The Difference Between Using AI and Owning an AI System
Most firms use AI. A smaller number own one. The gap between those two things is not about the tools - it is about where the expertise lives.
There are three stages to how professional services firms relate to AI right now. Most are in the first. A meaningful number are moving into the second. A smaller group - firms that have made a deliberate choice and followed through on it - are in the third. The stages are not just different points on the same line. They produce different outcomes, compound differently over time, and require different things from the firm to move between them.
The distinction that matters most is not between using AI and not using it. That distinction is increasingly moot. The one that matters is between using AI as a personal productivity tool and owning something the firm built for itself.
Stage one: personal productivity
This is where most firms are. People on the team have AI subscriptions. They use them to write faster, research more, summarize documents, draft emails. The results are real - time gets saved, some work gets better. But the knowledge lives with the individuals. When someone finds a way to use AI well on a project, that discovery does not automatically become available to anyone else at the firm.
There is nothing wrong with this. It is how most technologies get adopted - individuals first, then organizations. But there is a ceiling to what firm-level AI use that stays at the individual level can deliver. You are essentially running a collection of independent AI experiments, and the results stay private.
The marker of this stage: if the best AI user at the firm left tomorrow, almost none of their AI capability would stay. It is in their browser history, their personal project library, their own intuitions about how to prompt.
Stage two: firm-level infrastructure
Infrastructure is what happens when the firm starts treating AI capability as an organizational asset instead of a personal one. Good prompts become shared Skills that anyone can run. Workflows get documented with enough specificity that a new hire can execute them on day two. The firm's own files, data, and institutional knowledge get connected to the AI tools in a systematic way.
This stage changes what the firm can promise its clients and what it can expect from new hires. Quality becomes more consistent because it is tied to the firm's standards, not to whoever happened to work on a given project. The shift from using AI to building AI infrastructure is an organizational shift more than a technical one. It requires the firm to make decisions about what its standards are, write them down, and build them into the tools.
Most of what firms at the 202 level are building belongs in this stage - Skills libraries, data connections, documented workflows, shared knowledge bases.
Stage three: owning something proprietary
Infrastructure is still built on shared tools - the same AI platforms, the same interfaces, the same underlying models that any other firm can access. Stage three is different. It is when a firm builds something that did not exist before and that other firms cannot simply replicate by buying the same subscription.
Proprietary AI systems encode the firm's specific expertise, judgment, and institutional knowledge in ways that run automatically. They do not just make individual work faster - they take work that previously required the firm's best people and let it happen without them in the loop for every instance. The knowledge is in the system, not the people.
Infrastructure
- Built on shared platforms anyone can access
- Makes existing work faster and more consistent
- Requires human direction at each task
- Can be replicated by a firm willing to invest the time
Proprietary system
- Built for a specific problem this firm has
- Takes work that needed senior people and runs it automatically
- Runs without human direction on every instance
- Encodes institutional knowledge that took years to develop
This is not a small distinction. A firm with a well-built Skills library is meaningfully ahead of a firm running AI ad hoc. But a firm that has built a proprietary system that encodes its best judgment and runs reliably - that firm has built something that does not come with any subscription.
What moves a firm to stage three
The firms that get there share a few things in common. They have a clear picture of which parts of their work are genuinely repetitive - not tedious, but structurally similar enough that they could be systematized. They have the infrastructure in place (the data connections, the documented workflows, the tested Skills) to build something on top of. And they have identified a specific problem worth solving, not a general interest in AI development.
The clearest signal that a firm is ready is when they can articulate the thing they want to build in a single sentence. Not "we want to use AI more effectively" - that is stage one or two thinking. Something like: "We want a system that takes a new engagement brief and automatically produces the first draft of our standard market sizing analysis." Specific. Bounded. Something that currently takes a real person meaningful time.
Stage three is not for every firm. Many firms will get significant value from stage two and stay there indefinitely - which is a reasonable place to be. But for firms where a particular workflow is high-volume, high-stakes, and runs on expertise the firm has spent years developing, the proprietary path is worth understanding.
The work we do at 303 is stage three. If you have a specific workflow in mind - one that currently requires your best people and that you would like to run without them in the loop every time - that is the conversation worth having.