structure this as a 4-week, fixed-fee engagement to quantify value and build your business case" rather than "This POC is free."
Clarify pricing expectations: If your pilot costs $5K but commercial deals are $100K-$300K based on the value unlocked, state this explicitly to avoid anchoring.
Get alternative commitments if cash isn't possible: Letters of intent, logo rights, executive access—these signal seriousness when budgets are locked.
5 Rules for High-Impact POCs
- Start with success criteria, not scope: Define KPIs, outcomes, and decision-makers upfront.
- Time-box ruthlessly: 30-90 days maximum with weekly checkpoints. Open timelines kill momentum.
- Pre-commit next steps: "If we hit these metrics, will you present to stakeholders and move to commercial discussions?"
- Demand full buying center access: Technical users alone can't close deals. Meet decision-makers during the POC.
Document like a contract: Formalize scope, terms, and deliverables. This elevates the engagement beyond a casual test.
Final Thoughts
While it’s critical to think deeply about these models, don’t let analysis paralysis set in. Experiment with different models and iterate quickly. Try hybrid models for flexibility and learning.
That said, don’t let the model you choose get too complex for the buyer or even your team to grok. Ask yourself, “Can a junior sales rep 2-3 years out of college explain this easily to prospects?”
As your AI gets more autonomous and attribution becomes clearer, work towards more outcomes-based models over time. The companies that define the AI era will be those that align their pricing with the value they create.
Let us know what you’re learning as you experiment. We’re excited to keep learning from the folks on the field and plan to update our point of view as the market matures.
jake@emcap.com
rishub@emcap.com
mram@49palmsvc.com
joshua@49palmsvc.com
