For a leading banking and financial services institution in the Asia-Pacific region, internal developer support had become a growing efficiency challenge. The organization served millions across retail, business and institutional segments, supported a developer ecosystem of 1,000+ professionals and continued to invest in digital modernization to strengthen service agility and operational resilience.
The existing support model was under strain in several ways:
- Nearly 90 engineers were engaged full-time in handling tickets tied to repository access, JIRA configurations, software build support and related requests.
- Tickets were created manually in systems such as ServiceNow and JIRA.
- Engineers had to categorize issues, search repositories and internal documents and communicate resolutions back to users.
- Past fixes were scattered across systems, making reuse difficult.
- Similar issues were often resolved repeatedly because prior knowledge was hard to trace.
The impact was clear. The model was slow to respond and expensive to scale. As volumes grew, response quality also became inconsistent. Resolution times ranged from 5 to 30 minutes depending on complexity. Simple configuration issues sat in queues, while more complex requests triggered back-and-forth across multiple teams. The deeper problem was not just time lost. It was fragmentation, weak traceability and a high people-cost dependency for routine service operations.
That raised a sharper question: could ticket handling be redesigned as an intelligent resolution system instead of a manual support chain?
Designing a Resolution Engine, Not Another Support Layer
Persistent identified an opportunity to redesign the operating model itself. Instead of adding another layer into an already fragmented environment, the solution centred on a centralized, AI-enabled ticket resolution platform that combined Generative AI with Agentic orchestration. Built as a proof of concept on AWS Bedrock, the platform was designed to automate the full path from query interpretation to resolution retrieval, backend access and ticket creation.
The platform came together through five connected components:
- Agentic AI orchestration layer (Crew.AI): Intelligent agents identified the right data source for each query, whether ServiceNow, JIRA, or internal document repositories.
- Generative AI integration (AWS Bedrock): Titan and Nova Pro models supported natural language understanding and contextual resolution generation, while automated ticket creation and resolution retrieval with minimal latency helped move the platform from understanding a query to acting on it.
- Vector retrieval layer (Amazon OpenSearch): Previous resolutions could be retrieved instantly through a vector database.
- Unified data integration layer: Secure middleware connected 7–8 backend systems, while Microsoft Teams supported query intake and PowerApps enabled alerting and automation.
- Continuous learning and governance: Every AI-generated resolution was auto-documented, improving traceability, accuracy and speed over time.
Deloitte reports that improving productivity and efficiency is the most commonly achieved benefit from enterprise AI adoption, with 66% of organizations reporting gains. This platform was built around exactly that premise: not AI as a layer of assistance, but AI embedded into the path of operational work.
What emerged was a resolution engine designed to read context, locate the right source and act with minimal latency. The next test was whether that orchestration could hold up in live service workflows.
Where Context, Retrieval and Action Come Together
When a service engineer raises a query through Microsoft Teams, the Agentic AI platform initiates an end-to-end workflow designed for real-time issue resolution.
The process unfolds in a clear sequence:
- Interpret the query: The orchestration layer reads intent, identifies entities and context and determines whether the issue relates to JIRA access, system errors, repository permissions, or another support need.
- Check for prior resolution: The system cross-references the query against historical resolutions stored in the Amazon OpenSearch vector database.
- Return an instant answer where possible: If a similar issue has already been solved, the previous resolution is retrieved within seconds and presented in natural language through Teams.
- Route dynamically when no match exists: If no suitable prior match is found, the system identifies the right backend environment, queries the relevant source, extracts diagnostic information and formulates a recommendation.
- Create and assign tickets automatically when needed: If deeper investigation is required, the platform generates a JIRA ticket pre-filled with issue summary, logs, previous attempts, and suggested next steps, then assigns it to the appropriate support team.
- Feed the learning loop: Every resolution, whether AI-generated or human-assisted, is documented and fed back into the knowledge base.
McKinsey estimates that generative AI could add $2.6 trillion to $4.4 trillion annually across 63 use cases. The significance here is not the headline number alone, but where that value shows up: in work that depends on language, knowledge retrieval, decision support and faster resolution of repetitive tasks. That is precisely the terrain this implementation was designed to improve.
What previously required multiple manual steps across teams now happens in under 10 seconds. The gain is not just speed. It is consistency, continuity and a support experience that becomes smarter with each interaction.
Business Impact
From Human Queues to System-Level Efficiency
The implementation delivered measurable operational gains and set a benchmark for AI-driven service transformation.
- ~90% faster ticket resolution, reducing average handling time from minutes to seconds
- ~90% reduction in manual workload, freeing engineers for strategic work
- Significant cost savings through reduced dependency on human-led processes
- Improved accuracy and consistency through automated documentation and continuous feedback learning
- Enhanced traceability and compliance, with all resolutions recorded automatically
- Stronger internal efficiency, enabling skilled resources to shift toward innovation and higher-value engineering priorities
Faster resolution was the visible gain. The more important shift was that the model was now ready to extend beyond internal service operations.
Building Beyond Internal Support
The proof of concept did more than improve service operations. Following the success of the POC, it established a framework the client is now expanding beyond internal service management.
The next phase is already taking shape:
- External-facing AI assistants: Planned to support real-time customer queries across retail and commercial banking
- Cross-functional expansion: Broader adoption envisioned across finance, marketing and operations
- Enterprise-wide scale: A longer-term goal to build a unified, AI-driven automation layer across business units
Why Persistent Was Chosen for the Next Phase
Persistent was selected for a combination of delivery speed and platform depth:
- Proven expertise in Agentic AI frameworks and multi-agent orchestration
- Strong AWS partnership and hands-on experience with Bedrock, OpenSearch and Titan models
- Ability to deliver in 8–10 weeks, significantly faster than industry averages
- Deep integration capability across complex enterprise environments
- Secure, compliant and scalable solution designed for future expansion
This engagement marked a major milestone in the client’s AI transformation journey. It showed how internal service management can move from reactive ticket administration to proactive, intelligent resolution. What once took minutes—or even hours—now happens in seconds. The outcome not only redefined operational efficiency but also laid the foundation for broader adoption of AI, blending automation, learning and agility into one operating model.
The POC solved a support problem, but the architecture opened an enterprise path.
Explore Agentic Resolution at Scale. See What Your Support Model Is Costing You. Talk with Persistent.
