Ondatrace

Product thesis

Ondatrace isn't another project management tool.

Teams already have tools to track their work. What's missing is a way to see across them.

The first signs of trouble are usually scattered. A Jira ticket goes overdue. A customer sends a worried email. Someone mentions a problem in a meeting. Ondatrace brings those signs together so someone can act on them.

Six bets we're testing

  1. Start from what happened

    A question typed into a chatbot is gone once it's answered. An update from a project tool records who made it and when, and which project it's about. Keeping those details lets Ondatrace handle updates the same way whatever tool they came from, and build up a history.

  2. Let clear rules decide, and use AI to explain

    Many health checks come down to simple facts. Did a date move? Is a blocker still open? Does anyone own this task?

    When a rule can decide, the rule decides. The AI explains the result and handles the cases rules can't.

  3. Remember what came before

    If you only see today's status, you can't tell whether a problem is new or has been getting worse for weeks. Ondatrace keeps the history, because the trend matters as much as where things stand today.

  4. End with a decision or next step

    Summaries help, but managers mostly need to know what needs a decision and who has to make it.

    Ondatrace keeps decisions, risks and dependencies as separate records so they don't get lost inside a status paragraph.

  5. Earn trust one action at a time

    How much Ondatrace should do on its own depends on what's at stake. From low stakes to high:

    1. Refreshing data from a tool
    2. Drafting an internal update
    3. Editing a task
    4. Moving a committed milestone
    5. Sending a message to a customer

    Ondatrace takes on one kind of action at a time and keeps a record of what it did. A person can always step in.

  6. Work out the structure before asking for it

    Real work is messy. The same project goes by different names in different tools. Dates conflict. Dependencies come up in meetings and never reach the tracker. Some important work has no project record at all.

    Ondatrace connects to your tools as they are and suggests which projects exist and how they link up. People confirm or correct its suggestions. You don't have to build a tidy setup first.

    How it should work

    1. Read your tools
    2. Suggest the structure
    3. Show the evidence
    4. You confirm or fix it
    5. Keep watching

    Not this

    1. Set everything up
    2. Keep it updated by hand
    3. Ask AI about it

    The test we use

    Could Ondatrace work this out on its own before asking someone to type it in?

    People's time should go to decisions and judgment calls rather than data entry.

What would prove us wrong

We'd count Ondatrace as a failure if the people using it find that:

  • Seeing across tools doesn't tell them much beyond what their tools already show.
  • Its suggestions are too noisy to be worth a manager's time.
  • The evidence is there but too hard to check.
  • Keeping its records in order takes more effort than it saves.
  • Its review list becomes another inbox nobody trusts.
  • It can't explain why it did something, or why it didn't.
  • Fixing and supervising it takes more time than it saves.

The pilot is set up to answer these questions.

Try the demo Read the design decisions