Key highlights
110,000+ Copilot credits
Usage identified in a single month, prompting optimization to balance response quality with sustainable AI consumption.
70,000+ rows
Large customer datasets supported through a redesigned batch-processing approach, improving throughput and lowering Copilot-credit usage per record.
3 AI solutions
Complementary AI solutions address engineering validation, customer data verification, and specialized UV technical support.
Summary
Trojan Technologies advanced its adoption of agentic AI by deploying three practical solutions built on Microsoft Copilot Studio. The engagement improves access to trusted engineering knowledge, streamlines large-scale customer data verification, and provides sales and service teams with more consistent access to specialized UV guidance.
Challenges
Overcoming fragmented knowledge, manual data verification, and dependence on individual experts
Engineering validation information, including test reports, DVP&Rs, FMEAs, PNRs, requirements, and historical failure data, was distributed across multiple repositories, making searches and reconciliations time-consuming. Customer site and address validation across large datasets was also highly manual and error-prone. Meanwhile, sales and service teams relied heavily on individual UV subject-matter experts for technical guidance. Scaling AI introduced another challenge: managing response scope, processing throughput, Copilot-credit consumption, and approval controls.
Solution
Building three governed AI solutions around practical operational needs
Zensar developed three complementary AI solutions using Microsoft Copilot Studio.
Project VISTA: A Copilot Studio engineering assistant that searches approved information and supports risk lookup, report summarization, design-change reviews, requirements traceability, verification-gap identification, and cited answers.
Customer site verification: An end-to-end AI application that validates customer information against web sources and stores results in SQL Server, supported by batch tracking, error logging, resume capability, and optimized multi-record processing.
UV app specialist-agent: A conversational Microsoft Teams experience that centralizes specialized UV knowledge, giving sales and service teams more consistent access to guidance on UV dose, UV-AOP, disinfection, and related topics.
The work also addressed generative AI operational costs. After VISTA usage exceeded 110,000 Copilot credits in a single month, despite a 50,000-credit tenant pool, the team optimized the response instructions with leadership approval. The change reduced default response size while preserving the agent’s structured, evidence-based output. This reflects a deliberate approach to balancing user value, quality, and sustainable consumption.
Solution enablers
Microsoft Copilot Studio: Supports the development of specialized agents for engineering validation, customer data verification, and access to UV technical knowledge.
Microsoft Teams: Provides the conversational interface that enables sales and service teams to access centralized UV expertise.
Direct Line REST API: Connects the custom ASP.NET Core customer-verification application with the Copilot Studio agent.
SQL Server: Provides persistence for customer-verification results, supported by error logging, batch tracking, and resumable processing.
Business outcomes
Making enterprise AI more accessible, scalable, traceable, and sustainable across operational workflows
By leveraging Microsoft Copilot Studio across engineering, customer operations, sales, and service functions, the engagement demonstrated how agentic AI can address targeted business challenges while enabling broader enterprise AI adoption. The solutions improved access to trusted information, scaled data verification, and streamlined access to specialized UV knowledge through evidence-based responses and system integration. Supported by consumption controls, optimization strategies, and governance frameworks, the approach delivered operational efficiency while ensuring sustainable, cost-effective, and responsible AI adoption.
Faster access to engineering evidence
Reduced the effort required to locate and reconcile engineering information and improved traceability for validation and product development.
More scalable customer-data verification
Reduced manual effort and supported better data quality, with multi-record batch processing improving scalability across datasets.
Broader access to UV expertise
Reduced reliance on individual subject-matter experts by giving sales and service teams access to specialized data.
More sustainable AI consumption
Optimized response instructions after VISTA consumption exceeded 110,000 Copilot credits in a month against a 50,000-credit tenant pool.
