Smart Search: How to Use It

Keyword Search vs. Natural Language Search

When searching videos on the Search page, you'll see two options at the top:

• Keyword Search - Build your search term by term using keywords, proximity search, and filters. Best for precise, repeatable searches you want to save as Campaigns and monitor over time.
• Smart Search- Describe what you're looking for in plain English. Best for exploratory searches when you're not sure exactly which keywords to use or want to find meetings based on context rather than specific terms.

A council member doesn't say "sales opportunity" or "area where technology could help." They say "we've been doing this manually for years and it's killing us" or "the state is pushing us to modernize our permitting system." Your query should be about what was said in a conversation you're hoping to find — not a label you'd put on it afterward.

This is also where Smart Search differs from an LLM like ChatGPT. An LLM interprets your intent and reasons toward an answer. Smart Search works more like Google: it rewards topic-level specificity.

The more precisely you describe the content of the conversation, the better your results. You wouldn’t search google with “webpage about product management software”, you’d just say “product management software”.

Practical Rules

Words to avoid — if the word describes how you're thinking about the conversation rather than what it's about, cut it.

• Meta words (describe the act of talking, not the content): conversation, discussion, mention, topic, meeting
• Command words (address the search itself): find me, search for, show me, looking for, I want to see
• Intent words (what you want to do with the info, not the info itself): opportunity, sales potential, competitive threat
• Context-describing words (describe the data source, which is already uniform): government, agency, council, public sector, transcript

Shorter is often better. 3–6 words of precise, content-level language typically outperforms a long descriptive phrase. You're rewarded for specificity, not completeness.

If results feel off, go more literal. The most common failure mode is being too abstract. "Public safety technology investment" is abstract. "Body camera replacement" is literal. Literal wins.

Examples by use case

Competitive monitoring

• ❌ competitive threat discussion
• ✅ Axon Motorola Taser contract or switching vendors public safety

Budget signals

• ❌ budget opportunity conversation
• ✅ permitting software upgrade funding or ERP system replacement budget

Pain point discovery

• ❌ problems we could solve
• ✅ overwhelmed permit applications manual or outdated financial reporting system

Pre-call research

• ❌ what this official cares about
• ✅ Search the official's name + a topic — or use Speaker Profiles, which are purpose-built for tracking individuals

Questions? Contact us at support@cloverleaf.ai