Wrote a free vendor-neutral guide to building a first ontology — what did I get wrong?

Hi All

I am an independent author (not affiliated with Palantir).

I wrote a free guide, “Ontology Is All You Need: A Practical Implementation Guide for Midsize Companies,”. Link to PDF https://www.applayerai.com/ontology

Goal is to teach 50–1,000-person companies to create their first ontology.

It’s vendor-neutral by design. Many companies will start in tools they already own and graduate to Foundry later.

Before it goes wide, I’d genuinely value this community’s judgment and feedback.

Scope: I cap version one at 15–20 object types. Where do real first implementations actually land? is that ceiling right?

Sequencing: The guide builds the semantic layer first (weeks 3–8), then adds kinetics: two or three governed actions for a single workflow (weeks 8–11). Many of you work use-case-backwards instead. When does each approach win?

Timeline: Is 90 days honest for a midsize company with one half-time data lead and part-time domain experts, ending in one piloted agent behind approval gates?

Pitfalls: I list eight (god object, kitchen sink, database-shaped objects, action sprawl, and so on, credited to Palantir’s talks). What’s the #1 first-timer mistake you see that I’ve missed?

Accuracy: Anything in how I’ve represented the semantic/kinetic split or the ontology concept generally that would make a practitioner wince?

I’ll incorporate what I learn in this thread.

Thank you!

Not the feedback you asked for, but my AI slop detector went off immediately. The AI sentence structure and the barrage of AI tropes makes this a very hard read.

I would start by rewriting it in your own language. AI can be a great help, but think about this from the reader’s perspective: Why should I read this, when I can just get Claude to output the same quality of text, tailored to my specific situation?

Thank you for feedback. This is very helpful. I will definitely re-write it.