There was a time when making a phone call required strategy. You knew when you were calling long distance. You knew when you were calling internationally. You watched the clock because the economic unit was time. Telecom companies monetized minutes and distance. Voice was the product, and the meter was always running.
Then something shifted. Voice became just another application running on a data network. The real infrastructure was no longer about conversations — it was about packets. Broadband replaced per-minute billing. Mobile data plans replaced per-call charges. Eventually, we stopped thinking about calls at all. We paid for access to a data network. The pricing model changed because the architecture changed.
I believe we are approaching a similar shift in engineering software because of AI.
For decades, the economic unit in CAD, PDM, and PLM was the license. Then SaaS turned that into subscription seats. Named users. Concurrent users. Feature tiers. The entire stack, from design tools to product lifecycle platforms, converged around one assumption: value scales with the number of people who can access the system – CAD seats, PDM seats, PLM seats, approvers seats, reviewer seats, read only seats.
For a long time, seats made sense. Engineering work was human-centered. More engineers meant more licenses. More complexity meant more seats.
But what happens when productivity scales not with headcount, but with intelligence introduced by AI tools, and companies are considering hiring ‘AI agents’ to do some work. ? That is the question AI forces us to confront.
SaaS Solved Access — But Access Is Not the Same as Value
Let me be clear: SaaS was a breakthrough. It solved real problems.
Perpetual licenses were rigid and expensive. Upgrades were painful. Deployments were fragmented. SaaS aligned incentives, reduced upfront costs, simplified infrastructure, and accelerated adoption. It democratized engineering software.
Monica Schnitger recently wrote that SaaS is not going away in the age of AI. I agree. Subscription models corrected structural inefficiencies in enterprise software economics. But SaaS monetizes access – model geometry and drawings for CAD, versions in PDM, ECO process and governance in PLM.
Seats became the proxy for value. More engineers meant more seats. More projects meant more seats. Digital transformation often translated into expanding access.
For a long time, this worked well enough. But seats measure participation, not intelligence. They measure logins, not leverage. They measure how many humans touch the system, not how much structured knowledge the system contains. That distinction is becoming harder to ignore.
AI Changes the Center of Gravity
AI is not simply a productivity and intelligence layer on top of existing tools. It is beginning to reorganize where value sits.
When Satya Nadella talks about new economic models emerging in the AI era, he is not suggesting subscriptions disappear. He is hinting that the unit of value may evolve toward what fuels intelligence.
In engineering, AI is already crossing boundaries – it can generate geometry based on constraints. It can propose design alternatives. It can analyze change propagation across assemblies. It can extract structure from unstructured documentation. It can reason across cost, supplier, and configuration data. These capabilities do not depend primarily on seat access. They depend on structured, connected product knowledge.
If generative design operates on constraints, those constraints must exist in a structured form.
If automated change impact analysis is meaningful, dependencies must be explicitly modeled.
If AI is to suggest cost-optimized alternatives, supplier and cost relationships must be connected.
The leverage moves upward. CAD becomes less about drawing and more about executing geometry derived from structured intent. PDM becomes less about files and more about preserving history and protecting access. PLM becomes less about routing forms and more about governing product knowledge, change intent, validating product data before the release.
The intelligence layer sits above all of them.
When one engineer augmented by AI can accomplish what previously required several, seat count stops being a reliable indicator of value creation. The center of gravity shifts from access to intelligence.
Geometry Is Not Enough
For decades, we treated geometry as the embodiment of product knowledge. The CAD file was the artifact. PDM preserved it. PLM orchestrated its lifecycle. But geometry alone does not provide context.
AI does not reason effectively on isolated files. It reasons on relationships. It needs to understand how parts connect, how revisions evolve, how alternates relate, how lifecycle states change, how costs roll up, how suppliers impact availability. In other words, it needs a product graph.
Over the past few years, I have written about Product Memory and Context Graphs. At the time, these ideas were largely about digital thread, traceability, and architectural modernization. Today, they are becoming the substrate of intelligence. Intelligence scales with structure.
A simple archive of CAD files does not produce insight. A well-structured network of product relationships does. The difference is not incremental and it is architectural.
As organizations mature, their most strategic asset is no longer the number of tools they deploy. It is the depth and coherence of their structured product knowledge. That is what AI consumes and that is what enables automation. Eventually, that is what amplifies decision-making.
Here is a difficult question – if intelligence requires structured data, why is the dominant economic model still anchored to seats?
When Tools Become Thin Layers
This is where the discussion becomes interesting.
For decades, CAD systems were strategic fortresses. They encapsulated modeling expertise, proprietary kernels, parametric logic. PDM systems ensured file integrity. PLM systems enforced process rigor. The tools were the center. And AI begins to invert that relationship.
