Back to blog
A source of truth is the written, shared answer a company gives to two questions before it lets AI work for it. Who are we, what do we stand for and what is our uniqueness? And where does our information actually live? Both answers are needed for any AI strategy, and they come from a framework most leadership teams already know. Where to Play, How to Win asks which spaces are worth competing in and why the company would win there. For AI, where to play becomes where AI can work at all, which depends on where the company's records are. How to win becomes what AI has to carry once it gets there.
Where to play: where the records are
AI can only work where information is shared. A flow whose data sits in a system, a process or a written policy is a place AI can read, structure and act on. A flow whose data sits in a private spreadsheet, an inbox or one person's head is not, however good the tool. So the first map an AI strategy needs is not a map of use cases. It is a map of where the company's records live and who can open them.
In a recent case for us, the sales pipeline lived in a private spreadsheet, and the administrative routines lived with the people who ran them. Both had to move into a shared place before anything else, because the owner could leave and take the knowledge with them. The same mapping showed that six of the seven value-chain flows leaked in some way. The retention flow, for instance, was not orchestrated, and product development ran on signals and on consultants working ad hoc. Those leaks are the real where-to-play map. They show where AI can help, and where there is nothing yet for it to read.
AI can only work where the company's information is shared. Everything else is a gap, however good the tool.
How to win: the identity AI has to carry
The second half is the company's identity and value proposition. It tells AI what to protect and amplify. We leave the detail of it to Using AI Without Diluting Your Brand, which shows what happens when this half is missing and offers a simple test for spotting it. What matters here is that it is written down and kept next to the records, because the two halves only work together.
Writing the first version
Here the first version has two parts. The first is the map of records: a list of the company's flows, where the information for each lives, and who owns it, including every flow that sits with a single person today. The second is one decision about where the whole source of truth is stored, so that every tool reads from the same place and nobody has to ask which version is current.
The identity half comes from the work described in the article above. The records half needs no new technology, only a few hours of questions put to the people who run each flow. If the answers disagree, or nobody is sure who owns a flow, that is the finding, and it is worth having before any pilot starts.
Pick the flow where your team most often asks "where is the latest version?" Find out where its information lives, who owns it and whether anyone else can open it. That one answer is the first line of your list of records.
Key takeaways
A source of truth answers two questions in writing: who the company is and what it stands for, and where its information lives.
Where to play for AI is decided by where the records are. AI can only work where information is shared.
A map of where records live, including those held by a single person, comes before a map of use cases.
How to win for AI is decided by identity and the value proposition, including the list of what a machine must not write.
A first version has two parts: a map of records with an owner for each, and one decision about where the source of truth is stored. The identity half is written separately.
FAQ
What is a source of truth for AI? It is a written, shared account of who the company is and where its information lives. It sets what AI must protect and where it can work.
How does where to play, how to win apply to AI? Where to play becomes the question of where AI can work at all, decided by where records are shared and where value leaks. How to win becomes the question of what AI must carry: identity, value proposition and tone.
Why map the records before choosing use cases? Because AI can only work with information that is shared. A use case built on data that sits in one person's head or inbox will fail however good the tool is.
What should be in a first version? A map of where each flow's records live and who owns them, including those held by one person, plus one decision about where the source of truth is stored. The identity and value proposition half is written alongside it.
