Working slice · Data & Reporting · runs in your browser

The data analyst, automated.

Ask in English. The agentic backend has already translated each question to SQL — and it executes live, right here, no server, over a synthetic ENSEK-shaped energy-retail dataset. Billing reconciliation, the debt book, estimated-read exposure, revenue reporting. Don't trust it? Edit the SQL and re-run.

loading dataset…


Analyst questions

SQL — authored by the AI, runs live AI assist

Editable. This is the real query alasql runs against the in-memory dataset. Change it and hit Run to prove it isn't canned.

Result automated

Pick a question.

Decision-ready read AI drafted

⊕ Human control point — the irreducible core


Provenance — no fabricated capability

What's real here, exactly.

  • Live: every SQL query executes in your browser via alasql (vendored, MIT) against the loaded dataset. Row counts, headline sums, charts and the billing reconciliation are computed at click time from the actual result set — re-derived from tariff maths, never read from a planted flag.
  • AI-authored, then frozen: the natural-language→SQL translation and the decision narratives were written by the Paperclip agentic backend on 2026-06-08 and committed as data/pipeline.json so they're inspectable. The free-SQL box lets you go beyond them.
  • Synthetic data:
  • Not connected to ENSEK: nothing here touches a real ENSEK or customer system. This is a packaged proof of the operating model, not a deployment.

Where humans stay in the loop →