Many AI outputs feel generic — not because the tool is weak, but because the prompt never gave it enough to work with.
This conversation is with Schuyler Dragoo, Chief Strategy Officer at Kiingo AI. Her interest in how humans, AI, and other species misunderstand each other started when ChatGPT first came out — she realized it could help her understand social situations. That curiosity led her to an unusual research practice: spending time in Boston parks studying geese to understand communication breakdowns up close.
👉 Watch the full Tech for Founder Podcast episode with Schuyler on YouTube:[Full episode →](embed the youtube video here)
Why this matters
When an AI response comes back vague, it's tempting to read it as a flaw in the tool. What surfaces in this conversation is that the model is often just reflecting what it was given — a generic prompt produces a generic answer, because the AI is matching patterns from training data.
The same logic showed up away from any computer. Understanding how geese communicate required observation first, before expecting any response to land. Treating a new AI tool the same way — watching before relying on it — surfaced the same need.
More detail isn't always the fix, either. Too much at once can overwhelm a model, and the same word can mean one thing to a founder and something else to the system reading it.
Key ideas from this conversation
Generic input produces generic output. AI pulls from training data to predict the kind of response a vague prompt usually expects.
The same word can mean two different things. Terms like "minimalist" or "futurist" carry different definitions for different people — and for different AI models.
Treating AI like a new intern changes the result. Breaking a task into smaller, sequential prompts produced better outcomes than asking for everything at once.
The core insight
"AI is remarkably easy to use, but remarkably hard to use well."
Ease of use isn't the same as getting what you actually meant. Most generic outputs trace back to that gap, not to a weak model.
A simple founder lens
It's easy to assume the model didn't try hard enough. The conversation points one step earlier — to what it was actually given to work with.
What would someone brand new to my business need to know before they could do this task well?
Something to sit with
Communication rarely fails because someone wasn't listening. It usually fails because someone assumed the other side already understood.
What are you assuming someone else already knows?
Tech for Founder Playbook
Weekly insights from real founder conversations — turned into clear, actionable ideas.
→ Explore more: https://techforfounder.com