Who it's for
Operations and line-of-business leaders with a list of weekly checks, reviews and reconciliations that each cost a skilled person hours, and no safe way to hand them to AI.
IT who need something that connects to Oracle, SQL Server and PostgreSQL as they are, deploys in their own cloud, and exposes a stable interface to existing applications.
Compliance, audit and data owners** who need to see what ran, on which version, with which data, and who approved what.
What it does
Answers questions and runs queries against your data, safely. A semantic layer gives your metrics and dimensions business names. The model selects names; CrossApply compiles the query. Structured queries run with no model at all, so the same inputs give the same numbers every time.
Carries out work as a defined workflow. A workflow is a described job: fetch these figures, compute this comparison, ask this team to approve, update this record, send this message. Most steps need no model. Where judgment is required, the model's answer must cite the data it came from.
Keeps people in charge. Checkpoints, tasks and escalations route to the right people or groups, with deadlines, reminders and separation of duties, in an approvals application with single sign-on.
Plugs into what you run today. Each workflow has a versioned input and output contract. Your applications post a request and receive a result, with a webhook when it's done. The workflow can change underneath; the contract can't change silently.
Why it's different
The model never writes SQL. Access goes through a semantic layer your team reviewed. Queries are compiled, bound and capped. That is the control most "chat with your data" tools can't offer.
Writes are declared, previewed and approved.An action is one declared statement or procedure, one table, one transaction, an expected row count, and a request ID that stops repeats. The builder previews which rows would change before anything is saved.
Numbers come with evidence.When a workflow reasons freely, every figure in its answer is traced to the data it came from before the answer is accepted, and reviewers see which figures were copied and which were calculated.
Describe it and it builds it. Describe the job in plain language, or start from a template. The designer builds the workflow, trials it against your real data, and shows you the result. Give it a known answer and it won't publish until it matches.
A library of tested business calculations. About 150 ready-made, versioned calculations: reconciliation and matching, aging and SLA, validation with real check digits, rules and scoring, dates and periods, and domain methods for finance, AP/AR, HR and payroll, CRM, support, lending, wealth, marketing, risk, health administration and life and annuity. Limits and rates are inputs, never hard-coded.
Built for the audit. Immutable versions, every run pinned to the version it ran, full traces, decisions attributed to people, audit and error logs, scheduled checks that known answers still hold, archive and export to your warehouse.
Your cloud, your databases.Self-hosted. Oracle, SQL Server and PostgreSQL as they are today. No warehouse migration first.