Note: We’re hosting an event in our San Francisco office for the most ambitious founders and investors interested in exploring the future of AI pricing. If that’s you, register to join here.

Companies all over the world are currently onboarding AI workforces. It’s a profound shift – a fusion of the software market and the labor market. And it’s forcing founders to answer a key question: how much is this work worth, and how much should we be charging for it?

We’re working with dozens of companies on the edge of this shift. Pricing is one of the biggest strategic challenges they’re facing. Why? Because the old playbooks don’t work anymore. For four key reasons:

• Cost Pressures. This is the first time in the history of software where the marginal cost of adding more users is not near-zero, so nailing your pricing is essential to unit economics.
• AI is allowing startups to “sell work” and “outcomes” – revenue increases, time savings, cost reductions. This is a huge opportunity because labor budgets > software budgets. But the ability to measure those outcomes and directly attribute them to your product is crucial.
• A pilot budget is not the same as a workflow budget. Conflating experimental revenue with sustainable revenue is one of the costliest mistakes AI startups make.

AI pricing can’t be copy-and-pasted from previous tech waves. It demands understanding pricing from first principles. Here’s what we’re seeing, and how to apply it.

Why AI Has Created a New Pricing Wave

Pricing is about telling a story of value. Your customer has to see value in your product, recognize that you’ve delivered on it, and be willing to exchange money for that story.

But what many forget is that the “normal” ways we pay for products today didn’t always exist.

Every decade or so, industries fall into recognizable pricing patterns: it feels obvious now to pay monthly for a streaming service, or per-seat for a SaaS product. But these weren’t natural inevitabilities — they were strategic inventions. Salesforce reinvented enterprise software pricing, moving from six-figure upfront installations to per-user monthly subscriptions. In the late 2000s, everyone was used to paying per song for music – until companies like Spotify introduced the idea of using streaming to “own” a limitless library for a fixed fee.

Technology makes something newly possible, and startups craft the story that tells customers just how valuable that change is. That story is called the “value story” – and it becomes the foundation we use to create new pricing models along with new tech waves.

Over time, what once seemed novel – leasing, subscriptions, streaming, tokens – hardens into the “normal” way customers expect to pay. Once the language is set, everyone starts using it.

We’ve actually seen five of these pricing-based waves over the last twenty years or so. (We’ve outlined them in the chart at the end of this section.) Each time, we’ve seen a value story emerge that positions the new technology as a win for the customer, compared to what came before.

The AI pricing world feels uncertain because there are actually many optimal ways to tell a value story – and it differs largely depending on your product and your customer. Some AI companies still work with seat-based pricing, others are best using hybrid approaches, and a small percent are pioneering outcome-based models. This is a shift from the past few years or so, where pricing innovation had largely stalled.

Today, you need to understand this new technology, the new ways it can impact your customers, and what actually matters to them. Then, you can build a strong pricing strategy, and value story around those principles.

Eventually, new pricing models will calcify. It will get easier as we learn what works. (Hint: we are all headed towards outcome-based pricing). But for now, you need to deeply understand the mechanics of pricing if you want to succeed.

*Note that many of these pricing models still exist today. Think of them as overlapping ideas, rather than distinct moments in time.

Part I: How To Approach the AI Pricing Problem

We recently sat down with Madhavan Ramanujam, the author of Monetizing Innovation and Scaling Innovation, to discuss how early stage founders should be thinking about pricing in the new AI wave. Madhavan has worked with hundreds of companies on optimizing their pricing strategies (including Trulia). Today, he is a leading voice in the AI pricing conversation.

He recommends that companies discover what pricing model works for them by plotting their product along two axes: autonomy and attribution.

• Autonomy: refers to whether your AI product still needs humans in the loop (like a co-pilot) or can operate independently (like an autonomous agent)
• Attribution: refers to whether the value your AI generates can be clearly measured and attributed to its work.

Once you locate your company within this matrix, it’s far easier to choose a pricing model, and craft a values story that fits your technology and customer. Madhavan created this matrix after working with hundreds of top AI companies on their pricing strategies.

Here’s how the matrix quadrants break down:

Bottom Left: Low Autonomy, Low Attribution.

Productivity may improve, but the impact is hard to measure and tie to the product. Value still depends on the humans operating the product, so seat-based pricing still makes sense. This is the realm of companies like Slack, Figma and Grammarly.

Bottom Right: Low Autonomy, High Attribution

This is co-pilot land. Humans are the primary users, but it’s the AI creating measurable value – like lines of code generated, or warm leads created. Many emerging AI companies like Cursor and Clay live in this zone.

Top-Left: High Autonomy, Low Attribution

Here, the product is largely autonomous, but value capture isn’t direct. This is the world of infrastructure services that are essential to operations, but don’t drive the top-line value. Think of products like Twilio or AWS. For these, usage-based pricing makes sense.

Top-Right: High Autonomy, High Attribution

This is where the field is headed. Autonomous AI that drives measurable value. At this point, very few companies can actually clear this bar. Madhavan estimates only about 5% are in this quadrant today, but predicts that it will grow to about 25-30% in the next few years.

For early-stage AI startups, the reality is that most land on the right-hand side of the matrix: attribution is measurable, but autonomy is still developing. The key to unlocking the right pricing here is to design pilot programs that clearly demonstrate attribution and value delivery.

