Source URL: https://cloud.google.com/blog/topics/public-sector/indiana-dot-saved-360-hours-of-manual-effort-to-meet-a-30-day-executive-order-with-google-ai/
Source: Cloud Blog
Title: Indiana DOT saved 360 hours of manual effort to meet a 30-day executive order with Google AI
Feedly Summary: Public sector agencies are under increasing pressure to operate with greater speed and agility, yet are often hampered by decades of legacy data. Critical information, essential for meeting tight deadlines and fulfilling mandates, frequently lies buried within vast collections of unstructured documents. This challenge of transforming institutional knowledge into actionable insight is a common hurdle on the path to modernization.The Indiana Department of Transportation (INDOT) recently faced this exact scenario. To comply with Governor Mike Braun’s Executive Order 25-13, all state agencies were given 30 days to complete a government efficiency report, mapping all statutory responsibilities to their core purpose. For INDOT, the critical information needed to complete this report was buried in a mix of editable and static documents – decades of policies, procedures, and manuals scattered across internal sites. A manual review was projected to take hundreds of hours, making the deadline nearly impossible. This tight deadline necessitated an innovative approach to data processing and report generation.Recognizing a complex challenge as an opportunity for transformation, INDOT’s leadership envisioned an AI-powered solution. The agency chose to build its pilot program on its existing Google Cloud environment, which allowed it to deploy Gemini’s capabilities immediately. By taking this strategic approach, the team was able to turn a difficult compliance requirement into a powerful demonstration of government efficiency.From manual analysis to an AI-powered pilot in one weekOperating in an agile week-long sprint, INDOT’s team built an innovative workflow centered on Retrieval-Augmented Generation (RAG). This technique enhances generative AI models by grounding them in specific, private data, allowing them to provide accurate, context-aware answers.The technical workflow began with data ingestion and pre-processing. The team quickly developed Python scripts to perform “Extract, Transform, Load" (ETL) on the fly, scraping internal websites for statutes and parsing text from numerous internal files. This crucial step cleaned and structured the data for the next stage: indexing. Using Vertex AI Search, they created a robust, searchable vector index of the curated documents, which formed the definitive knowledge base for the generative model.With the data indexed, the RAG engine in Vertex AI could efficiently retrieve the most relevant document snippets in response to a query. This contextual information was then passed to Gemini via Vertex AI. This two-step process was critical, as it ensured the model’s responses were based solely on INDOT’s official documents, not on public internet data.Setting a new standard for government efficiencyWithin an intensive, week-long effort, the team delivered a functioning pilot that generated draft reports across nine INDOT divisions with an impressive 98% fidelity – a measure of how accurately the new reports reflected the information in the original source documents. This innovative approach saved an estimated 360 hours of manual effort, freeing agency staff from tedious data collection to focus on the high-value work of refining and validating the reports. The solution enabled INDOT to become the largest Indiana state agency to submit its government efficiency report on time.
The government efficiency report was a novel experience for many on our executive team, demonstrating firsthand the transformative potential of large language models like Gemini. This project didn’t just help us meet a critical deadline; it paved the way for broader executive support of AI initiatives that will ultimately enhance our ability to serve Indiana’s transportation needs.
Alison Grand
Deputy Commissioner and Chief Legal Counsel, Indiana Department of Transportation
The AI-generated report framework was so effective that it became the official template for 60 other state agencies, powerfully demonstrating a responsible use of AI and building significant trust in INDOT as a leader in statewide policy. By building a scalable, secure RAG system on Google Cloud, INDOT not only met its tight deadline but also created a reusable model for future innovation, accelerating its mission to better serve the people of Indiana.Join us at Google Public Sector SummitTo see Google’s latest AI innovations in action, and learn more about how Google Cloud technology is empowering state and local government agencies, register to attend the Google Public Sector Summit taking place on October 29 in Washington, D.C.
AI Summary and Description: Yes
Summary: This text outlines how the Indiana Department of Transportation (INDOT) leveraged AI on Google Cloud to efficiently process legacy data and generate a government efficiency report. The deployment of an AI solution using Retrieval-Augmented Generation (RAG) significantly streamlined their compliance efforts, demonstrating the transformative potential of AI in the public sector.
Detailed Description: The Indiana Department of Transportation (INDOT) faced a significant challenge in meeting an executive mandate for a government efficiency report, primarily due to the presence of extensive legacy data scattered across numerous unstructured documents. The following key points highlight the significance of this transformation through AI:
– **Context of the Challenge**:
– Public sector agencies face pressure for speed and agility amid legacy data.
– INDOT had a tight 30-day deadline to compile an efficiency report in response to an executive order.
– **Innovative AI Solution**:
– INDOT adopted AI capabilities using Google Cloud, focusing on the Retrieval-Augmented Generation (RAG) technique.
– RAG enables generative AI models to use specific, private data to deliver accurate, contextual answers.
– **Data Processing Workflow**:
– A week-long agile sprint led to the development of an AI workflow, starting with data ingestion and preprocessing.
– Python scripts were used for Extract, Transform, Load (ETL) processes, extracting and structuring data from various internal sources.
– **Implementation and Impact**:
– A robust searchable vector index was created, forming the knowledge base for the generative RAG model.
– The AI-driven pilot produced draft reports with 98% fidelity, dramatically reducing the expected 360 hours of manual analysis.
– The pilot further established INDOT as a leader in efficient government operations.
– **Broader Implications**:
– The successful implementation not only helped INDOT meet its deadline but also set a precedent for 60 other state agencies to adopt the AI-generated report framework.
– INDOT’s approach demonstrated responsible AI use and built public trust in their capabilities.
– **Future Initiatives**:
– The success of this initiative supports broader executive backing for future AI projects aimed at enhancing transportation services in Indiana.
Overall, INDOT’s experience showcases the potential of AI technologies in transforming public sector efficiency, paving the way for more innovative practices in government operations.