What AI is good at
AI is good at three things in sourcing: understanding imprecise natural-language queries, cross-referencing multiple attribute filters at once, and finding 'similar items' across vendors based on visual and structural features. 'A transitional tight-back sofa around 84 inches with track arms and performance fabric' used to require manual filtering across six sites — now it returns results in one query.
What AI is not good at
AI doesn't understand relationships, pricing negotiations, or rush orders. It can't call your rep for you when you need a fabric swatch overnight. And it can't replace the judgment call of 'this frame works in this room' — that's still a human decision.
The honest picture
AI search replaces the 'open 10 tabs, hunt through filters, compare in a spreadsheet' part of sourcing. It doesn't replace the relationship part, the judgment part, or the follow-up part. Designers who expect AI to do the whole job end up disappointed; designers who use it to cut 80% of the clicking get their sourcing time back.
Where it's headed
The direction is integration — AI search connecting to project management, to budget tools, to client presentation generation. The endpoint isn't 'AI picks your sofa,' it's 'sourcing becomes a 10-minute task inside a larger design workflow.'