AI in Reportworq#

Reportworq's approach to AI is deliberately different from pointing a model at a database. The idea is governed AI: AI works on top of the financial content your finance team has already built, validated, and approved, not on raw databases, ERP tables, or the general ledger. This page explains the model behind that choice and what it means for you as an administrator or integration user before you turn any AI feature on.

The problem with pointing AI at raw data#

Most AI projects begin by connecting a model directly to an ERP, a data warehouse, or a database. That sounds direct, but it forces the model to do the hardest parts of finance on its own: understand the schema, understand the business logic, work out how calculations are defined, decide which numbers are correct, and then do the arithmetic. Every one of those steps adds complexity and raises the chance of a wrong or invented answer.

Finance already solved those problems. Your team knows which reports are correct, and it validates those numbers every close. The reports carry years of curated terminology, business logic, and presentation decisions.

Governed AI: AI on trusted content#

Reportworq puts AI on top of that trusted, finance-approved content rather than the raw systems underneath it. Instead of asking a model to recreate financial logic, Reportworq exposes governed reports the model can read, so the AI analyzes trusted answers instead of trying to calculate them.

This is the same shift described in the product's "AI done differently" idea: rather than forcing AI to rediscover the business on every prompt, Reportworq teaches AI how the business already communicates. Report authors enrich reports with structured business context, purpose, audience, strengths, limitations, terminology, and caveats, in a report's AI Profile. See Create an AI Profile for a report.

The practical results are:

• Lower hallucination. The model reasons over answers finance already trusts, not schema it has to reverse-engineer.
• Lower token use. Curated context is smaller and cleaner than raw data dumps, so prompts are cheaper.
• Higher trust. Answers trace back to the same reports finance already stands behind.

Every question and every answer is auditable#

Governance is not only about which data the model sees. Reportworq knows who the user is, what they are allowed to see, which financial assets are trusted, how those assets are secured, every question asked, and every answer returned. Every interaction is recorded. Administrators can review AI activity in the audit logs. See Audit logs.

Users see only what they are authorized to see#

AI in Reportworq respects the same access boundaries as the rest of the platform. A user asking a question through an AI feature sees only content they are already entitled to see. A department manager who receives a governed P&L can ask about it and get authorized detail back, without ever being given direct access to the general ledger. The AI does not widen access; it works inside the user's existing permissions.

Where AI shows up in the product#

AI in Reportworq is delivered through a small set of features, each covered in this section:

• Connect an AI provider is the one prerequisite that turns on every AI feature. Nothing below works until an administrator has added and enabled at least one provider connection.
• AI-generated insights in reports attaches model-written narrative to a distributed report, and covers the RWAIINSIGHT worksheet function that writes AI narrative into a workbook.
• The AI assistant covers the in-app assistant surfaces, which are present in the product but not yet switched on in the current build.
• The AI prompt store is where an administrator reviews and tunes the internal prompts Reportworq sends to your provider.

Who this section is for#

This section is written for administrators, who configure providers, govern access, and review audit trails, and for integration users, who consume AI-authored content in delivered outputs and downstream tools. Report authors who want to enrich a report for AI should start with Create an AI Profile for a report.

Going deeper. The single most important next step is connecting a model. See Connect an AI provider.