Chat & ad-hoc analysis
Ask your data anything. In English.
Your team asks a question the way they'd ask a colleague. Synopsis writes the SQL, runs it against your whole warehouse, and answers with the number — and where it came from.
Fourteen accounts have falling order volume and rising support contacts. Together they're $412K of annual revenue. The steepest drop is concentrated in your two largest regions.
Sources: CRM · billing · support desk
The whole dataset, not a sample
It reasons over every row you have.
A chat window that pastes your data into a prompt can only see the sliver that fits. Synopsis is different: the AI writes SQL against your warehouse, so the answer is computed over your entire history — millions of rows, every system, joined on one key.
- Real queries, real math. Counts, sums, and cohorts are computed by the database — not estimated by a language model.
- Cross-system by default. "Revenue by rep" spans CRM, billing, and the GL without anyone building a join.
- Follow-ups keep context. Drill from the answer to the why without re-explaining the question.
Three reps have two consecutive down quarters. Here's each one against their trailing average — the biggest gap is −18%.
Trust every number
Every answer shows its work.
Ask where a number came from and Synopsis shows you: the SQL it wrote, the tables it touched, the systems those tables came from. When a metric has a company-specific definition, the semantic layer applies your definition — not a guess.
- The query, visible. Anyone can open the SQL behind an answer.
- Your definitions, enforced. "Active customer" means what your team says it means, every time.
- Sources, cited. Answers name the systems they drew from.
-- generated for: "customers slipping this quarter"SELECT account_id, arr, qoq_deltaFROM metrics.account_healthWHERE qoq_delta <0AND support_contacts_30d >3ORDER BY arrDESC-- definition: "slipping" = metrics.yml › account_healthTraced to: CRM · billing · support desk
From question to artifact
Good answers become dashboards.
When an answer is worth keeping, say so. Synopsis turns the query into a live widget, adds it to a dashboard, and keeps it current — so the question never has to be asked twice.
- One prompt, one widget. "Track this weekly" is the whole setup.
- Stays live. Widgets re-run on fresh data, not saved snapshots.
- Or set a watch. Turn the same question into an alert that fires when the answer changes.
What your team gets
Built for the people who have the questions.
No SQL required
Plain English in, answers out. The people closest to the business ask directly instead of filing a ticket.
SQL when you want it
Analysts can open, edit, and reuse every generated query. The warehouse is yours to query directly.
Charts on demand
Ask for the trend, the breakdown, or the comparison — answers come back visual when a visual says it better.
Shareable answers
Send a result to a teammate with its full trace, so the debate is about the decision — not the number.
Knows your vocabulary
Products, regions, segments, seasons — the semantic layer maps your team's words to your data.
Honest about ambiguity
When a question is genuinely ambiguous, Synopsis asks a clarifying question instead of guessing.
Ask the question you've been sitting on.
Thirty minutes with your data, and you'll hear the answer out loud. Bring the question your team has never been able to get answered.
