A year ago, a consulting or tech services firm could point to the AI it used and call it an advantage. That argument no longer holds. Open-weight models now match much of what the leading proprietary models produce, at a fraction of the cost, and the performance gap many firms built their story on is closing faster than most expected.
For founders, this is less a technology question than a value question. If your proposition is "we use AI to do X faster", and every competitor can say the same with an open model on commodity compute, you do not have a proposition. You have a delivery method. That distinction becomes very real the moment a buyer asks you to describe your intellectual property in diligence.
Using a model is not owning IP
Most firms blur two different things: access to AI capability and ownership of AI-enabled IP. Access is a subscription. Ownership is something a buyer can value independently of you, your team and your current vendors.
An enterprise licence and a prompt library is not an asset. A domain methodology encoded into a structured, repeatable process that produces a differentiated outcome, whichever model sits underneath it, is.
The test is simple. If your AI provider changed its pricing, changed its API or closed your account tomorrow, what would your clients actually lose? If the honest answer is "not much, we'd switch", your clients already know that. A buyer will work it out quickly.
Separate your IP from the model
This is the point most AI narratives miss. For AI-enabled IP to add value to the business, it has to sit in a layer you own, distinct from the model and contributing more than the model does on its own. If you cannot draw a clear line between what the model does and what your firm adds, a buyer will assume the model is doing most of the work.
In a consulting or tech services firm, that layer lives in three places, and none of them is the model:
Methodology. The structured sequence of questions, decisions and interventions your firm applies to a client problem. A named, documented approach that produces consistent outcomes is defensible. Brilliant improvisation is talent, not IP, and it leaves with whoever leaves.
Context and data. Proprietary data sets, client benchmarks and anonymised outcome libraries built across engagements. If you have run 80 implementations and can tell a new client what failure looks like at step three, that accumulated pattern recognition has real value, and no model provider has it.
The layer around the model. The configuration, workflow integration, validation and quality control your firm has built and tested in real client environments. Crucially, it should be model-agnostic. If you can swap the underlying model and the output still holds, the value is yours. If swapping the model breaks the proposition, it never was.
What diligence will expose
Buyers now expect AI to feature in the value narrative. Every seller mentions it, so sophisticated trade buyers and PE investors have learned to probe past the slide, and where a meaningful share of value rests on AI, some bring in specialist AI diligence providers.
The questions are precise. Who owns this capability if your senior technical people leave? Can you show a client outcome produced by this specific approach, rather than by strong individual consultants? Is the methodology documented, trained into the team and reproducible? Is the tooling proprietary in any meaningful sense, or could a competitor license it next week?
Where the story rests on access rather than methodology, this is where it tends to unravel. Not because the technology is weak, but because the capability cannot be separated from the people using it or the tool they use. An IP conversation becomes a key-person dependency conversation. It is not scalable or reproducible, and that shows up in the valuation.
Build the asset before you need it
The firms that hold a defensible position will not necessarily be the ones with the most sophisticated AI. They will be the ones that have codified what they know into something that survives a change of people and a change of model.
That means treating IP as a tangible, commercial asset. Document the methodology and train it into the team. Turn client outcome data into a proprietary benchmark. Map the repeatable workflow and be explicit about where AI accelerates it and where human judgement makes the critical call. Make the AI a component of the product, not the product itself.
Then connect it to the numbers. IP a buyer will pay for shows up in revenue mix, margin and positioning, not only in a methodology document. And test one thing honestly: can you explain your proposition without naming a tool? If you cannot, fix that before a buyer finds it.
"Access to AI tools is not IP. The methodology, context and structured output your firm builds around those tools, and owns independently of them, is what a buyer will pay for."
The open-model trap
A firm builds its differentiation on the gap between frontier models and everything else. The gap closes, the story goes with it, and what is left is a delivery capability any well-funded competitor can replicate for the cost of a cloud account.
The way out is not better technology. It is sharper methodology, richer proprietary context, and a proposition a buyer can value separately from whichever model happens to be running it this year.
If you are not certain where your firm's real IP sits, the Equity Blueprint diagnostic is a good place to start. The Value Proposition pillar will show you quickly whether what you describe to clients and buyers is an asset or an access agreement. Take the diagnostic at www.viverogroup.com/diagnostic. It takes three minutes, and you will know which problem you are actually solving.