Dynamics NAV · Business Central · GP · AX · NetSuite · SAP
Your ERP knows the answer. It just won’t tell you.
Cetari gets the data out, puts it in one place, and lets anything query it.
You do not need a six-figure data stack to ask your ERP a question.
Trapped data looks like this.
-
The report that takes three days.
-
The JET export that times out at row 40,000.
-
The one person who knows how to pull the item ledger.
-
The inventory sheet on SharePoint that disagrees with the ERP.
-
The board question nobody can answer by Friday.
The data exists. It has existed the whole time, sitting in the ERP and the spreadsheets around it. Getting it out is the part nobody budgeted for.
Out of the ERP, into one place, open to everything.
-
Your data is trapped.
It lives in a legacy ERP that answers questions slowly or not at all, plus spreadsheets and side systems scattered around it. Every answer requires a ticket, an export, an analyst, or a prayer.
-
We get it out. All of it.
Cetari extracts from the ERP, the SharePoint workbooks, and the auxiliary databases, and aggregates everything into a single managed warehouse that belongs to you — single-tenant storage, in open Parquet files you could walk away with tomorrow. The answer reflects the whole organization, not one system.
-
Then anything can query it.
People in plain English. Excel and Power BI through ODBC. Analysts through SQL. AI agents through MCP. Same data, same answers, every surface.
Diagram: five sources — a legacy ERP such as Dynamics NAV, Business Central, GP, AX, NetSuite, or SAP; SharePoint workbooks; auxiliary databases; operational side systems; and spreadsheet exports — all flow into one warehouse owned by the customer, built on open Parquet files, queried by DuckDB, and stored single-tenant. From that one warehouse, four query surfaces fan out: plain-English chat, Excel and Power BI over ODBC and JDBC, direct SQL, and AI agents over MCP.
Questions that used to be projects.
These are the kinds of questions a warehouse answers in seconds. No ticket, no export, no waiting on the one person who knows the tables.
Which customers ordered less this month?
What inventory is aging past 60 days?
Which vendors are late most often?
What did we ship by region last quarter?
What changed in margin by product line?
Before A JET report that times out.
After The answer in four seconds.
Before “Ask Dana — she has the spreadsheet.”
After Ask the warehouse directly.
Before A three-day wait for a custom report.
After Type the question, read the table.
Assistant
connected · cetari warehouse
Which customers ordered less this month, and which product categories drove the decline?
Three accounts account for most of the decline, led by frozen goods and dry grocery. Largest month-over-month drops:
| Customer | MoM change | Primary category | Est. impact |
|---|---|---|---|
| Northstar Foods | -18.4% | Frozen goods | $184k |
| Metro Provisions | -13.1% | Dry grocery | $91k |
| Valley Market Group | -9.7% | Dairy | $64k |
Queried 2 tables · sales_invoice_lines, customers · 1.8s
What you have to do: almost nothing.
-
Cetari connects to your ERP read-only. Nothing in your system of record changes.
-
No migration. The ERP stays where it is, doing what it does.
-
No new software for your team to learn. Excel still works. The browser still works.
-
Running in days, not quarters.
This is a managed service. We run the extraction, the modeling, and the warehouse. You ask questions.
For the skeptical IT person.
Here is why your IT director will sign off.
- Open Parquet, not a proprietary store
- Your warehouse is plain Parquet files in single-tenant cloud storage. If you leave, you take the files. The data is never hostage.
- Single-tenant isolation
- One customer per warehouse. No shared compute, no shared storage, no pooled tenancy. Access is governed — who sees what is a rule, not a habit.
- Standards you already support
- DuckDB under the hood. ODBC and JDBC on the wire. Power BI, Tableau, and Excel connect the way they connect to anything else.
- AI access without production credentials
- Agents query through an MCP server scoped to the warehouse. No AI tool ever holds raw ERP credentials or touches the production system.
- Read-only extraction
- Cetari reads from the ERP and writes nothing back. Your system of record is untouched by design, not by promise.
The chain this replaces
ERP → ingestion vendor → cloud warehouse → transformation layer → semantic layer → BI tool → analyst → spreadsheet
ERP → Cetari → answer
| The usual stack | Cetari |
|---|---|
| Five vendors, five contracts, five invoices | One managed service |
| An analyst queue for every new question | Self-serve through chat, Excel, SQL, or MCP |
| Data locked in a proprietary warehouse | Open Parquet you can take with you |
Built for the people stuck waiting on reports.
Operations
See what is aging in inventory and which vendors are slipping on the day it happens, not at month-end. No ticket filed.
Finance
Settle questions about margin and revenue in minutes. The numbers match across the room because there is one set of numbers.
IT
Retire the report backlog without building a data platform. A read-only connection and nothing new to babysit.
Analysts
Clean modeled tables in Power BI, Excel, or raw SQL. Stop re-exporting the same ledger every Monday.
Bring us the question you can’t get answered.
Tell us what system you run and what you are trying to find out. We’ll show you the same question answered from a warehouse in seconds.