About project

The Challenge

Cassen’s reporting function is core to its business, delivering the scientific results on which clients subsequently base their actions. Each Safety Data Sheet and chemical report must be precise, comprehensive, and fully traceable to proven sources.

Before this project, an earlier GenAI pilot already showed promise but also revealed several structural bottlenecks:

• Throughput limitations — Every analysis run was capped at approximately 50 chemical markers, which constrained the complexity of cases that could be tested simultaneously.
• Manual regulatory testing — Manual verifications of environmental and safety compliance slowed down shipments.
• Multi-day turnaround times — Lengthy processing had the potential to delay client decision-making and response efforts.
• Massive per-report cost — The labor involved made expansion expensive and laborious.
• B2C barrier — The process wasn’t architected for a consumer-based model, where fast, simple reports are paramount to adoption.

Operationally, the lab was at maximum capacity, with growing client demand and increasingly sophisticated test cases putting pressure on its systems. Strategically, not being able to expand into consumer markets meant leaving money on the table.

Technology & Approach

Dedicatted designed and deployed a secure, AWS-hosted agentic AI solution that was appropriate for Cassen’s needs, using Amazon Bedrock with Amazon Nova models for top-level reasoning, summarizing, and accuracy in a compliant environment.

Solution components of note:

1. Multi-Agent Orchestration — Multiple AI agents perform in sequence to execute pattern detection, retrieval from past reports (through RAG), internet enrichment at the discretion of the user, and auto-assembly into a formatted output.
2. Compound→Pattern→Source Knowledge Base — A relational database that manages and stores compounds as relationships to identified patterns, risks, and source references, ensuring outputs remain valid and traceable.
3. Role-Based, Chat-First Interface — Chemists interact conversationally with the platform to guide and influence outputs. Administrators oversee the knowledge base, control internet access for compliance, and manage approved source lists.
4. Domain-Specific Fine-Tuning Loop — An ongoing improvement process involving chemist-reviewed output, which fine-tunes AI to Cassen’s workflows, vocabulary, and regulatory environment.
5. Scalable AWS Architecture — A modular setup that allows for future expansion, such as integration with lab instruments or the addition of specialized analytical models.

Impact on workflows:

• The system can now analyze over 150 markers per run—three times its previous capacity.
• Advanced features, such as VOC signature analysis, provide new forensic insights, like identifying potential fire origins.
• The platform produces comprehensive, compliant reports and streamlined, consumer-ready summaries from a single dataset, targeting various market segments without requiring duplicate effort.

User Experience & Adoption

Dedicatted set usability as a design driving principle from day one. Chemists were not accustomed to using AI tools in the first place, so the system had to feel natural from the outset. The chat interface enables users to interact with the platform like a trusted colleague—asking for analysis, requesting specific information, and facilitating real-time editing.

Training needs were minimal:

• The majority of chemists were self-assured after 1–2 sessions.
• Interactive refinement proved to be a significant boost in productivity, enabling chemists to see and modify the AI’s outputs instantly instead of starting from scratch. 
• Built-in citations removed the need for manual reference checks separately.

Adoption has been robust throughout the laboratory. Managers highly commended the role-based controls, which ensured compliance while maintaining workflow efficiency. Additionally, the direct updating of the knowledge base has facilitated the system’s adaptation to evolving research priorities and regulatory standards.

Outcome

Quantitative results:

Qualitative results:

• Reports are more comprehensive, fully sourced, and consistent, which boosts confidence among regulators and clients.
• The consumer-facing summary layer opens a new pathway for market expansion without impacting regulated B2B processes.
• The modular, AWS-based architecture enables quick adaptation to changing testing needs and regulatory updates.

For Dedicatted’s Cassen Laboratories, this initiative is about more than just an efficiency gain—it’s a foundation for growth. By scaling AI-driven workflows and providing the flexibility to serve both institutional and consumer markets, Cassen is now a Canadian leader in applying cutting-edge AI to environmental testing.