When intelligence is embedded in agents and contextual graphs that sit above the tools, those tools risk becoming execution layers rather than strategic anchors.
- CAD executes geometry.
- PDM synchronizes data.
- PLM orchestrates governance.
The intelligence layer reasons across all of them. This does not mean CAD disappears. It does not mean PLM becomes irrelevant. It means their relative position in the value hierarchy shifts.
If design intent is captured as structured constraints, geometry becomes downstream from intelligence rather than upstream. If product relationships are modeled explicitly, change management becomes computational rather than manual. If supplier intelligence is integrated, cost decisions become algorithmic rather than reactive.
In such a world, the strategic leverage resides in the structured product memory and not in the individual seat.
That is how commoditization happens. Not through feature stagnation, but through architectural displacement. When intelligence becomes the primary value engine, execution tools begin to look more like utilities. Utilities are rarely priced based on how many people touch them. They are priced based on the resource consumed. In engineering, that resource is increasingly structured product data.
Why Seat Economics Start to Crack
Consider a practical scenario – an engineering organization employs twenty designers and twenty engineers. In a traditional SaaS model, growth correlates with adding more seats. More people equals more revenue for the vendor.
Now introduce AI into the workflow. Generative tools reduce manual modeling effort. Automated impact analysis reduces engineering overhead. AI-assisted configuration eliminates repetitive tasks. The company may produce more with the same team. It may even reduce the need for additional hires while increasing product complexity.
- Revenue grows.
- Product variants grow.
- Data volume grows.
Seat count does not necessarily grow. In fact, seat-based economics may begin to penalize efficiency. The more the organization leverages automation and intelligence, the less seat growth aligns with value creation.
Meanwhile, something else grows rapidly – data, revision records, configuration logic, richness of decision process, the complexity of product graph. Value is accumulated in data, not in seats.
That misalignment between value creation and monetization is the crack in the seat based model.
From Seats to Data
When telecom companies realized that voice was no longer the primary value driver, they shifted to monetizing data. We may be approaching a similar moment in engineering software.
If intelligence depends on structured product knowledge, then perhaps the economic unit should reflect the scale and quality of that knowledge. “Pay per data” does not mean charging for every byte stored. It means aligning economics with structured product memory – with the nodes and relationships that enable intelligence.
For customers, this could mean paying based on the scale of structured product knowledge they manage.
For vendors, it means focusing on enabling data quality, connectivity, and reasoning capabilities rather than maximizing seat count.
This is not about extracting more revenue. It is about aligning incentives with where architectural value resides.
The Hard Questions
Every economic shift brings legitimate concerns, and this one is no different. If we move toward data-centric economics, we have to confront uncomfortable questions. How do we measure structured data in a way that reflects real value rather than raw volume? How do we prevent runaway costs as product graphs grow more complex? How do we ensure that customers retain ownership and portability of their product memory? And perhaps most importantly, how do we avoid turning intelligence itself into a new form of lock-in?
Data-based economics cannot become an opaque billing abstraction. If structured product knowledge becomes the foundation of intelligence, then transparency and governance become even more critical. Customers must remain stewards of their product memory. Exportability, interoperability, and architectural openness matter more, not less. This is not a superficial pricing adjustment. It represents a deeper architectural reorientation of the engineering software stack, and architectural shifts demand architectural responsibility.
Customers as the Real Protagonists
The hero of this transition is not the vendor inventing a clever pricing mechanism. It is the customer who recognizes that structured product knowledge is the strategic asset and a better way to measure the value.
Those companies will build intelligence-ready organizations. They will not abandon CAD, and they will not discard PLM. Instead, they will reposition them. CAD becomes the execution environment for geometry and constraints. PLM becomes governance and coordination. Product memory — the structured, connected graph of knowledge — becomes the intelligence core.
When customers begin demanding economic alignment with where value truly resides, the industry will respond. Vendors always follow architecture, and architecture follows value.
What is my conclusion?
Are we at the “Broadband Moment” of PLM software? License economics reflected the era of desktop tools, when software was installed locally and value was measured in ownership. Seat-based SaaS reflected the era of cloud access, when value was measured in participation and connectivity.
AI may usher in the era of intelligence economics.
If intelligence runs on structured product memory, and if product graphs become the substrate for automated reasoning and decision support, then the unit of value in engineering software will inevitably shift toward that foundation. Seats will not disappear overnight. CAD will not collapse into irrelevance. SaaS is not going away.
But the center of gravity is moving and economic models eventually follow.The vendors that understand this shift will not win by purchasing more seats. They will win by mastering data intelligence and optimizing workflows.
Just my thoughts…
Best, Oleg
Disclaimer: I’m the co-founder and CEO of OpenBOM, a collaborative digital thread platform that helps engineering and manufacturing teams work with connected, structured product data, increasingly augmented by AI-powered automation and insights.
