Chat with your data

Ask a question — from root cause analysis to what-if scenarios — and get in-depth answers with IBCS-standard charts and written explanations. No BI ticket, no waiting. The AI inherits your business context, so you can rely on the numbers the same way you'd rely on a trusted analyst.

01 — Analysis types

What kinds of analysis can I run in chat?

Zebra AI's chat works with whatever data source is connected to your story — Excel, CSV, SQL databases, Power BI semantic models, Snowflake, Databricks, or OneLake. Ask in your own words, and the AI delivers in-depth answers: charts, variance breakdowns, written explanations, and actionable recommendations. Questions that used to require a BI ticket or a data team's time are now something any business user can answer directly.

The types of analysis span five core areas, and they work across finance, sales, people analytics, operations, or any domain your data covers.

Root cause analysis

Understand what's driving a number and why.

What-if and scenario planning

Model the impact of a change before it happens.

Ad hoc exploration

Answer any question without building a report first.

Benchmarking

Compare performance across segments, time periods, or standards.

Forecasting

Project future performance from historical patterns.

The AI understands business concepts natively — variances, KPIs, seasonality, plan vs. actuals, year-over-year comparisons — and adapts to the domain of your data. Every answer includes the reasoning, not just the result.

02 — Starting a conversation

How do I start a conversation?

Once Zebra AI generates your dashboard, you'll see suggested questions as clickable pills above the chat input. These are AI-generated based on your data, the filters applied to the dashboard, and the commentary on it — so the suggestions reflect exactly what you're looking at.

Tap a suggestion to send it instantly, or type your own question. While the AI works, you'll see a "Thinking" indicator. Once the response is ready, it appears in the conversation — often with a chart and a written explanation side by side.

Every follow-up builds on the full conversation history. Ask "Now break that down by region" or "What about Q3 instead?" and the AI carries context forward — no need to repeat yourself. If your question is ambiguous, the AI asks a clarifying question before proceeding.

03 — Business context

How does the AI understand my data and my business?

The chat doesn't start from scratch — it inherits everything you've set up along the way.

When you connect data and configure it in the data preview, you can provide business context that shapes how the entire product works: your output language, fiscal year, custom terminology, KPI definitions, seasonality rules, and instructions at both the dataset level and the individual column level. Any calculated measures you create in the data preview are also part of this context.

All of that carries forward. The dashboard is built with this context, and the chat has full access to it too. When you ask a question, the AI already knows your dimensions, measures, scenarios (actuals, prior year, plan, forecast), financial statement logic, and what your columns mean. Variance comparisons automatically use the right baseline. KPIs you defined are available to plot and reference.

The chat starts in sync with your dashboard. Whatever filters you've set — a specific region, time period, or segment — the AI picks them up the moment you open the chat, so your first question already reflects the view you're looking at. Ask "What are the top drivers of the variance?" and you get an answer scoped to your filters, without re-stating them.

You can see exactly how the AI reasoned through your question by expanding the Thought process on any response. It shows which analysis approach was chosen, which columns and measures were used, and why the answer is structured the way it is. Collapse it when you just need the headline; expand it when you need to verify the methodology before presenting. For more on how the numbers behind each answer are computed, see the trust and accuracy section of the knowledge base.

If you modify your dataset between questions — editing columns on a connection, adding a calculated measure — the chat detects the change automatically and adjusts. No need to start a new conversation.

04 — Charts and visuals

What charts can the chat create?

When the AI answers with a visual, it uses Zebra BI charts — the same IBCS-standard visuals on your dashboard. It picks the chart type that best fits your question, but you can always request a specific type.

Chart type	Best for
Variance	Comparing actuals vs. plan, prior year, or forecast — the signature IBCS view
Bar	Ranking categories (top products, regions, cost centers)
Line	Trends over time (monthly revenue, quarterly growth)
Area	Trends with volume emphasis
Waterfall	Showing how components add up to a total (revenue to EBITDA bridge)
Pin	Highlighting individual data points across categories

05 — Editing and measures

Can I edit or delete a question?

You can refine or remove your most recent question directly in the conversation. Hover over your last message to see the options:

To edit: Click the pencil icon ("Edit question"), modify the text, and click Update to send the revised question. Click Cancel to discard. The updated question is sent as a new message — the original stays in the history.

To delete: Click the trash icon ("Delete question and response"). Both your question and the AI's answer are removed.

Edit and delete are available on your last message only. Edit is disabled if your data is disconnected; delete is always available.

How do calculated measures work in chat?

Calculated measures — reusable KPIs like gross margin %, revenue per employee, or cost ratios — are created in the data preview using a dedicated UI, as part of setting up your business context. Once defined, they carry through to both the dashboard and the chat: the AI can plot them in visuals, reference them in analysis, and use them to answer questions.

You can also ask the chat to create a calculated measure on the fly. Type something like "Create a gross margin percentage measure" and the AI adds it to your data model using standard arithmetic. Measures can be defined across scenarios (actuals, prior year, plan, forecast) and include a direction — whether higher is better (revenue) or worse (cost) — so variance charts color-code correctly.

For one-off calculations that don't need to persist as a named measure, the AI runs the analysis directly in the conversation.

06 — FAQ