A Technical Look at an AI-Assisted Tableau-to-Power BI Migration

For more than 20 years, we have designed and implemented business intelligence and analytics solutions. Tableau-to-Power BI migration is not new for us. We outline the strategic side of it in Migrating Tableau to Power BI. What has changed is how we now support selected parts of the implementation with AI.

The technical question was not whether AI could convert a workbook automatically, but which parts of an AI-assisted Tableau-to-Power BI migration could be standardized without compromising architecture, business logic, or quality.

This approach is already part of how we deliver Tableau-to-Power BI migrations for our customers. We use it to create a structured baseline for complex workbooks with numerous dashboards, worksheets, data sources, and business-critical calculations, while our developers remain responsible for architecture, validation, and report-specific decisions.

A migration is not a file conversion

Tableau and Power BI use different technical concepts. Tableau calculations are often tied to the worksheet and its visual context. Power BI places more logic in the semantic model through DAX, Power Query, relationships, and filter context.

A calculated field therefore cannot simply be copied from one tool to the other. Dependencies must be identified, parameters redesigned, nested logic consolidated, and visual behavior checked against the original report. Customized Tableau dashboards also require design decisions because some chart types, interactions, and tooltip patterns have no direct Power BI equivalent.

Some of the most demanding work lies in the underlying architecture rather than the visual conversion. Where required, source structures are redesigned into efficient Power BI star schemas, existing access logic is rebuilt using Row-Level Security, and complex calculations are reworked to resolve circular dependencies or improve DAX performance. These decisions require a thorough understanding of both the original Tableau logic and the target Power BI architecture.

Our workflow separates this work into two AI-assisted stages.

Stage 1: Building the AI-Assisted Tableau-to-Power BI Migration Baseline

The first stage covers phases 1 to 3 of our technical workflow. It creates a consistent starting point before a developer begins detailed report work.

The inputs are clearly defined: the Tableau .twb workbook, an empty Power BI .pbip project already connected to the required data sources, and any report assets. The workflow creates a controlled workspace and keeps the Tableau source read-only.

For the model migration, we developed purpose-built Python scripts to extract and structure XML metadata from the Tableau workbooks. The scripts identify dashboards, worksheets, data sources, filters, parameters, calculated fields, Level of Detail expressions, and their dependencies. This automated complexity analysis makes nested calculations, filter contexts, dependency chains, and particularly complex workbooks visible before implementation begins.

The workflow then builds a dependency graph so calculations can be processed in the correct order and classifies each field as a DAX measure, Power Query transformation, calculated column, parameter table, or another model object.

This classification is not accepted blindly. The workflow presents naming conflicts, consolidation choices, and ambiguous parameter behavior to the developer before writing changes. Once confirmed, it generates DAX measures, Power Query logic, parameter tables, and model definitions in the Power BI project, including TMDL objects where appropriate.

For the visual migration, the Tableau workbook XML is parsed to identify dashboards, visual zones, coordinates, dimensions, and chart types. These are mapped to suitable Power BI visual types. After the visual inventory is reviewed, the workflow writes the initial PBIR report structure, including pages, visual JSON, bindings, and reusable formatting elements.

At the end of the first stage, the developer receives a populated Power BI project, an initial report structure, migrated model objects, and a list of open decisions and required validations.

This result is not a finished dashboard. It is a technically structured baseline in which a large share of the recurring preparation work has already been completed.

Stage 2: Report-Specific Refinement by Developers

Phase 4 begins when the Power BI developer opens the generated project and continues working directly on the dashboard.

In our current implementation, Claude Code coordinates targeted changes, while an MCP connection provides a direct interface to Power BI Desktop (see also our article on Semantic Layer vs. Context Layer, where we describe how Claude queries Power BI semantic models via MCP). Developers can request specific adjustments, such as repositioning visuals, changing data bindings, inserting background images, adjusting DAX, adding icons, or standardizing formatting.

The workflow follows an iterative cycle: edit project files, validate the result, reload Power BI Desktop, capture and review the updated report. This cycle continues until the page meets both functional and visual requirements.

When refining individual reports based on the generated baseline, Power BI developers use AI deliberately where it provides real value. This includes translating calculations, recurring page structures, positioning visuals, formatting, and cleanup work.

AI supports these repeatable implementation tasks. Decisions about model architecture, complex business logic, Row-Level Security, performance, and report-specific solutions remain the developers’ responsibility. They determine whether a workaround is technically appropriate and whether the migrated result is analytically correct.

Traceability and Validation Are Built In

Each migration step produces a report. The model report records which fields were translated, which objects were created, and which items require verification. The visual report lists created pages, mappings, approximations, warnings, and manual follow-up actions.

This makes the migration reviewable. Unsupported patterns are not hidden, and uncertain decisions are surfaced before they become defects. The reliability of the AI-generated baseline also depends on the quality and completeness of the Tableau metadata provided to it. Clean parsing and structuring of calculations, dependencies, filters, parameters, and data sources before they enter the AI workflow reduces the risk of incorrect interpretations, particularly in complex business logic.

Validation also includes technical optimization. Developers review the semantic model, DAX logic, Row-Level Security, refresh behavior, and report performance throughout implementation.

Where the original Tableau logic does not translate well into the Power BI architecture, it is redesigned rather than copied directly. Developers still validate calculations, totals, filters, interactions, and visual behavior against the Tableau original and the business requirements.

In our latest migration project, this combination of reusable automation and expert review helped us deliver all priority workbooks on schedule.

Compared with a previous project of similar complexity, we estimated that AI-assisted formula translation, page replication, positioning, formatting, and cleanup reduced total effort by approximately 25%.

The reduction was particularly significant in repetitive implementation work, while data architecture, complex business logic, security, and final validation remained primarily developer-led.

The result is a more efficient AI-assisted Tableau-to-Power BI migration process without treating AI output as production-ready by default. Our experience defines the architecture, the quality controls, and the migration decisions. AI helps us execute the repeatable parts faster.

Dr. Özgün Köksal is Senior Data Analytics Specialist at Hopmann Marketing Analytics, focusing on building data and AI competence within marketing and data teams and helping organizations translate technical AI-driven workflows into practical, everyday adoption.

Can Weidemeier is Senior Data Engineering Specialist in the AI-ready Data Stack team at Hopmann Marketing Analytics, building tailored dashboards with Tableau and Power BI to help clients make data-driven decisions.

Note: This article was prepared with AI assistance and reviewed editorially. Responsible: Hopmann Marketing Analytics.