START WITH A PROBLEM

Reports that need engineering

Every new question becomes a ticket, and the queue is weeks long. By the time the report arrives, the question has changed.

Business Leadership PROBLEM 2 min read
Today
  • Question
  • Ticket
  • Engineeringwrites SQL
  • Reportalready stale
Every new question joins the queue. The follow-up joins it again.
With delivered dashboards and chat
  • Recurring question
  • Dashboard · saved viewopen it
  • Answer
  • New question
  • Ask the assistantplain language
  • Answer · ask next
Engineering builds the data layer once, not every report.
The queue goes away when the question does not need a query.

The symptom

A manager wants orders by region for last month, split by category. There is no report for that, so it becomes a ticket. The ticket waits behind the features that were promised to customers. Weeks later a developer writes the SQL, checks it, runs it off-hours so it does not slow the system, and sends a spreadsheet. By then the question has moved on, and the follow-up - "now only the rejected ones" - starts the cycle again.

Nobody is slow. The business asks questions faster than any engineering team can hand-build answers, and the queue is the proof.

Why it happens

Answering a question means querying the production database, and that is engineering work for three reasons. The query has to be right, which means knowing the schema and its traps. It has to be safe, because a careless scan slows order entry for everyone. And it has to be reproduced every time the question changes, because a spreadsheet does not take follow-ups. Each report is a small software project, and the business has more questions than there are projects in a quarter.

Self-serve BI tools promise to remove engineering from the loop and usually move the loop instead: now someone in the business has to learn the tool, model the data and build the report, and the queue reappears in a different team.

What changes with an indexed layer

Engineering builds one thing: the indexed copy of the collections the business asks about - orders, customers, products, invoices - refreshed on a schedule, never touching production for a read. On top of it, two kinds of answer. The recurring questions become delivered dashboards and saved views, configured to the business and opened in seconds: the Urgent orders grid at 14, the regions leaderboard with North at 395. The new questions go to an assistant that maps a sentence to the same collections and answers in minutes, with a chart or table and the next question suggested.

The ticket queue does not get faster. It gets shorter, because most of what was in it no longer needs a query written by a person. Engineering's time goes into the data layer and the application, which is what it was for.

See it on real data.

The demo instance runs dashboards, data grids and the AI Assistant on real business data. No sign-up.