Getting the Most Out of BlazeSQL

Your team now has an AI data analyst that understands your database and your business — no SQL skills required, no waiting on anyone else. This guide will help you make the most of it.

Talking to Blaze

BlazeSQL is a conversation, not a search bar. You don't type one question and start over — you have a real back-and-forth dialogue.

Say you ask, "What were our sales last month?" and get a table back. From there, just keep going:

• "Add the region column"
• "Filter to just the enterprise segment"
• "Make it a bar chart"
• "Now show the last 90 days instead"

BlazeSQL remembers the full conversation, so each message builds on what came before. If something isn't right, just say so — "That's not what I meant, I want X instead" — and it'll adjust.

But here's what really sets Blaze apart: you don't have to know exactly what you're looking for. Give it a goal and let it work through the problem with you:

• "Help me figure out why churn increased last quarter"
• "What patterns do you see in our customer acquisition over the last six months?"
• "I'm trying to reduce support ticket volume — what's driving the most tickets?"

BlazeSQL has broad business knowledge and can think strategically about your data. Give it a goal, and it'll break it into steps, run analyses, spot patterns, and suggest what to explore next. Think of it as a smart colleague with instant access to all your data — an analytical partner, not a calculator.

If you're ever unsure whether Blaze can do something, just ask.

Teaching It Your World

BlazeSQL already knows your database structure. But your business has its own language, logic, and quirks. The more you share, the better it gets.

• Your terminology. When your team says "active user," what exactly does that mean? Logged in within 30 days? Has a current subscription? BlazeSQL needs your definition, not a guess.
• Business logic. Revenue excludes refunds. "Enterprise" means 50+ seats. These rules matter for accurate answers.
• Database nuances. Status codes (1 = active, 2 = churned), timezone quirks, naming conventions — small details that prevent big misunderstandings.
• Standing instructions. "Always exclude test accounts." "Our primary metric is ARR, not MRR." Guardrails that make every answer more reliable.
• KPI definitions. Your formula for churn rate, LTV, conversion rate. When everyone — including BlazeSQL — uses the same definitions, you can trust the numbers.

Teach BlazeSQL all of this using "Help your AI Assistant learn." Think of it like onboarding a new team member — you wouldn't expect a new hire to know your internal jargon without being told. The investment pays off quickly: every question gets a better answer, automatically.

BlazeSQL also learns as you use it. When you correct or refine a result mid-conversation, Blaze detects this and asks "Did I get it right this time?" If you confirm, it automatically generates knowledge notes from the correction — no manual work needed. You'll also see suggested notes appear in the sidebar after conversations, which you can review and save. Over time, Blaze gets smarter just from your team using it.

Why Letting Blaze See Your Results Matters

By default on the web app, Blaze can see query results and use them to power its analysis. On the desktop app, offline mode is enabled by default — meaning Blaze cannot see your results. We strongly recommend disabling offline mode (by enabling "Send query results to AI" in settings) unless your organization has strict regulatory requirements that mandate keeping data entirely local. Here's why.

When Blaze can see results, it becomes a true analytical partner:

• Spotting anomalies — numbers that look off, unexpected zeros, duplicates, or data that doesn't add up.
• Fuzzy matching — if you ask about "Acme Corp" but the database has "ACME Corporation," Blaze can look at the actual values and find the right match. Without visibility, it's guessing.
• Iterative analysis — data analysis is naturally iterative. You look at results, notice something interesting, dig deeper. When Blaze can see results, it drives this process. Otherwise, you're the bottleneck interpreting every result.
• Being a thought partner — interpreting results, making recommendations, suggesting what to explore next. Without seeing results, Blaze is more of a query translator than a coworker.

When Blaze cannot see results (offline/desktop mode):

• It still generates queries and understands your database, but it's essentially going in blind.
• It can't verify whether filters returned the expected data.
• It can't self-correct when a query returns zero rows or unexpected results.
• You have to interpret everything yourself — the analytical intelligence is bottlenecked.

You might hesitate to disable offline mode, assuming that keeping data local is inherently more secure. It isn't — with "Send query results to AI" enabled, your data is fully protected: Zero Data Retention on all AI model calls (prompts and responses are not stored by Google), your data is never used to train models, everything is encrypted at rest (AES-256) and in transit (TLS), customer data is used only to provide the BlazeSQL service, and you can delete any data at any time. Your data is protected whether or not offline mode is enabled. See our Data Privacy & Security article for full details.

Offline mode exists for organizations where strict regulatory or on-premise policies mandate that data never leaves the local device. If that doesn't apply to you, disable it — you'll get a significantly better experience.

With offline mode on, Blaze still maps your questions to SQL, understands your schema, and applies all your training. But you're leaving significant analytical power on the table.

What You Can Do with Results

Many users ask a question, glance at the answer, and move on. But there's much more you can do:

• View as table or chart — bar, line, pie, scatter, map. Pick what tells the story best.
• Export to CSV or Excel — one click to bring data into a spreadsheet.
• Share a link — send colleagues the exact same results, no screenshots needed.
• Save queries — rerun your regular checks (weekly sales, daily signups) without retyping.
• Build dashboards — combine visualizations into a single view for standups or exec updates.
• Add to weekly email reports — key insights land in your inbox on a schedule, automatically.
• Customize in the graph editor — tweak colors, labels, axes, or chart type.

An insight in your chat window doesn't change anything. An insight that's shared, saved, or on a dashboard can change how your team operates.

Building an Analytical Culture

Before BlazeSQL, data questions meant asking an analyst and waiting, building a spreadsheet and hoping it was right, or — most often — just not asking. That friction is gone now.

Ask the "small" questions. "Is Tuesday or Wednesday better for email sends?" "Which rep has the fastest time-to-close?" Questions you'd never bother a data team with — but they add up to better decisions every day.

Check your assumptions. "Our best customers come from referrals." Maybe — or maybe that was true three years ago. Now you can check in 30 seconds instead of debating in meetings.

Make it a daily habit. The teams that get the most from BlazeSQL don't save it for monthly reports. Before a meeting, check the numbers. After a launch, look at the impact. When a question comes up in Slack, answer it in real time.

Share what you find. Build shared dashboards for each team. Set up weekly email reports. When someone finds a surprising insight, post the link. That's how the habit spreads.

Over time, conversations shift from "I think our conversion rate is around 3%" to "It's 3.2%, up from 2.8% last month, strongest in mid-market." Opinions get replaced with evidence. That's the real transformation — becoming a team that moves from "I think" to "I know."

Updated on: 05/02/2026