Context

The client, a benchmark in the conversational commerce technology sector, decided to modernize its data and Business Intelligence platform. In this context, the journey involved migrating from an environment built on Databricks and Power BI to the Google Cloud ecosystem, using BigQuery as the data layer and Looker as the visualization layer.

However, before executing the migration, the client needed to answer an essential question: “What exactly exists today, and what is the actual effort required to rebuild it in the new stack?”

Therefore, Programmers was brought in to conduct this comprehensive technical assessment, mapping the current environment, sizing the complexity of each artifact, and delivering a solid foundation for migration planning.

The Challenge of Dashboard Analysis

The legacy BI environment was large, heterogeneous, and opaque:

• Massive Scope: More than 210 active dashboards within the assessment’s scope.

• High Complexity: Panels containing 15 to 20 interconnected tables, many with many-to-many and bidirectional relationships.

• Excessive Redundancy: A single dashboard had up to 317 metrics, of which approximately 70% were duplicates with slight variations—something only noticeable upon individual inspection.

• Lack of Standardization: Each table came from a different data source, with no standardization across teams.

• Inconsistent Modeling: Different levels of technical maturity among the teams that created the panels resulted in inconsistent modeling styles.

As a result, conducting this survey manually would take an average of 1 day per dashboard, projecting over 200 business days just for the discovery phase, even before any migration could begin. In other words, a timeframe completely incompatible with the client’s modernization goals.

The Solution

The Programmers team structured the assessment into three stages: notebook analysis, data artifact analysis, and a complete scan of the BI environment. We also developed four proprietary AI-augmented engineering tools, using Claude as a development copilot:

• Automated dashboard discovery: A script integrated with the Power BI REST API (authenticated via Service Principal in Azure) that lists all workspaces and dashboards currently in use, discarding discontinued artifacts.

• Secure content extraction: Using PBIP (Power BI Project) and PBIT (Power BI Template) formats, which separate the semantic model, report, and templates into open files (JSON and DAX). This allowed the AI to analyze the structure of the panels without exposing any sensitive client data.

• Intelligent modeling audit: Customized commands in Claude to audit relationships between tables, filter system tables (which artificially inflated the complexity score), and map data lineage.

• Descriptive summary & complexity score: Automated generation of a natural language description for each dashboard, accompanied by a complexity score that served as direct input for sizing the migration effort.

Throughout the process, the team applied prompt engineering to refine the AI’s behavior. For example, guiding it to ignore system tables and distinguish legitimate metrics from duplicates, ensuring the results reflected operational reality. Most importantly, the workflow preserved a human-in-the-loop approach: AI accelerates, but the engineer decides.

Results

• Unprecedented Speed: Reduction from 1 day to just minutes per dashboard. The analysis that would have required over 200 business days was executed in a fraction of that time.

• Discovery of Hidden Inefficiencies: Automated identification of duplicate metrics, abandoned dashboards, misused system tables, and configuration errors that would have easily gone unnoticed in a manual analysis.

• Accurate Sizing of Migration Effort: A standardized complexity score allowed the client to plan sprints, allocate teams, and forecast deadlines based on objective data rather than subjective estimates.

• Secure Analysis: Zero exposure of sensitive data. Using PBIP and PBIT formats ensured the AI only worked with metadata, strictly adhering to the client’s data governance.

• Applicability Beyond Migration: The same tools became the foundation for ongoing BI governance—cleaning up orphaned reports, identifying redundant metrics, auditing quality, and detecting outliers.

• Scaling Without Losing Control: The human-in-the-loop model ensured that architectural and prioritization decisions remained with the engineers, while the AI handled the heavy lifting.

Conclusion

This case demonstrates our vision of AI-augmented engineering: empowering the professional. In other words, transforming weeks of manual labor into minutes of intelligent analysis, without compromising technical rigor, data governance, or human judgment.

When AI is applied with purpose, method, and proper engineering, it ceases to be just a promise and becomes a real competitive advantage in terms of deadlines, quality, and decision-making capabilities.

Therefore, if your company is facing a data modernization or platform migration challenge, or simply needs greater efficiency through applied AI, reach out to Programmers. We transform complexity into clarity, and clarity into results.