AWS AgentCore and Joulica Agentic Analytics

The next year is set to see the widespread adoption of Agentic AI, where agents collaborate, plan tasks, call tools, and work autonomously. Nowhere is this shift more transformative than in Customer Experience (CX), where autonomous agents will shape customer journeys moment by moment. While Agentic is accelerating, most organisations are still operating with fragmented insights. Real-time visibility in one system, historical reporting in another, and no unified way to understand how AI-driven decisions impact business outcomes. Joulica brings these worlds together with real-time + historical Agentic Analytics, delivering a continuous, coherent picture of agent behavior, governance, and performance.

AWS AgentCore is one of the most important steps forward in terms of delivering a production grade framework that lets teams build multi agent applications at scale with structure and repeatability. But as organizations move from prototypes to real deployments, a new challenge is emerging:

How do we monitor, measure, and improve what these AI agents are actually doing?

Joulica delivers the realtime and historical observability layer to govern Agentic applications. This ensures that every agentic system is delivering ROI and is aligned with business goals.

Why Agentic AI Needs Analytics

Agentic AI behaves very differently from traditional IVRs, chatbots, and rule-based routing engines. In CX, a single customer request may activate multiple specialised agents — for intent understanding, information retrieval, task execution, summarisation, routing, and agent assist — each making autonomous decisions that impact the customer journey.

Without analytics, CX leaders and operations teams have significant blind spots:

• Fragmented analytics: Customer-journey insights are split across channels and platforms, while real-time troubleshooting sits in one system and historical optimisation lives in another. This makes it impossible to understand how agentic decisions truly impact CX outcomes.
• Containment drops: Why did the system escalate when it should have resolved the query?
• Misrouting: Which agent or reasoning step misinterpreted customer intent?
• Agent-assist accuracy issues: Why did the assistant surface the wrong suggestion or incomplete context?
• Automation inconsistency: Which tools, prompts, or models are driving escalations or unnecessary handovers?
• Debugging gaps: Why did the bot suddenly enter clarification loops or generate irrelevant steps?

Joulica’s Agentic Analytics links every action an AI agent takes to actual CX outcomes: containment, AHT, CSAT, resolution, sentiment, deflection, and cost-to-serve.

The example below is an example of some Joulica realtime KPIs for a multi-agent system built using AgentCore.

This allows organisations to understand not just what the AI did, but why, and how that decision affected the customer and the business.

Joulica unifies real-time observability and historical journey intelligence, giving teams a single, continuous view of agent performance, tool use, and business outcomes. This enables safer, smarter, and more scalable contact-center automation.

A Practical Example: An AgentCore Powered Travel Planning Application

To illustrate how AgentCore and Joulica complement each other, consider a travel-planning agentic application. While the scenario is leisure travel, the underlying mechanics mirror what happens when Agentic AI is deployed in any industry vertical: multiple AI agents collaborating, interpreting intent, calling tools, and delivering outcomes that directly influence customer satisfaction, revenue and costs.

1. A Simple Customer Request

"Plan a 3-day trip to Barcelona on a €1,500 budget."

2. AgentCore Orchestrates the Workflow

AgentCore routes the request to:

• Trip Planner Agent – interprets the customer request
• Flight Agent – responsible for flight searches and bookings
• Activity Agent – handles activity planning and bookings
• Car Agent – manages car rentals
• Hotel Agent – responsible for hotel reservations

3. Agents Call Tools to Complete the Task

Agents interact with domain tools such as:

1. Search and book flights
2. Search and book hotels
3. Search and book activities
4. Search and book car rentals

Each tool decision affects the overall outcome, whether automation succeeds, misroutes, escalates, or even frustrates the customer.

The Joulica dashboard below shows how the Trip Agent utilizes the sub agents, as well as showing the paths that lead to escalations that need the assistance of a human.

Where Joulica Adds Value

The success of any agentic workflow depends on the decisions agents make at each step. These decisions not only effect the course of an individual customer interaction but also can affect subsequent interactions and the entire customer journey.

With Joulica, organisations gain:

• Higher automation and containment by identifying the decision paths that help or hinder successful resolution. 
• Improved accuracy and reduced handling time through visibility into which steps accelerate outcomes versus those that introduce friction or unnecessary loops.
• More reliable and cost-efficient tool usage by understanding which tools deliver results and where inefficiencies or failure patterns exist.
• Safer, more explainable AI behaviour by revealing why agents select or reject certain options — critical for governance, trust, and regulatory assurance.
• A continuous improvement loop by correlating real-time behaviour with historical patterns to detect anomalies, improve prompts, refine policies, and optimise performance over time.

The paths that an Agentic application takes can quickly become complex, as illustrated below.

Why AgentCore and Joulica Is a Powerful Combination 

AgentCore gives organisations an enterprise-grade framework for building agentic applications, but the moment these systems touch real customer journeys, analytics becomes the deciding factor for whether they can operate safely, efficiently, and at scale. This is where Joulica transforms AgentCore from a powerful orchestration engine into an enterprise-ready CX platform.

AgentCore empowers organisations to

• Design and orchestrate agentic workflows with consistency and governance
• Combine intent understanding, reasoning, and tool calling through Amazon Bedrock
• Rapidly create complex automations that replace static IVRs and rigid decision trees

Joulica turns these workflows into measurable business outcomes by

• Linking agent behaviour directly to CX outcomes — containment, AHT, CSAT, deflection, resolution, agent-assist accuracy, and cost-to-serve
• Creating a continuous optimisation loop that improves accuracy, reduces escalations, and increases automation rates over time
• Delivering governance and explainability so enterprises can deploy agentic automation with confidence
• Providing real-time and historical visibility into every agent decision, tool call, and interaction
• Supporting analytics across all the platforms underpinning the customer journey, allowing organisations to measure performance across CCaaS platforms, AgentCore, CRMs and the myriad of other applications that are found in a typical enterprise.

When you pair the orchestration and flexibility of AgentCore with the observability and outcome-driven analytics of Joulica, organisations gain:

• Faster time-to-production for agentic CX solutions
• Better operational performance through detection of anomalous behavior, misrouting, and tool failures
• Stronger governance and compliance with transparent AI decision trails
• Higher automation, lower cost, and improved customer experience driven by insights that traditional analytics cannot provide

Summary

Agentic AI is moving rapidly from experimentation to production in real CX use cases. But building an agentic application is only half the battle, with the real challenge coming on day two, with the need to understand how every agent decision impacts containment, routing accuracy, agent-assist quality, customer satisfaction, and cost-to-serve.

By unifying real-time and historical intelligence, Joulica gives organisations the clarity to:

• trust the decisions their AI agents are making,
• identify and fix failure points before they impact customers,
• continuously improve automation rates and operational efficiency, and
• ensure agentic systems stay aligned with policies, compliance, and business goals.

For enterprises adopting AgentCore, Joulica becomes the key capability that turns powerful agentic workflows into safe, scalable, and value-generating CX automation.