Our Copilot and Article Writer AI Agents can now understand files and links provided to them as part of the context data. 

This means you can upload files or copy and paste them into the composer, and when the AI Agent processes your question or request, it will also review any images or files that you provide. That data from these additional content sources data will be added to the context of the session. Additionally, if you provide a URL, the agent will automatically review the content on that URL (if it is public) and use that data as part of its context as well.

This new capability allows you to do more with our AI Agents. A few examples include:

1. Give an agent an image of an system error, and the agent can use that error to search your knowledge base for the answer.
2. Include a PDF of a customer contract and ask the agent to review the contract for specific terms. 
3. Give an a URL of an external knowledge base article, and the agent can gather all the information and use that to answer your question.

We hope this enhancement will make your experience with our AI Agents even more powerful and efficient. We look forward to seeing how you use this new feature to streamline your tasks and improve your workflow.

Release Date: 04/21/24: 

CX Manager

• Added the ability to use custom fields in Email Templates. Previously only standard fields were available.
• Fixed a bug where the Email from Chat Room email template was not respecting the filters for Owner Team on a Case. It will now use the proper email template based on the Owner Team of the Case.
• Fixed an issue where cloned Cases from inbound emails to closed Cases were sometimes being assigned to the incorrect Owner Team.
• Fixed an issue with outbound emails sometimes not appearing in the message history. We will continue to harden this specific flow.
• Improved the loading time and process for email history on Cases.

Knowledge

• Migrated all Knowledge Libraries from the default theme to the modern theme for bug fixing, performance improvements, and design updates.
• Fixed a bug related to accessing internal articles for non content editors when the Article was restricted to certain Teams.

Journeys

• Fixed a bug where Tasks in multiple Parallel Steps would not all get properly created as the Journey progressed.
• Fixed a bug where Automation loading spinners on the Journey Timeline sometimes would not disappear when the automation completed.

Forms

• Fixed a bug where signatures were not properly saving and exporting on Form exports. 

APIs

• Updated API/v3 to allow cutomers/location/put and customers/user/put APIs to accept requests without customer_id when id is provided. This will help integrations where a Location moves to a new Customer and the original Customer ID is no longer valid.
• Added SAML SSO integrations for the new CXME application along improved SSO configuration workflows.
• Reduced the latency when redirecting a login after successful SSO login
• Improved the resiliency of our eventing system.

Data

• Fixed an issue where email transcripts were not always being included in AI-generated Case Summaries.

Reporting & Dashboards

• We are in the process of launching our new Journey and Task analytics data and pre-built dashboards. We will be rolling these out over the comings weeks.

To move AI from a novelty to a core banking utility, you need more than a prompt - you need an operational foundation to help govern this intelligence at all times. That is why we built  OvationCXM’s AI Studio: the unified command center designed to create, configure, and continuously improve high-performance AI agents within a secure, governed framework.

By layering AI directly into structured, governed workflows, we ensure that every automated step is anchored in a real-world operational context, allowing work to move predictably across your entire ecosystem. Without a central point of control, AI risks becoming another siloed application that increases operational complexity rather than reducing it - a significant risk in a high-stakes, regulated environment.

OvationCXM’s AI Studio provides the transparency needed to audit AI logic, the laboratory to safely test new models, prompts and rules in a sandbox, and the governance to ensure every digital assistant remains a compliant, audited extension of your institution’s policies.  

In this post, we discuss how we take an AI agent from a concept to a high-performing operational asset in four distinct phases. 

Phase 1: Build

In OvationCXM’s AI Studio, an agent is not a generic chatbot; it is a specialized digital employee with a specific role, voice, and set of boundaries. The process begins by defining the Agent’s functional purpose from one of our four specialized types. For example, you might deploy a Summarizer agent to instantly distill years of fragmented email threads into a single, actionable case history, or OvationCXM’s internal-facing agent (copilot) to work alongside your staff, suggesting the next best step based on live orchestration data. For direct engagement, the external-facing conversational agent handles real-time customer inquiries, while the article writer/editor ensures your knowledge base stays current by drafting and refining baseline documentation based on successfully resolved cases.

Once the agent’s purpose is defined, it’s time to select the best model from a library of over 300 industry-leading models. This flexibility allows you to assign a high-reasoning model for complex analysis while using a faster, lightweight model for high-volume data routing. Crucially, you can approve a subset of models for your organization or even bring your own keys (BYOK), enabling you to use your own enterprise accounts with the model provider of your choice. 

To ensure consistency, you can also configure your global branding with information about your specific organization, including your brand voice and desired tone. By anchoring agents in these specific Global Branding configurations, you ensure that every agent speaks with the professional tone and consistency your brand demands, regardless of its function.

Phase 2: Execute

In this second phase, OvationCXM’s AI Studio bridges the gap between conversation and execution. In addition to the high-quality responses and content you receive from agents, they can also take direct action on your behalf through integrated workflows and tools. Because banking tasks are rarely linear, you define advanced workflows that guide the agent through multi-step scenarios - such as verifying a customer’s identity across three different systems before it ever suggests a journey change. 

When the agent needs to interact with your systems, it uses a set of deterministic tools, including secure API calls that allow the agent to perform key tasks. These tasks can include fetching a real-time account balance from a core system or updating a record in your CRM. By using these tools, the agent doesn't have to guess. Instead, it acts with precision, using the authenticated data already present in your ecosystem.

Phase 3: Observe

With OvationCXM’s AI Studio, every interaction is captured in a detailed trace, providing an auditable, step-by-step view of the agent's reasoning, the context used, and the final response. If a logic gap appears, you can see exactly where the agent diverged from the expected path.

This transparency is bolstered by a continuous feedback loop. Users can rate AI interactions, and any poor rating is routed to a review queue immediately. This allows your team to take immediate corrective action - whether that means adding a new rule, updating a knowledge article, or assigning a task to a human team member to step in. The point is that you aren't just watching the AI - you are managing it in real-time.

Phase 4: Improve

The final phase is about evolution without risk. In traditional software, you use unit tests to ensure code doesn't break; in AI Studio, you use Evaluations. First, you define your Data Sets based on questions and verified answers that the agent must handle correctly.

Before any change to a prompt or a rule goes live, you run an Experiment. AI Studio utilizes an LLM-as-a-Judge to automatically grade the agent’s new responses against your pre-defined criteria. This allows you to iterate and improve your agents with total confidence, knowing that your quality standards are being both met and proven before the AI ever interacts with a customer or a sensitive workflow.

The Bottom Line

AI Studio isn't just a builder - it’s a comprehensive governance layer that bridges the gap between modern intelligence and legacy core systems.  It provides the infrastructure to ensure that as your bank scales its capabilities, it does so with the observability, safety, and precision that the industry demands. By prioritizing structured orchestration first and on-demand intelligence, OvationCXM helps you transform AI from a speculative novelty into a permanent competitive advantage.

Interested in seeing a quick demo? Contact us today.