The vendor landscape - who to hire for what
"I need AI help" can mean wildly different things requiring different people: a prompt engineer (good prompting/evals, Cluster B/C), a no-code automation builder (n8n/Make, D2), a full-stack AI engineer (APIs, RAG, agents, Clusters C/E), or a fine-tuning/ML specialist (rare need, F1). Knowing which one you actually need avoids hiring mismatches - the single most common mistake non-technical founders make.
Most day-to-day AI feature work (like most of what you already build) sits in the "full-stack AI engineer" or "no-code automation builder" bands - API integration, prompting, some RAG, some agent orchestration - not deep ML research. Genuine ML/fine-tuning specialists are needed far less often than the market suggests, and are usually overkill (and overpriced) for typical product needs.
Advisory case: next time you're scoping a hire or vendor for Kasper or Realla.AI, write down (before talking to them) which of the four hats you actually need - then see whether their pitch matches that, or tries to upsell you into a fancier (unnecessary) category.
Why does hiring the wrong "AI person" happen so often?