Ask your AI a question. Ask your competitor's AI the same question. The answers will be almost identical.
Same model. Same training data. Same capability. Same output. The only reason most firms have not noticed this yet is that most firms are not comparing. They are too busy being impressed that AI can draft an email or summarise a document to ask the harder question: if everyone can do this, where is the advantage?
The answer is uncomfortable. There is no advantage. Not yet. Not for most firms.
The commodity layer
Every firm in your sector can buy the same AI. Anthropic, OpenAI, Google — the models are available to everyone at the same price point. The capability is extraordinary. But extraordinary and exclusive are not the same thing. When everyone has access to extraordinary, extraordinary becomes ordinary. It becomes commodity.
This is not speculation. This is already happening. The firms that adopted AI early — 2023, early 2024 — believed they had a competitive edge. They did. But it was a timing advantage, not a knowledge advantage. Timing advantages are, by definition, finite. As adoption spreads, the edge disappears. The model does not get better for you because you adopted first. It gets available to everyone because the provider scaled.
Right now, every firm using AI with default knowledge — the general information the model was trained on — is producing the same outputs as every other firm using AI with default knowledge. The same client emails. The same contract summaries. The same HR responses. The same compliance drafts. If you put the outputs side by side, you could not tell which firm produced which.
That is not transformation. That is commodity with a better interface.
The three layers
AI competitive advantage exists on three layers. Most firms are stuck on layer one.
Layer one is the LLM itself. This is commodity. Everyone has it. It is infrastructure, like broadband — essential but not differentiating. You would not claim competitive advantage because your office has an internet connection. The LLM is the same.
Layer two is timing. The early adopters. The firms that moved first, figured out use cases, built internal capability before their competitors did. This advantage is real but temporary. It lasts exactly as long as it takes for competitors to adopt the same tools — which, given the pace of AI adoption, is months, not years. Layer two is already eroding for most sectors.
Layer three is the knowledge hub. This is the only layer that is sustainable. This is where your firm's specific knowledge — your processes, your standards, your client intelligence, your operational nuances, your brand voice, your domain expertise — is curated into a living system that AI draws from instead of the open web.
A firm operating at layer three does not get the same output as a competitor. The AI is not searching the internet for a generic answer. It is searching the firm's own knowledge for a specific one. The client email sounds like the firm. The compliance response reflects the firm's actual policies. The HR answer draws from the firm's own procedures, not a generic best-practice guide.
The experience is identical — the user types a question and gets an answer, just like ChatGPT or Claude. But the output is completely different because the source is different.
Why most firms never reach layer three
Layer three requires something most firms do not have: a mechanism to curate their knowledge centrally, attribute contributions, and make it available to AI in real time.
The knowledge exists. Every firm has it. It lives in the heads of the people who do the work — the operational staff who know which processes actually work, which client needs come up repeatedly, which edge cases the textbook does not cover. This is tacit knowledge. You only know where the bodies are buried if you dug the graves.
But tacit knowledge, by definition, is difficult to transfer. It lives in people, not systems. When someone leaves, it leaves with them. When a new person joins, they start from zero. Every meeting is an attempt to share it. Every handover is an attempt to capture it. Most of the time, it fails — not because people are unwilling, but because there has never been a mechanism that makes it easy.
The knowledge hub changes this. It is not a document library. It is not a wiki. It is not a shared drive. It is a living, curated, authored system where every person in the team can contribute the element they know. When someone discovers a nuance on Friday, every query on Monday includes it. The hub gets smarter every day. The gap between a firm with a knowledge hub and a firm without one widens with every interaction.
The invisible shift
When a firm operates with a knowledge hub, nothing changes for the end user. They open the AI. They type a question. They get an answer. The interface looks the same. The experience feels the same.
But the answer is completely different.
Instead of the AI searching the web for a generic response, it searches the firm's terms of business. Instead of a generic HR answer, it draws from the firm's actual HR policies. Instead of drafting from scratch, it draws from FAQs the firm has answered a hundred times. Instead of the same output as every competitor, it produces output that sounds like the firm and nobody else.
This is not optimisation. This is a structural change in what AI is drawing from — and it is the only change that produces a sustainable competitive advantage.
The compounding effect
Knowledge hubs compound. Every client interaction refines them. Every edge case resolved becomes a pattern. Every contribution from every team member makes every future query better. This is not a tool where everyone starts equal and stays equal. Firms that build their knowledge hub first develop an advantage that accelerates with use.
A competitor starting six months later does not just need to build the same hub. They need to build the same hub plus six months of refinement, edge cases, and operational learning that is already embedded in yours. The gap is not static. It grows.
The question every firm should ask
The question is not whether AI is useful. It is. The question is not whether to adopt AI. You should. The question is: when every firm in your sector has the same AI capability — and they will — what will make yours different?
If the answer is nothing, you are the commodity.
If the answer is your knowledge — curated, configured, and deployed through AI that draws from your firm instead of the web — then you have something nobody else can replicate.
The LLM is the same for everyone. What you put into it is not.