Policy in Practice accelerates AI innovation on AWS with SBS

Policy in Practice runs an annual AWS hackathon designed to help its developers explore emerging AI and machine learning services, prototype innovative solutions, and drive novel solutions benefiting its councils and citizens. However, the manual process that was required to provision secure, temporary AWS sandbox environments had become a significant bottleneck. Preparing AWS sandbox accounts for the previous hackathon had taken approximately 3 weeks of manual configuration work, and those accounts supported only half the number of participants that the team hoped to accommodate in future events.

This time-intensive setup created a heavy dependency on senior engineers, pulling valuable technical resources away from strategic initiatives. The organization lacked automation to limit spend, control permissions, or manage AWS account lifecycles, introducing both financial and security concerns. As participation grew, this approach made it increasingly difficult to maintain the security and compliance standards that are essential for a public sector organization. “We wanted a way to give our developers completely free access within safe boundaries—unrestricted freedom in a locked box,” says Joseph Hollingworth, chief technical officer at Policy in Practice. “The challenge was finding a solution that could deliver both innovation velocity and governance rigor.”

SBS, an AWS Partner whose software automates sandbox environment provisioning, developed a production-ready solution that was specifically designed for organizations like Policy in Practice. The engagement between Policy in Practice, SBS, and AWS began with discovery sessions to align on business and technical requirements and delivery timelines. Through this collaboration, the teams identified Sandbox Studio—which is available through AWS Marketplace—as the orchestration layer to standardize, govern, and rapidly deploy secure, policy-compliant AWS sandbox environments at scale for the hackathon. Sandbox Studio was deployed directly within Policy in Practice’s AWS Organization, enabling automated, template-based AWS sandbox account provisioning for five distinct AI use cases. Each team received preconfigured sandbox accounts with access to Amazon SageMaker AI to build, train, and deploy AI models as well as Amazon Bedrock to build generative AI applications and agents. With infrastructure in place from the start, developers could immediately begin prototyping.

SBS’s automated lifecycle management simplified the entire process from account creation through decommissioning. Teams could collaborate through a shared sandbox account feature, while each team could use automated spend tracking to stay within its $500 budget allocation. Policy in Practice configured Amazon Simple Storage Service (Amazon S3)—an object storage service—bucket replication for safe demo data access and established Amazon SageMaker workspace access for all participants, with developers able to access the AWS console, command line interface, and sandbox accounts through the Sandbox Studio web interface. “By using Sandbox Studio from SBS, we can focus on building, and not monitoring, AWS accounts,” says Jonathan Moss, senior infrastructure engineer at Policy in Practice.

The rapid progress on this project was facilitated by the AWS Think Big for Small Business Program (AWS TBSB Program), which offers small and/or minority-owned public sector organizations access to business, technical, and marketing support.

The new approach delivered immediate results for Policy in Practice’s hackathon. Environment setup time plummeted from approximately 3 weeks to under 12 hours for double the number of participants. Policy in Practice estimates that future events will require less than 3 hours of preparation, which is a reduction of 95 percent. Every team stayed within budget, and developers experimented freely with zero governance exceptions or production risk. The initiative also saved 60–80 hours of senior engineer time per event. During the hackathon, teams produced four AI prototypes, with two advancing to production. The Natural-Language to SQL Query Engine converts plain-language queries into SQL queries across Policy in Practice’s data warehouse, reducing query turnaround time from days to seconds. The Auto-Parser Updater detects and fixes schema changes in thousands of council data files daily, significantly reducing the workload for three full-time apprentices who handle up to 1,050 files per day.

Building on the hackathon’s success, Policy in Practice is embedding Sandbox Studio into its everyday AWS development workflow, transforming it from an event-specific tool into a platform for continuous innovation. As a result, there has been an 80 percent increase in new product deployment frequency on AWS. “The hackathon wasn’t just about building prototypes; it opened a new way for us to deliver innovation on AWS. We now see a direct path from experimentation to production, with measurable impact for our clients,” says Genevieve Orford, chief product officer at Policy in Practice. “We can experiment across Amazon SageMaker AI, Amazon Bedrock, and other AWS AI tools without friction. That speed is transforming how we think about product development and model deployment.”