8 Comments
User's avatar
Paras Doshi's avatar

love the 5 tips! my default now is basically if AI gives me slop, it's not AI's fault, it's me who is accountable and i need to iterate to make it useful for whatever i am doing.

-Paras

Sharique Nisar's avatar

High engagement metrics are often deceptive without the right product category for context. When data stays in one dimension, AI tends to default to generic patterns. True clarity comes from layering independent signals that force a more specific and accurate business narrative.

Kay Wilson's avatar

Great read. Every time I hear “AI hallucination,” my first instinct is to ask, “What context did the model actually have?” This is a common blind spot, especially as teams rush AI adoption. We’re optimizing for speed, but skipping the harder work of input curation. That tradeoff shows up directly in output quality.

Om Prakash Pant's avatar

I’ve seen this play out a lot in teams. when the inputs are vague, AI fills in the gaps with something that looks right but isn’t really useful.

In real work the difference usually comes from context - constraints, data, and a clear sense of what the output is actually for. and without that, the output ends up generic no matter how good the model is.

ralph's avatar

That’s such a novel approach to it. Still doesn’t Ai get generic answers because it wants to produce generic answers?

Chandra Narayanan's avatar

Great question. Not really. We have done extensive testing. If you increase the amount of the right level of specificity, it does produce highly specific results.

IN fact, even now, it can be highly specific but wrong. (so my title is actually misleading) because it makes the wrong assumptions.

ralph's avatar

Does it get answers close but wrong a lot? Someone suggested it does that a lot

Chandra Narayanan's avatar

It depends on if the context it missed is critical or not.