Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
.png)
There is a well-established framework in technology strategy for understanding where profits accumulate in a value chain. Profits flow away from layers that are modular and commoditized, and toward layers that are integrated and differentiated. The canonical example is Apple: its hardware is tightly integrated with its software. This integration is why Apple captures nearly all of the profit in the PC and smartphone industries despite being a minority player by unit volume.
The question that has dominated AI industry analysis since 2023 has been: where in the AI value chain will this integration premium emerge? For a long time, the answer appeared to be “nowhere near the model companies.” If models are commodities, then OpenAI and Anthropic are building the equivalent of commodity chip fabs. Important, capital-intensive, structurally low-margin businesses squeezed between hyperscaler suppliers on one side and application-layer distributors on the other.
The agentic paradigm upends this analysis. If the harness must be integrated with the model to deliver a compelling agentic product, then the model is not a standalone product that can be evaluated and swapped independently. The relevant unit of competition is the model-harness combination. The companies building both, Anthropic with Claude Code, OpenAI with Codex, are the ones creating the integration that commands pricing power.
The clearest evidence came from Microsoft. For over a year, Microsoft’s stated AI strategy was to be model-agnostic. The company talked about its Core AI infrastructure as a platform that could accommodate any model, abstractable and interchangeable, with Microsoft’s value residing in its enterprise distribution, data layer, and security infrastructure. This was the commodity-model thesis operationalized as corporate strategy.
Then Microsoft launched Copilot Cowork and bundled it into a new E7 enterprise tier at $99 per user per month, roughly 65% above the previous top-tier E5 offering. Copilot Cowork was built on Anthropic’s Claude technology. It could not be model-agnostic. An agentic product that needs to orchestrate multi-step workflows, manage state, verify results, coordinate multiple model invocations cannot treat the model as a swappable component without significant performance degradation. Microsoft, in practice, abandoned model agnosticism the moment it needed to ship a product that enterprises would pay a premium for. The harness and the model had to be integrated. The commodity thesis broke on contact with the agentic reality.
For companies building agentic AI products and trying to determine how to price them, this is the central insight. The pricing power does not come from the model alone, and it does not come from the orchestration layer alone. It comes from the integration between them. From the compounding advantages that accrue over time as the harness is tuned to the model’s evolving capabilities and the model is trained with awareness of how the harness will use it.
Join companies like Zoom, DocuSign, and Twilio using our systematic pricing approach to increase revenue by 12-40% year-over-year.