Anyone who has built a Power BI model on top of SAP knows the pain. The table names are cryptic, the field names are shorter than they should be, and the relationships between headers and line items are rarely where a newcomer expects them. Writing correct DAX against BSEG, BKPF, MARA, and the rest takes hours of schema spelunking before you ever get to the measure itself.
If your team is still debating whether Power BI belongs in the stack, start with Why Every Tech Team Needs Power BI in Their Stack. This post assumes you are already past that conversation and are trying to ship reports faster.
The Problem: SAP Schemas Are Hostile to Newcomers
SAP's table design predates modern BI tooling by decades. Financial postings live in BKPF (header) and BSEG (line items) with a composite key that must be joined correctly; otherwise, your totals will double. Material data in MARA links out to MAKT for descriptions, MARC for plant data, and MBEW for valuation. Nothing is self-explanatory, and the field names (MATNR, WERKS, DMBTR) give you no hint about what they mean unless you already know.
The result is that even an experienced Power BI developer can spend the first half of a project just reading documentation and testing joins. The actual DAX, the part that delivers business value, gets pushed to the end of the sprint.
What Changes When AI Generates the DAX for You
The shortcut most developers reach for is to paste a schema into ChatGPT or Copilot and ask for a measure. That works, technically, but it has two real problems. First, the AI has no idea what your table conventions look like, so it guesses. Second, you just pasted your client's schema into a general-purpose chatbot.
An ERP-aware tool closes both gaps. It already knows the structure of SAP's standard tables, understands Star Schema best practices, and generates measures that follow Microsoft's DAX function reference patterns. A request like "monthly revenue by profit center, current year vs prior year" becomes a complete measure with CALCULATE, SAMEPERIODLASTYEAR, and the right filter context, in seconds.
The Privacy Problem Nobody Talks About
Pasting a real ERP schema into a public AI tool is a disclosure risk most consultants do not think through. Table names, custom Z-tables, field comments, and naming conventions can identify a client, reveal a migration in progress, or expose business logic that should not leave the engagement.
Velqur solves this with client-side sanitization. Before your schema ever reaches the AI, the browser strips SQL comments, DBML annotations, CDS notes, schema prefixes, linked server references, default values, and constraint names. The AI sees only the structural shape it needs to generate correct code.
Nothing that could identify your organization leaves your machine. You get a report afterward showing exactly what was removed and why.
A Practical Example
Say you need a measure for "year-to-date net revenue, excluding intercompany postings, by company code."
By hand, that is a lookup in BSEG for the posting keys, a join to BKPF for the document date, a filter against T001 for the company code, and a CALCULATE with a date filter. Easily an hour if you are
careful, longer if the model is new to you.
With Velqur, you describe the measure in plain English, the tool generates the DAX against its knowledge
of the SAP financial tables, and you paste the result into Power BI. The time savings compound across every measure in the model, and the output is consistent, which matters when another developer inherits the report six months later.
Frequently Asked Questions
What is BSEG in SAP?
BSEG is the line-item table for financial accounting documents. It stores one row per posting line and joins to BKPF, the header table, via the document number, company code, and fiscal year. Most revenue,
expense, and balance sheet measures in Power BI start from some combination of BSEG and BKPF.
Can ChatGPT write DAX?
Yes, but with caveats. General-purpose chatbots can produce syntactically valid DAX, but they do not know your table names, do not follow your naming conventions, and have no awareness of whether your model is a proper star schema. They also retain whatever schema you paste in, depending on the provider's policy. An ERP-aware, privacy-sanitized tool produces better code and does not create a disclosure problem.
Do I need to import my SAP schema into Velqur?
Velqur ships with a pre-built catalog of standard SAP tables, so for out-of-the-box reporting, you can start generating DAX immediately. If your organization uses custom Z-tables or modified field names, you can import the schema, and the client-side sanitizer will strip identifying details before anything reaches the AI.
The Bottom Line
SAP is not going to get easier to query. The tables are what they are, and the learning curve is real. But the part of the job that is pure translation, from business question to correct DAX syntax, no longer needs to take hours. With ERP-aware generation and schema sanitization built in, you can move straight from requirement to working measure without compromising client data or spending half the project relearning SAP internals.
Velqur is currently in early access and free to try at velqur.edgexene.io.