How it works

Start from the data you have. It proposes, you decide.

Working with us

  1. 1

    Survey the data

    Read-only. Nothing in your systems changes.

  2. 2

    Build on real data

    The first prototype runs on your own data.

  3. 3

    Go live

    It proposes, people decide, every question is logged.

  4. 4

    Measure, then grow

    See results first, then add more.

Ten checks before it acts

Every check is code. It runs before the AI is called, so every decision can be traced.

  1. 1

    Filter noise

    Drop what has no content

  2. 2

    Find the thread

    Know the conversation

  3. 3

    Identity

    More certainty, wider scope

  4. 4

    Clarity

    Unclear? Ask back

    May ask
  5. 5

    Idle work

    Stalled? Hand to a person

    May ask
  6. 6

    Mandate

    Not allowed, not done

  7. 7

    Resources

    Can it really be done?

  8. 8

    Blast radius

    Irreversible? A person decides

    May ask
  9. 9

    Confirm

    Wait for the real result

  10. 10

    Act

    AI, code, or a person

All clear · it acts
Any check fails · it asks

What never changes

Unclear? It asks

No guessing

Irreversible? You decide

Always through a person

Scoped access

Each group sees its own

Everything logged

Nothing rewritten later

The full principles

What should your AI take on?

Talk to us