I've spent the last 8 months building a tool that helps founders keep their docs up to date and I though the hardest part would be AI.

    And I was wrong.

    For context, what we’re building does two main things:

    • It watches product changes (code, releases, UI flows) and drafts doc updates
    • It looks at support conversations and finds missing or outdated articles

    Simple idea on paper. Much harder in practice.

    Here are two things I got wrong early on.

    Mistake #1: Thinking “automation” was the main value

    At first, I was obsessed with making everything automatic.

    Auto-write.
    Auto-update.
    Auto-publish.

    In my head, less human involvement = better product.

    That was wrong.

    Early users didn’t care if it was automatic. They cared if it was correct.

    If one article was wrong, it didn’t matter that ten others were updated perfectly. Trust was gone.

    We learned quickly that:

    Suggestions > blind automation.

    Review flows > full autonomy.

    Accuracy beats speed every time.

    Now most of the work goes into making sure updates are easy to review, not just fast to generate.

    Mistake #2: Underestimating how messy real teams are

    I assumed most teams had:

    • clear release notes
    • clean changelogs
    • organized support inboxes
    • structured internal docs

    Almost none of them do.

    Reality looks more like:

    • features shipped without notes
    • changes mentioned in Slack
    • half-written tickets
    • outdated Notion pages

    So our early versions kept breaking because we were building for “ideal teams,” not real ones.

    We had to redesign everything to work with imperfect inputs.

    That slowed us down a lot, but made the product usable.

    TL;DR for anyone building ops / AI tools

    If you’re building something that touches core workflows:

    1. Don’t optimize for automation first Optimize for trust.
    2. Assume user data is messy Design for it.
    3. Solve the maintenance problem Not just the creation problem.

    Still early so yeah still learning and changing things constantly.

    But these two mistakes probably saved me months once I understood them.

    Hope this helps someone else avoid the same traps.

    Two things I got wrong while building my SaaS
    byu/aswin_kp inEntrepreneur



    Posted by aswin_kp

    Leave A Reply
    Share via