The Story at a Glance
A mid-sized North American SaaS company serving the public sector sought to reduce project abandonment and expand into private markets. Users often dropped off midway due to the platform’s manual, time-intensive design process, and the small internal team lacked bandwidth to innovate or scale.
Partnering with CleanSlate Technology Group, the company implemented a custom AI-powered solution on AWS—leveraging technologies like Amazon SageMaker, Amazon Rekognition, and custom CNNs to automate image processing, project grouping, and design workflows.
In just six weeks, a Minimum Viable Product (MVP) launched with over 90% project completion rates, reduced internal workload, and enabled market expansion. A second phase is now underway, focused on enhanced security and advanced AI features.
This success underscores CleanSlate’s ability to drive innovation, streamline workflows, and support long-term growth through intelligent automation.
What Had to Change
Although users were excited about the projects they started, many became fatigued with the manual effort and only completed 60–70% of the process. The existing process produced great outcomes when finished, but required too many manual and intensive steps to complete a project.
Additionally, the SaaS company had a small team that was struggling to keep up with the demands of helping public sector clients finish their projects. The company did not have time to explore new market opportunities outside the public sector. The company needed to adopt an automation-first solution that would:
• Increase project completion rates
• Eliminate manual bottlenecks
• Support a high volume of users and projects
• Maintain professional quality
How CleanSlate Responded
Working closely with the client’s product and technical teams, CleanSlate designed a multi-phase, cloud-native AI solution that reduced friction in the design process and delivered end-to-end automation.
Phase 1: AI-Based Image Processing and Enhancements
• Applied convolutional neural networks (CNNs)to identify important design elements
• Built custom deep learning modelsto auto-enhance images based on industry standards.
• Delivered automated high-quality image enhancements consistently with minimal user input
Phase 2: Automated Project Groups and Design Selection
• Used Amazon Rekognition for object and scene recognition
• Grouped projects by context: event type, similarity, or time
• Recommended templates and designs for many customer groups to each group
• Empowered users with adjustable design preferences
Phase 3: Operationalizing for Scale
• Leveraged Amazon SageMaker, Lambda, and serverless architecture for scalability
• Combined traditional machine learning models (e.g., k-nearest neighbors) with proprietary business logic
• Applied design quality scoring and template-matching to ensure consistency
• Designed for elastic performance with minimal manual oversight
Technologies and Services Used
• Amazon SageMaker
• Amazon Rekognition
• Lambda (serverless compute)
• Custom deep learning models
• Machine learning-based image quality assessment
• Infrastructure-as-Code (IaC) for repeatable deployment
What Changed
• 90%+ completion rate in AI-assisted projects
• Increased revenue by enabling the company to expand to private sector industries
• Improved operational efficiency by reducing manual workload
• Accelerated speed to market for new offerings
• Delivered a Minimum Viable Product (MVP) in just 6 weeks
Learn more about the Project Lead: Nathan Liston
Ready to Move Forward?
Start with a structured assessment of your environment, priorities, and constraints. CleanSlate can help turn that into a practical path forward.