Home
Case Studies
Arizona DCS
Arizona Department of Child Safety (DCS) simplifies processes with genAI
Manual-based processes hinder critical decision making
Arizona Department of Child Safety (DCS) manages complex processes as part of its mission to protect children, reunify families, ensure safe living environments and prevent unnecessary trauma.
In the past, the department struggled to manage complex casework and keep up with constantly evolving policies. Furthermore, manual processes slowed decision-making, leading to inefficiencies and delays.
Consistency and collaboration in design and technology innovation
The delicate nature of Arizona DCS’s work required a thoughtful approach. To make sure process changes were handled with care and consideration, Arizona DCS partnered with IBM Consulting® and applied a phased approach using Microsoft® Azure & Microsoft genAI capabilities. This combination enabled a strategic, innovative and outcome-focused transformation, specifically tailored to meet the distinctive demands of DCS operations. IBM Consulting delivery teams also used IBM Consulting Advantage for Cloud Transformation (powered by IBM watsonx™) for key program support materials and internal workflows, which boosted accuracy and lead to faster, better-informed decisions.
The Arizona DCS transformation & modernization, resulted in improved central registry classification, virtual assistant support for policy changes and streamlined development processes that reduce workload and increased productivity. All this allowed DCS professionals to spend more time with families and community members instead of completing administrative tasks.
“IBM Consulting team helped us implement key Microsoft gen AI solutions that have significantly improved the efficiency of caseworkers, enabling them to focus on what really matters—helping families,” says Frank Sweeney, Chief Information Officer of Arizona DCS.
Central registry classification and improved categorization
Arizona DCS faced a challenge in classifying historical allegation findings within their central registry, a database of reports on children’s neglect and abuse. The current system relied on manually going through plain text, which made it difficult to efficiently manage and analyze the vast amount of data. Manual classification led to delays in decision-making and the lack of subclassification made it challenging to identify specific allegations and their corresponding seriousness. The introduction of changes in subclassification definitions through statute, court cases, policy and procedures further complicated the overall classification process.
In Phase 1 of the transformation, the combined Arizona DCS and IBM Consulting team focused on an enhanced and efficient classification utilizing Microsoft OpenAI capabilities to develop a system for classification of allegations into 4 main categories.
In Phase 2, the team worked on improving categorization using Microsoft Azure OpenAI tools to generate subcategories for classified allegations based on narratives and subclassification definitions. This success helped to enhance Arizona DCS categorization capability to better align with the seriousness of offenses. Additionally, this automation helps fuel human decision-making at DCS.
Arizonia DCS is experiencing several benefits to the central registry classification and categorization:
- New DCS classification accuracy helps improve retention period decisions and enables informed decision-making regarding retention periods in the Central Registry.
- Seriousness of offenses is now an additional determining factor in retention period decisions.
- Enhanced categorization enables data-driven decision-making with improved accuracy, for better child and family outcomes.
- Cost avoidance and productivity gains are achieved via correct classification with minimized manual intervention and manual processing.
Virtual assistant for policy changes and information access
Arizona DCS faced issues in keeping the caseworkers current with the policy changes and procedural updates on the Program Policy portal, a tool created to provide information and answer questions about policies. The complexity of policy documents and other accessibility issues led to inefficiencies in information retrieval and comprehension, resulting in lost productivity and challenging compliance efforts.
Working together with IBM Consulting and using Microsoft Azure gen AI, the team delivered a solution to address the challenges faced by Arizona DCS professionals; an AI-powered virtual assistant was created to simplify the process of accessing and understanding policy documents.
The new virtual assistant features include a user-friendly interface that allows Arizona DCS professionals to navigate complex policy documents and provide a summary of policy documents. All summaries are sorted based on confidence level and reliability to ensure accurate information, along with links to the information source. The virtual assistant saves time for the user in searching for policy information and helps ensure clear understanding to be compliant.
This gen AI-powered virtual assistant solution is now live and providing these benefits:
- Enhanced user experience via intuitive interface offers seamless navigation and access to policy documents for DCS professionals.
- Swift search and retrieval facilitate quick, precise search and retrieval of policy information in response to user inquiries.
- Simplified policy language for clear explanations helps minimize the likelihood of policy violations.
- Increased productivity and reduced review time for policy alterations
This solution has also improved user experience, simplified policy language and reduced the risk of policy violations.
DevOps integration for enhanced development and productivity
Arizona DCS were challenged to handle roadblocks in their process as product owners and business analysts were managing additional responsibilities supporting case workers, this led to decreased productivity. The time-consuming process for product owners and business analysts to create user stories and acceptance criteria also delayed task development, impacted project timelines and decreased productivity across teams.
After detailed design and planning, the team worked to integrate TachyonGPT, a gen AI tool provided by Neudesic®, an IBM company, with an Azure DevOps Board to generate and refine features, user stories, acceptance criteria and test cases. The team utilized Gherkin syntax for user stories and acceptance criteria, to gain consistency and clarity. Given this, Arizona DCS was able to establish a consistent way of writing Azure DevOps work items and reduce errors and increasing productivity, which ultimately helped enable product owners and business analysts to focus on high-priority tasks and reduce their workload.
Additional benefits for Arizona DCS include:
- Productivity gains with a 40% increase allow more efficiency.
- Accelerated development timelines enable faster delivery of projects and features.
- Gen AI-driven development empowers the team without requiring extensive technical expertise and effectively managing risks.
- TachyonGPT integration with Azure DevOps streamlines development processes, reducing workload, and proving the potential of AI to transform the way teams work and deliver projects together.
Processing thousands of documents related to casework each month, was a significant challenge for Arizona DCS. The existing process was time-consuming and inefficient. As an example, documents could only be associated with one case at a time and associating documents required several steps and clicks, increasing overall processing time.
To solve this, the team implemented a mechanism to read documents that automatically prefills forms using the AI Builder document processing capabilities of Microsoft Power Platform. Then, the focus evolved to training the AI Model for better accuracy, which is inflight now, with the team currently training the AI Builder document processing model on approximately 200 different types of documents. This automation helps fuel the team’s decision-making at Arizona DCS.
“These improvements have translated into faster, more reliable case management, better stakeholder satisfaction, and substantial cost savings.”-Sweeney concluded.
With the power of AI Builder, the team has significantly improved the efficiency of document processing for achieving multiple benefits:
- Reduced processing time frees up resources for more critical tasks.
- Automated uploading of multiple documents reduces document upload time by 500%.
- Streamlined document processing saves time and resources for providers and DCS staff.
Microsoft Copilot capabilities are also under consideration by Arizona DCS for future genAI transformation program use cases, as the overall program continues to help address the needs of children and families across the state.
The Arizona Department of Child Safety is charged with child protection for the state. They work to reunify families, ensure safe living environments for the children, and prevent unnecessary trauma.
Legal
© Copyright IBM Corporation. IBM, the IBM logo, IBM Consulting, Neudesic, and watonsx are trademarks or registered trademarks of IBM Corp., in the U.S. and/or other countries. This document is current as of the initial date of publication and may be changed by IBM at any time. Not all offerings are available in every country in which IBM operates.
Client examples are presented as illustrations of how those clients have used IBM products and the results they may have achieved. Actual performance, cost, savings or other results in other operating environments may vary.
Microsoft, and Microsoft Azure are trademarks of Microsoft Corporation in the United States, other countries, or both.