Traditional search fails when you can't remember the exact term. Semantic search finds what you mean, even when you don't use the right words.
You're looking for a ticket about the payment integration, but you can't remember if it was called "Stripe integration," "payment gateway," or "card processing." With keyword search, you'd need to try all three. With semantic search, you type "credit card payments" and it finds all of them.
How semantic search differs from keyword search
Keyword search looks for exact matches. If the ticket is called "Stripe integration — card payments" and you search for "credit card processing," keyword search returns nothing. The words don't match.
Semantic search converts both your query and the data into mathematical representations of meaning (embeddings). "Credit card processing" and "Stripe integration — card payments" are semantically similar even though they share few words. The search finds the match because the meaning overlaps.
Where this matters in practice
Semantic search is most valuable when your data uses inconsistent terminology — which is always, in agency work. One project calls it "QA," another calls it "testing," a third calls it "quality assurance." Semantic search treats them as the same concept.
It also helps with natural language queries. "tickets that are about the front page" will find a ticket called "Build homepage hero section" even though none of the words match. This is the difference between search that works like a database query and search that works like talking to a person.