KodBees LabsContact

Our approach

How an idea becomes a product (or doesn't).

Every product passes the same four stages. Each one asks a single question, and a product only moves on when the answer comes from what people actually do. Most ideas stop before the end — that's the method working.

Four gates

The stages map onto the four cells of our logo. A project's mark fills as it advances.

  1. 01 — Observe

    Is the problem real, and does it keep coming back?

    What we do
    We sit with the people doing the work, map how it gets done today, and note where time and attention leak away.
    Evidence we look for
    The same friction, described independently by different people, with a cost they can name.
    Leaves this stage when
    We can state the problem in one sentence that the people who have it agree with.
    We stop if
    The pain is occasional, already well solved, or exists mostly in our heads.

  2. 02 — Prototype

    Can we make it meaningfully easier?

    What we do
    We build the smallest thing that tests the core idea — a script, a spreadsheet, a rough interface — and put it in front of the same people.
    Evidence we look for
    People finish the task faster or with less effort, and tell us what they would change.
    Leaves this stage when
    One workflow works end to end, and people ask to keep using it.
    We stop if
    The improvement is marginal, or the hard part is something software can't fix.

  3. 03 — Validate

    Do people choose it over what they use today?

    What we do
    We run the prototype inside real work for weeks, not minutes, and watch whether it sticks.
    Evidence we look for
    Repeat use without reminders, and a willingness to commit time or money.
    Leaves this stage when
    Usage holds without our help, and the economics can work.
    We stop if
    Usage fades once the novelty does.

  4. 04 — Ship

    Can it run as a product people rely on?

    What we do
    We harden it — reliability, security, onboarding, billing, support — and launch it under its own name.
    Evidence we look for
    It runs without the lab holding its hand.
    Leaves this stage when
    It graduates from the lab and gets a home of its own.
    We stop if
    We can't support it at the standard the people using it deserve.

What we hold ourselves to.

  1. P1

    Problems before products

    We don't start from a technology looking for a use. We start from work people repeat, and stay with it until we understand it.

  2. P2

    Evidence over enthusiasm

    Every stage ends with a question we answer with what people do, not what they say they would do.

  3. P3

    Small surface, deep job

    A product should do one job thoroughly. We cut scope until the core workflow fits in a sentence.

  4. P4

    AI where it changes the outcome

    Models go where they change the result — reading, reasoning, generating — and stay out where ordinary software is faster, cheaper, or more predictable.

  5. P5

    Work in the open

    We write down what we're testing and why we stop. A shelved project is a result, not a secret.

What we won't build

  • Features nobody asked for, added to make a demo look bigger.
  • Thin wrappers around a model with no real workflow underneath.
  • Products we can't support properly once people depend on them.
  • Anything that needs dark patterns to keep people using it.

See the method applied.

Every project in the lab shows its stage, its open questions, and its next gate.

See our products