DEFINITION
What is an AI platform?
The term everyone uses and almost nobody explains.
Almost every organisation now uses AI. Almost none gets it past a trial. The difference is not the model but what sits underneath it.
An AI platform is a coherent set of AI agents on a shared knowledge layer, with connections to existing systems, oversight of what happens and someone responsible for keeping it running. The difference from standalone AI tools is not the model but the foundation: on a platform every agent works from the same facts and every action stays traceable. W69 AI Consultancy in Amsterdam designs and builds those platforms.
McKinsey, fieldwork November 2025: 88 per cent of organisations use AI, but in no single function had more than roughly ten per cent scaled it. PwC asked 4,500 chief executives in January 2026: 56 per cent said AI had yet to deliver revenue or savings.
Why it stalls for almost everyone
Almost every organisation now has AI tools. Almost none has a foundation for those tools to stand on. That difference explains why the pilot succeeds and the roll-out does not.
The shared knowledge layer
The place that records what is true inside your organisation: agreements, decisions, rates and exceptions. Without that layer every tool works from its own version of the truth, and nobody notices when those versions start contradicting each other.
The agents
Specialised agents that each have a clear task and hand work to one another, rather than five separate assistants that know nothing about each other. Every agent has its own scope, its own permissions and a place where its work ends up.
The connections, via MCP
Agents need to reach the systems you already run: your finance system, your CRM, your documents. That happens through the Model Context Protocol, an open standard, rather than a bespoke connector per system that you have to maintain yourself.
Standalone tools versus a platform
On the left, five tools that know nothing of each other. On the right, the same work on a shared foundation, with oversight around it.
What an AI platform is made of
You do not build a platform in one go. You start with the question of where things go wrong, then lay the foundation, and only put agents on top once that foundation holds.
AI Navigator™
First we establish where your organisation stands and where things genuinely go wrong. Often that is not the model but the knowledge beneath it. The outcome is a list of places where a platform makes a difference, ordered by return.
Set up the knowledge layer
The shared layer every agent works from: what has been agreed, decided and learned, with a date and a source attached. This is the step that takes the most work and the one that separates a pilot from a platform.
Build and connect the agents
Only now do the agents arrive, and not all at once. Two or three that genuinely take work off people’s hands, connected to your systems, with a written record of what they may do, where a human must step in and what gets logged.
Run it and keep developing it
Models get replaced, connections break and rules move on. Without ongoing management a platform falls back into a set of standalone tools within a year. That is why management is part of the platform rather than an extra line on the invoice.
What people ask about this
Standalone tools know nothing of each other. They share no knowledge, no oversight and no maintenance, and each holds its own version of the truth. A platform puts a shared layer underneath: the same facts, the same oversight, the same upkeep. That is precisely why standalone tools stall at the pilot stage and a platform scales.
Because the model has no reliable context. It does not know what has been agreed in your organisation, what was decided earlier or where the right data sits. McKinsey research from November 2025 found that 88 per cent of organisations use AI, but in no function had more than roughly ten per cent scaled it. PwC asked 4,500 executives in January 2026: 56 per cent said AI had yet to deliver revenue or savings.
Not necessarily. The heavy costs sit in connecting to older systems with no open interface, and in audit work for regulated sectors. The knowledge layer itself and the first agents can be set up in days rather than months. Start small with two or three agents on a working knowledge layer, and only scale once it demonstrably works.
Since 2 August 2026 the transparency obligations in Article 50 apply: people must know they are dealing with AI, and AI-generated content must be marked in a machine-readable way. That touches almost every organisation, including those with no high-risk use case. A platform makes this easier than standalone tools, because you configure it once rather than per tool.
That has to be settled up front. Models get replaced, connections break, regulation moves and your own processes change. Without ongoing management a platform falls apart into standalone tools within a year. At W69, management is part of the platform rather than an additional charge.
Want to know where your organisation stands?
The AI Navigator™ maps where you are today, including the point where it usually comes unstuck: the foundation beneath your AI.