When you ask an AI analytics agent about your data, you're rarely looking for a single metric. You want to know what changed, why it changed, and what to do next. A drop in conversion. A spike in traffic. An underperforming page. An unexpected customer journey. These are all signals of something bigger. The real challenge is making sense of what's driving these changes.
We analyzed 44,201 customer conversations with our AI analytics assistant Sense Analyst from Q1 2026 to understand how people are actually using analytics agents. Here's what we found.
Key insights
Most people use analytics agents to investigate change, not just retrieve data. We found 40.5% of conversations were diagnostic, focused on explaining a drop, spike, or anomaly.
Users want different levels of help. Some hand off a specific task, others want the agent to run a full investigation. A good analytics agent needs to support both.
The journey often starts before the analysis. Around 18% of conversations were pre-analytical—people asking how to navigate the platform or where to begin.
Context drives intent more than role does. The same person may need a quick answer one day and a deep investigation the next. A fixed experience can't serve all of those moments.
1. People want to understand the ‘why’ behind the numbers
The first finding was clear: people aren't using analytics agents to retrieve metrics. They're using them to investigate change.
In 40.5% of the conversations we analyzed, users were trying to explain a change or anomaly.
They already had something in front of them:
A drop in conversion
A spike in traffic
An underperforming page
An unexpected customer journey
The analytics agent became the next step in the investigation—moving from a signal to a question to an explanation they could act on.
Signal → Question → Investigation → Explanation → Action
This is an important shift. Traditional analytics starts with a dashboard. Agentic analytics can start with a question that emerges while someone is already looking at the numbers. The goal isn't just to show people what happened. It's to help them understand why—and what to do about it.
2. Users give agents different levels of autonomy
Not everyone wants the same kind of help, and that's exactly what we saw in the conversations.
Sometimes users had a specific task in mind or they wanted to test a hypothesis. Other times, they wanted the agent to investigate a broader problem or track something over time.
User intent | Share of conversations |
|---|---|
‘Do this task’ | 16.5% |
‘Investigate the problem’ | 11.6% |
‘Check my hypothesis’ | 4.6% |
‘Keep track of this’ | 3.3% |
The further users move along that spectrum, the more of the analytical process they hand off to the agent.
That doesn't mean users always want maximum automation. It means a good analytics agent needs to be flexible—supporting everything from a focused, one-shot request to an open-ended investigation.
The best experience isn't more. It's the right level of support for what that person needs, right now.
3. The journey starts before the analysis
This was the finding that surprised us most: not every conversation began with an analytical question.
Around 18% of conversations were pre-analytical—users asking for help with setup, navigation, or platform questions. Sometimes the challenge isn't finding the answer. It's knowing where or how best to start.
A user might have a business question but not know which report to open, which segment to build, or where in the platform the relevant data lives. Before they can get to the insight, they need to get oriented.
An analytics agent shouldn't just help people interpret data. It should help them reach the right data in the first place.
The analytical journey often starts before the first chart appears.
Want to learn more? Download our practical guide to Sense.
4. Context is everything
What users ask depends on their context, not their role or team.
The same person might want a quick answer one day, investigate a performance drop the next, and check whether a recent change actually worked the week after.
Quick check → Deep investigation → Validation
And it’s not just the type of question that changes, but the content they’re asking agents about. They could be asking about conversion and drop-offs one day, and bugs or errors the next.
A fixed experience built for one team can't serve all of those moments. The future of agentic analytics is an intelligent layer that adapts to what people need, when they need it.
That might look like:
A quick answer
A deeper investigation
Decision validation
Ongoing monitoring
Actionable recommendations
What this tells us about the future of analytics
Analytics agents aren't just changing how people access and understand data points. They're changing how organizations investigate, decide, and act.
The process is evolving: Form the question → Investigate → Validate → Monitor → Take action
This is what we built Sense for.
Sense is Contentsquare’s AI layer, helping everyone on your team dive deep into the data—whatever their role or level of analytics experience. It answers complex questions, delivers clear results, and walks you through actionable recommendations.
Our data shows teams across organizations want more than data access. They want help making sense of what they see, and moving forward with confidence.
And that’s what Sense does: helps more people turn signals into understanding, and understanding into action.
Methodology:
The analysis covers 44,201 Sense Chat and Sense Analyst conversations from Q1 2026. Conversations were classified by behavioral pattern, user objective, product surface, and user role, using aggregated patterns rather than individual customer data.
![[Headshot] Sara Hefny - Product Researcher at Contentsquare](http://images.ctfassets.net/gwbpo1m641r7/4HVhhh6PSwVI0DSXP3vdyM/d5e6f0140503de24363da15209a8f7c5/Sara_Hefny.jpeg?w=512&q=85&fit=scale&fm=avif)
Sara trained as an anthropologist uncovering the hidden friction in systems refugees navigate during displacement and resettlement, and has spent her career applying that same lens to digital products. She’s still surprised by how much the two have in common.
![[Visual] Blog header how people use analytics agents](http://images.ctfassets.net/gwbpo1m641r7/156ROMbtUMD1v7mfaaXSCb/3a04911fa7c2dd111ab3e9eeccfb135c/AdobeStock_963706233.jpeg?w=1280&q=85&fit=scale&fm=avif)

![[Asset] Blog — AI-referred traffic Hero](http://images.ctfassets.net/gwbpo1m641r7/2rkGm0XqF2971lPIa2dIXL/699b4b8ff946438773996ca09e304d34/AdobeStock_920265101.png?w=624&q=85&fit=scale&fm=avif)