Every business already has the answers. They sit in the ERP, the CRM, the accounting system, a dozen spreadsheets and an inbox. The problem was never that the data did not exist. The problem was that getting an answer meant asking someone to build a report, waiting for it, and then asking a follow-up question that started the whole cycle again.

OneSage is our attempt to remove that cycle. You connect your systems, you ask a question in plain language, and you get an answer from live data in seconds. Under the hood there is a lot happening to make that feel simple.

This post walks through how it works, layer by layer, and ends with the reason we think it replaces the dashboard rather than decorating it.

Reading the schema before answering the question

A database does not describe itself in business terms. It has tables with names like sales_invoice and columns like posting_date and outstanding_amount. A person in finance thinks in terms of overdue invoices, customer groups and this quarter.

The first thing OneSage does when you connect a system is read its schema and build a semantic map. It learns which tables hold what, how they relate, and what the columns mean in the language of the business. That map is what lets a question like "Show me overdue invoices above $10K" resolve to the right tables, the right joins and the right filter without anyone writing a line of SQL. It works across PostgreSQL, MySQL, SQLite, MSSQL, Oracle and MongoDB, and across the systems people actually run: ERPNext, Zoho, Salesforce, SAP, Excel and SQL, Google and Amazon, among more than 100 connectors.

Turning language into a query, safely

Once the intent is understood, OneSage writes the query. This is the part that makes people nervous, and rightly so. A model that can write SQL against a production database needs guardrails, not good intentions.

So the query layer is read-only by default. It applies row limits and timeouts, so a badly phrased question cannot lock up a system or pull the entire ledger into memory. Access follows the same SSO and role-based permissions the rest of your stack uses, so a person sees what they are allowed to see and nothing more. The result comes back as a chart or a table built from live data. The published response time is 0.8 seconds. Answers in seconds, not days.

One question, many systems

Most real questions cross a boundary. Pipeline lives in the CRM. Cash lives in accounting. Stock lives in the ERP. A dashboard is usually built on one of them, which is why so many decisions get made on partial pictures.

OneSage joins across sources. A founder's daily briefing on revenue, pipeline, cash and alerts is one question over four systems, not four logins. It can also put your internal numbers next to live web research: competitor prices, industry news, a change in regulation. And it does not care what language you ask in. Ask in Arabic, get the answer in German, send the report in French. It works in more than 140 languages.

Documents and agents: the work after the answer

An answer on screen is often only half the job. Someone still has to put it in a deck for the board or a spreadsheet for the bank. OneSage generates PPTX, PDF, XLSX and DOCX directly from your data. "Create a board presentation from last quarter's data" returns a formatted deck in under 30 seconds. The numbers are pulled at generation time, so the document reflects the system as it is, not as it was when someone last exported it.

The last layer is the one we think matters most. Answering questions is useful. Not needing to ask is better. An autonomous agent in OneSage does what a diligent analyst would do if they never went home:

  • watches a KPI and tells you when it moves
  • monitors competitors and researches leads
  • sends a scheduled briefing to your email or WhatsApp every morning

Deploying one takes about 30 minutes. From then on it runs without prompting. Agents that work while you sleep.

Why this beats a dashboard

A dashboard is a set of answers to questions someone asked months ago. It is fixed at build time. When the business changes, or the question changes, it goes stale, and the person who built it becomes the bottleneck for fixing it. Ask a follow-up and there is no follow-up. Only another ticket.

Live questions invert that. The question is asked at the moment of decision, against the data as it stands. There is no report to maintain because there is no report. And because the whole path from question to answer sits under one set of controls, the answer is safe by construction: read-only, limited, timed, permissioned.

None of this requires handing over your data. OneSage is built to enterprise standards, with SOC 2, GDPR and ISO 27001, and on the Enterprise tier it runs on your own infrastructure. Getting started is short: sign up, connect your apps in about five minutes, ask a question in ten, deploy an agent in thirty, and act on what you find.

If that sounds like the way you would rather run a business, you can start at onesage.ai.

Salman Mulani Founder, Onesage Inc.