5

Where does the data live?

Module 3 · Structure. After this lesson you can: identify when the same fact lives in multiple places, choose a single authoritative home for it, and explain why this is what makes AI delegation reliable.

Consider a team whose contact records for a key client exist in three places: a phone's address book, a shared spreadsheet, and a chat thread where actual conversations happen. All three hold different values. One has an old phone number. One has a misspelled email address. One is current. No one on the team knows with certainty which version is correct.

The problem sounds like a minor inconvenience. The actual cost is visible in what happens around it. Every time the team's AI assistant needs to reach that contact, it has to choose which source to trust, and there is no principled basis for that choice. Every project involving the contact starts with informal reconciliation: which email is live, which number is active. Any system built on top inherits the confusion, the way a building inherits a crooked foundation. And the problem does not stay contained to contacts. Deal values recorded in one tool, deal status updated in another. Tasks tracked in two applications that show different states. Dates in messages, in calendars, in people's heads.

The instinct is to solve this with effort: update everything everywhere, be more careful, enforce a convention. That is treating a structural problem as a behavioral one, and the behavioral fix does not hold. The structural fix is simpler and more durable: every fact gets exactly one home. One location where the authoritative version lives. Every other system reads from that home and writes back to it. Copies stop being rival versions and become what they should always have been: views onto a single record.

When teams apply this principle to a genuinely messy area, an entire category of work tends to disappear. Not decrease. Disappear. The reconciliation step at the start of each project goes away, because nothing has drifted from anything else. The disputes about whose version is correct go away, because there is only one version. The assistant stops having to choose between sources, because there is only one source to point it at.

This is what the phrase single source of truth means in practice. It is not a rule about tidiness. It is a structural decision: choose where each fact lives, and treat everything else as a derived view. A view can be read. It should not be written to directly. When a copy is written to as though it were an original, it becomes a rival original, and the problem returns.

This lesson belongs in a program about AI for a specific reason. An assistant is only as reliable as the ground it stands on. Point one at three contradictory records and you get a confident amplifier of contradictions: outputs that are precise in form and unreliable in substance. Give it one source of truth and its precision becomes trustworthy. Structure is not housekeeping. It is what makes delegation safe at all.

Stop here. Actually sit with this before you scroll on.

If the same fact lives in two places, which one is true?

Try this nowUnder 30 minutes
  1. Pick one mess you actually live with. Contacts, passwords, family dates, project notes, customer info. Pick the one that has burned you most recently.
  2. Map it with your AI. Tell it: "Here is a fact I manage. Help me list every place a version of it lives." You will find more places than you expect.
  3. Choose the one home. Not the fanciest tool. The one you will actually keep open.
  4. Merge with the machine's help. Have it reconcile the versions, flag the conflicts, and produce the single clean record for the home you chose.
  5. Demote everything else to a view. Delete it, or mark it clearly as "copy, do not edit." Notice, over the next week, how a small background hum of anxiety goes quiet.

Two questions before you go

Answer, then say whether you were sure or guessing. Being honest about which is the skill being trained.