If you can prove impact, customers will pay. Here’s how to do it.

Part II: Run Pilots That Turns Experimental Revenue into Real Revenue

The dominant GTM method for many AI companies (at least in the B2B space) is still the POC or pilot project.

There are strong tailwinds driving adoption of AI pilots. No one wants to miss the AI boat, so enterprises and SMBs are very open to conversations about integrating AI. But getting enthusiasm for a pilot and actually running an effective one are two different things.

The truth is many founders fumble here. A recent MIT study suggested that only 5% of AI pilots actually deliver value. This isn’t because AI itself isn’t beneficial, it’s that many companies don’t know how to define and run effective pilots that measure real value.

In our recent conversation with Madhavan, he outlined three traps founders commonly fall into when designing pilots:

1. Pilots are not pure technical validation. They are an opportunity for you to demonstrate value in terms that are relevant to your customer, not just your internal engineering team.
2. Pilots should not be free – this trains your customers to expect more for less, and it often attracts non-serious buyers.
3. They fail to set measurable success parameters in the language of the customer, i.e. to co-create an ROI model with the buyer.

Traps 1 and 2 are easy to avoid. But trap 3 is the killer. The hardest (and most important) part of designing an AI pilot is proving that you can deliver measurable value.

Successful pilots are structured around a value story that resonates with a specific internal champion at your customer’s organization. The goal is to translate what your product does into ROI, and the metrics that the decision-maker already cares about. Madhavan has also outlined a few stories that tend to help founders begin designing these POCs.

The Incremental Top Line Story: Show that your AI can drive revenue impact, reduce churn, or improve conversion. This is appealing especially to revenue-focused executives.

The Cost Savings Story: Show that your AI can reduce licensing costs, reduce headcount, introduce operational efficiencies. This works especially well with CFOs.

The Opportunity Cost Story: Show that your AI can free up time for higher-value work. CEOs and management teams tend to focus here.

Today, pricing success is not about pulling a model off the shelf; it’s about communicating value in terms your internal champions care about (and can explain themselves). This is critical, because it sets up your own transition within the organization from “experimental AI pilot” to actual workflow partner.

That’s the inflection point where pilots turn into recurring, sustainable revenue.

Part III: Don’t Mistake Pricing Strategy for Defensibility

Earlier, we mentioned that startups typically lead charge in translating new technologies into new value stories. This opens up an important point: pricing is a wedge, especially if you’re competing against incumbents.

Incumbents are rarely incentivized to change their pricing models (unless something major happens). They’re usually slow to the game. Google, for example, was late to the game when it came to realizing that AI is starting to encroach on the search category. Why? Google created the SEO business. Why would they invest in technology that undermines paying for traffic, search ranking and clicks? It’s classic Innovator’s Dilemma dynamics.

Today, they’re now in the game (AI overview was an early attempt). But it allowed room for AI-native search engines, like Perplexity, to gain significant ground.

Startups can often attack incumbents through pricing innovation. But it’s an attack strategy – nothing more. Often, pricing doesn’t lead to sustained defensibility over time.

If your primary differentiation is a better pricing model, competitors can copy that model relatively quickly. This creates a strategic tension: the same pricing simplicity that makes your product attractive to customers also makes it easier for competitors to replicate your approach. As more companies adopt similar pricing strategies, pricing advantages tend to erode.

The same applies to outcome-based pricing – what Madhavan calls the “holy grail” of AI pricing models. The beauty of it is that it is the ultimate “pay for success” model. (If the outcome is delivered, I pay. If it’s not, I don’t.) Adoption of a product with this model is usually faster, which is another reason why so many companies are eager to move to this model. But this will never be a long-term defensibility. You need to use that initial advantage to then build a separate moat.

Lots of companies mistake pricing innovation for defensibility. It’s easy to see why. Companies with strong network effects or other defensibilities often gain significant pricing power over time. But this pricing power is a result of their defensibility, not the source of it.

Sustainable competitive advantages come from deeper moats that pricing alone cannot provide. That means network effects, scale advantages, embedding, and brand. We cover all those here.

The lesson for AI founders is nuanced: you can use pricing as a tool of disruption, but don’t mistake it for a sustained edge.

The Key Takeaways

Many AI companies struggle with pricing today, because the old playbooks don’t work anymore. This is actually an opportunity. Every new technology wave reshapes how customers perceive value. The challenge is translating that shift into concrete terms for your customers.

It’s even harder today because AI has opened up so many ways of packaging value, and so few forms of concrete language to communicate that value. This means each company finds the right pathway to prove attribution to customers on their own.

From our conversation with Madhavan, and our experience with early-stage startups, it comes down to five key steps:

1. Understand where the pricing world is going (toward outcome-based models)
2. Know where you actually are on the autonomy/attribution matrix (Likely between hybrid and outcome based)
3. Design a POC or pilot with clear value benchmarks in terms your customer cares about
4. Deliver
5. Don’t mistake pricing innovation for long-term defensibility

It’s simple to say, harder to execute. But the companies that crack the AI pricing problem are literally writing the new playbooks for the AI era. They’ll coin the terms we use to convey value for the duration of the AI technology window.

The only thing better than following a playbook is writing one before anyone else. This space is still developing. It could be you.