Using AI search and analysis in Condens

There are several places where AI search and analysis lives in Condens:

AI search builds lets you ask questions in natural language instead of constructing keyword queries. Rather than matching exact terms, it looks for content that matches the meaning of what you're asking and surfaces results backed by evidence.

1. You stay in control

Whatever AI search surfaces, you can always trace it back to the original moment in your data. Results are linked to their source so you can read them in context and verify them yourself. Where AI makes suggestions, nothing is applied without your input: you can accept, edit, or discard anything before it becomes part of your research.

2. It retrieves, it doesn't interpret

When you ask a question, the AI looks through your data for moments, quotes, and content that are relevant to your question. It doesn't draw conclusions, form recommendations, or synthesize patterns on its own.

This means it works best when your question has a specific, concrete answer that exists somewhere in your data. The more you're asking it to interpret or conclude, the less reliable the results will be.

3. It won't make things up

One thing worth knowing: the AI won't make something up just to give you an answer. If there isn't enough relevant data to respond to your question, it will tell you that directly. That's intentional. A non-result is still useful information. It tells you either that the topic hasn't come up in your research, or that you need to adjust how you're asking.

4. Some questions are bigger than a search bar

Some questions are genuinely large enough to be research projects on their own. "What are our biggest product problems?" or "What should we prioritize this quarter?" won't yield useful results unless someone in a Session explicitly stated those answers or you documented them in an Artifact. That's because the AI is looking for that answer in your data, not forming one. Those questions are better tackled through your analysis and synthesis work, where you can bring together evidence and draw your own conclusions.

Used for the right things (like finding who mentioned something, surfacing specific feedback, pulling together evidence around a concrete topic), AI search is fast and genuinely useful.

The right kind of question depends on what you're trying to find out. Here are the most common goals and what works well for each.

Finding specific moments or evidence

Use this when you're looking for something concrete: a quote, a reaction, a mention of a specific topic or feature. Session search, Participant profile search, and Home search all work well here. Keep your questions specific and use the language your participants actually used.

✅ "What did participants say about the checkout flow?"
✅ "Were there any moments where users expressed frustration with onboarding?"
✅ "Did anyone mention switching from a competitor? What did they say?"
✅ "Where did users struggle during the profile setup task?"

❌ "What do users think about our product?"
❌ "What are the main pain points?"

These broader questions work better as a starting point for your own synthesis. AI search is best at surfacing evidence, not drawing conclusions from it.

Exploring patterns across data

Use this when you want to understand what comes up repeatedly across multiple participants or sessions. Home search and workspace-level Session search work well here. The broader your data set, the more specific your question needs to be.

✅ "Which participants mentioned issues with the dashboard?"
✅ "What recurring feedback came up around the notification settings?"
✅ "Who asked for a dark mode, and in what context?"
✅ "What themes came up around collaboration features across Q1 interviews?"

❌ "What should we fix first?"
❌ "What's the most important feedback we received?"

The more you narrow your scope with filters before searching, the more relevant your results will be, especially across a large workspace.

Making sense of what you've tagged

Use this when you've finished a round of sessions and want to synthesize what you've pulled out. Ask your Highlights is the right surface here, since you're working with data you've already reviewed and selected.

✅ "What topics come up most often across these highlights?"
✅ "Are there any recurring frustrations in what I've tagged?"
✅ "What are the most common feature requests mentioned here?"
✅ "Cluster these highlights by theme."

❌ "What should I focus on?"
❌ "What are the main insights?"

Grouping and mapping themes

Use this on the Whiteboard once you've pulled highlights together and want to explore structure. Clustering prompts can carry more context than a search question since you're giving the AI a frame for how to think about the material, not just matching keywords.

Simple, open-ended prompts work well when you want to explore broadly:

✅ "Group these by the type of problem described."
✅ "What are the main suggestions for improvement and areas of satisfaction in this feedback?"

More detailed prompts work well when you have a specific research context:

✅ "These highlights are from usability tests on our onboarding flow. Group the findings into areas of confusion, smooth interactions, and overall task success."
✅ "These are highlights from discovery interviews with first-time users. Group them by what surprised people, what felt familiar, and what felt missing."

Checking what your team already knows

Use this before starting new research to avoid duplicating work, or when a stakeholder asks a question and you want to surface what's already been documented. Home search and Magazine search both work here: Home if you want to search across everything (including raw data), and Magazine if a stakeholder needs quick answers based on validated findings.

✅ "What have we learned about enterprise onboarding?"
✅ "Do we have any research on why customers downgrade their plan?"
✅ "What do we know about how teams onboard new members?"
✅ "Is there any research on our mobile checkout flow?"

❌ "What should we research next?"
❌ "What are the gaps in our research?"

Phrase these as knowledge queries rather than analysis requests: "What do we know about X?" works better than "What should we do about X?" because you're looking for documented findings, not asking AI to form a recommendation.

• Filter before you search
  The broader your data set, the more a scoped query helps. Before searching at project or workspace level, apply filters by project, session, participant type, date, or custom metadata fields to narrow down to what's actually relevant.

• Use specific language
  Use the feature, flow, or product names your participants actually used. "Profile setup" will surface more relevant results than "onboarding" if that's the language in your transcripts.

• Try rephrasing if results feel off
  A slightly different question can surface different moments, especially if your participants used varied language to describe the same thing.

• Follow the source links
  AI search surfaces as much relevant evidence as it can, and everything links back to its source. Use it to get oriented and find the key evidence, then follow the source links to verify in context.

• For Whiteboard clustering, try multiple prompts
  Reapply clustering with different prompts to see how groupings shift; it's a good way to look at the same evidence from different angles. Start with an open-ended prompt to get an overview, then reapply with a more focused one to dig into a specific theme.

• Search before starting new research
  Use Home search or the Magazine to check what your team already knows before kicking off a new study. It's a quick way to find existing answers, spot gaps, and avoid duplicating research that's already been done.

You don't have to use AI search to find things in Condens. Search is available at every level and gives you full control over your query.

You can build complex searches using multiple keywords, apply filters, use and/or logic, and change sorting options to get a comprehensive view of your results. You can search across everything or scope it specifically to artifacts, projects, sessions, participants, tags, or highlights.

It's a good starting point when you know exactly what you're looking for, or when you want to combine precise filtering with an AI search on top.