The Business Is Its Knowledge

The world and business is talking about AI.  How and why we should all be using it.

We can explain the why, simplify the how, and execute the results. You can’t just pretend this will go away, it will change everything.

The AI is the engine. Your knowledge is the fuel. Without the fuel, the engine runs on nothing.

The Shift

Here I take a look at why the AI revolution rewards those who productise what they know.

AI is eliminating routine execution work. Summarisation, basic analysis, repetitive operations, and workflow execution are increasingly handled by tools and agents. This is not a future prediction: it is happening now, and accelerating.

This is not a threat. It is a liberation. But ONLY if you act.

For too long, talented people have spent their days working as machines: processing, summarising, reformatting, chasing, checking. As we automate away every area where a human is "working as a machine", we free up human capacity to be truly human. To be curious. To be creative. To build relationships. To innovate. The organisations that thrive will be those that redirect this freed capacity toward the work that only humans can do.

The natural instinct is to focus on the interface: which AI assistant to use, which chat tool to deploy, which agent framework to adopt. This is a mistake. Chat interfaces are commoditising rapidly. OpenAI, Anthropic, and Google are in an arms’ race that no single business can outpace. Betting on a proprietary interface is betting against the largest technology companies in the world.

The winning position is different: build the knowledge layer that every tool depends on.

My Thesis

Businesses compete through proprietary knowledge: what they know, how they do it, and why it works. This knowledge takes many forms. There is structured data: the numbers in your financial systems, the records in your CRM, the metrics in your reports, the facts that drive decisions. And there is unstructured knowledge: the documents, emails, meeting transcripts, project histories, contracts, and process descriptions that capture how the business actually operates.

Both types matter. A question like "what was our margin on similar projects?" requires structured financial data. A question like "what did we learn from that project?" requires unstructured knowledge. Most business questions require both.

The opportunity is to make all of this knowledge ingestible (captured from the messy real world of systems, files, and conversations), trustworthy (quality-controlled, versioned, governed), and accessible (available to any tool the business chooses to use).

We call this the "Business Brain": a knowledge operating system that captures institutional knowledge, makes it queryable, and exposes it to the tools and agents that execute work. The Brain is not another chat interface. It is the substrate that makes any interface useful.

The Limiting Factor

Large language models and AI agents will continue to improve at an exponential rate. Each generation will be smarter, faster, and more capable than the last. This is not in doubt and a “generation” is measured in months now, not years.

But here is the critical insight: no matter how powerful these tools become, they will always be limited by your proprietary knowledge and data. The most advanced model in the world cannot answer questions about your contracts, your project history, your client relationships, or your internal processes unless that information has been captured, structured, and made available.

What This Means in Practice

Operationally, the Brain becomes the place where truth lives. Contract terms are centralised so delivery discussions can be checked automatically. Meeting outputs become reliable tasks. Past projects inform pricing, estimating, proposals, and risk assessment. When someone asks a question, they get a correct, grounded answer with sources - and a simple mechanism to correct the system when it is wrong.

The hardest problem is not technology. Organisations often have many years of data: multiple versions of the same document, outdated processes, contradictory truths. They cannot simply ingest everything and hope for the best - that produces confident answers that are partially wrong, which is worse than ignorance. The Brain must distinguish between what happened (immutable evidence) and what we currently believe to be true (curated knowledge that can change over time with traceability).

The Conversation to Have Now

AI is not replacing knowledge work. It is making knowledge work executable. The organisations that will thrive are those that treat their institutional knowledge as a strategic asset - captured, governed, and made accessible to whatever tools prove most useful.

The interface layer will continue to evolve. Models will improve exponentially. New tools will emerge. But they will all hit the same ceiling: the quality and accessibility of your proprietary knowledge. That ceiling is yours to raise.

The question is not whether to do this work, but when and how. Every month of delay is a month where your knowledge remains locked in documents, systems, and heads - inaccessible to the tools that could put it to work.

I would welcome the opportunity to explore what this looks like for your organisation. The first step is a conversation: understanding where your knowledge lives today, what questions you need it to answer, and what a practical path to implementation looks like. No two organisations are the same, and the approach must be tailored to your context.

If this resonates, let us start that conversation - human to human.

Let’s get a bit more technical..

Market Validation

This thesis is no longer speculative. Three developments have validated the core insight:

First, analyst consensus has shifted. Gartner now advises enterprises to prioritise knowledge infrastructure over model selection. The competitive moat is governance, retrieval architecture, and data curation - not the AI model itself. Models are commoditising; knowledge is not.

Second, the protocol wars are over. In late 2024, Anthropic released the Model Context Protocol (MCP) as an open standard for connecting AI tools to external data. By mid-2025, OpenAI, Perplexity, and every major automation platform had adopted it. MCP is now the universal language for AI-to-data connectivity. This means a single knowledge layer can serve every AI assistant - Claude, ChatGPT, Perplexity, workflow tools - without custom integration for each.

Third, the ROI data is in. Organisations implementing knowledge infrastructure report average returns of 3x for every pound invested. The payoff is real and measurable.

Strategic Optionality

A common misconception is that businesses are rapidly switching between AI assistants. The data shows otherwise: most users stick with their primary tool. What enterprises actually want is optionality without lock-in - the ability to adopt new tools as they emerge, without re-platforming the knowledge layer.

The Business Brain provides exactly this. Because MCP is now universal, the same knowledge infrastructure serves any tool the business chooses. Teams can work in Claude today, experiment with ChatGPT tomorrow, and automate with workflow tools next week - all querying the same governed knowledge base. The interface is flexible; the knowledge is stable.