# Aisera Product Release notes - Mar 25, 2026

## Features

The following features have been added to the Aisera Gen AI Platform in this release.

### Flows

The following features have been added to **Flows** in this release. 

#### Hyperflow Test and Debug Panel

The Hyperflow Details screen now includes a built-in test and debug panel. You can interact with your Hyperflow in a webchat interface, allowing you to validate tools, workflows, and inspecting execution details before publishing with impacting end users.

**Testing Modes**

You can test in two modes, selected at the start of each session:

<table><thead><tr><th width="119.60003662109375">Mode</th><th>Details</th></tr></thead><tbody><tr><td>Mode</td><td>Details</td></tr><tr><td>Draft</td><td>Test unpublished changes, including new tools and prompts.</td></tr><tr><td>Published</td><td>Interact with the live version exactly as end users do. Ideal for validation and issue reproduction.</td></tr></tbody></table>

This allows you to test in **Draft** first, then verify behavior in Published before releasing changes.

**To start a test session:**

1. In the Aisera Admin UI navigate to **AI Automation > Hyperflows**.
2. Select the Hyperflow you want to test.
3. Sele**ct Test** to open the **Test Hyperflow** window.
4. Enter a **User Query** and select a **Status**.
   1. Testing a Draft will only be available if you are currently editing a draft of the Hyperflow, otherwise only the Published status will be available for testing.
5. Select **OK** to begin interacting with the Hyperflow. A webchat panel will open to continue interacting with the Hyperflow.

**The Hyperflow responds in the same format as seen by end users, including:**

* **Text Responses:** Natural language answers, confirmations, and explanations.
* **Tool Results:** Outputs from Hyperflow tools such as ticket IDs, account details, and so forth.
* **Clarification Questions:** Prompts for additional information when required.
* **Out-of-scope messages:** Clear messages when a request falls outside the agent’s capabilities.

**Conversation Status**

The test panel shows the current status of the interaction:

<table><thead><tr><th width="147.5999755859375">Status</th><th>Details</th></tr></thead><tbody><tr><td>In Progress</td><td>The Hyperflow is processing the request or waiting for additional input.</td></tr><tr><td>Completed</td><td>The Hyperflow has successfully completed processing the request.</td></tr><tr><td>Failed</td><td>Something went wrong during execution.</td></tr></tbody></table>

To start a new session, select Stop Test to close the current session, then select Test again.

#### Debug Information for Generate RAG Answer action in Workflows.

The **Generate RAG Answer** action node in Workflows now returns a full set of RAG debug information, enabling detailed inspection of how and why documents were selected during retrieval and generation.

The debug information is accessible in two places:

* In **AI Lens**, when testing a workflow that contains the Generate RAG Answer action node.
* In **AI Workflow Studio**, by opening a workflow that contains the Generate RAG Answer action node and selecting **Test**. The debug information appears under **Nodes** > **Generate RAG Answer**.

The Generate RAG Answer action node can be added to a workflow by navigating to **AI Automation** > **AI Workflow Studio**, opening a new or existing workflow, and selecting an **Action** node.

### Conversational AI

The following features have been added to Conversational AI in this release.

#### App Instructions for ConvAI 2.0

Conversational AI 2.0 applications now support App Instructions, a new feature allowing you to configure exception-based routing rules that override the standard fulfillment pipeline and redirect specific user queries directly to a designated workflow. This feature is best suited for high priority cases such as financial fraud reports, medical emergencies, or mental health disclosures, where deterministic routing is required regardless of the user’s phrasing.

App Instructions support a maximum of three exception cases. It is not advisable to add more than three overrides.

The behaviors are nondeterministic in the case a user inputs a multi-intent query, and overrides should be thoroughly tested to ensure they achieve the desired effect.

App Instructions uses generic natural language matching and cannot reference session, channel, or user variables.

**To configure App Instructions for a Conversational AI 2.0 application:**

1. In the Aisera Admin UI navigate to **LLMs > Prompt Studio**.
2. Locate and select the `convai_app_instructions_prompt_gpt4o_chat_completions` prompt.
3. Select **Tune**. This will override the global prompt, creating an application-level prompt, allowing you to edit the prompt for the application you are configuring.
4. In the prompt, add an instruction for each exception case using the following format.<br>

   <div data-gb-custom-block data-tag="hint" data-style="info" class="hint hint-info"><p>The title must exactly match the name of the workflow you want to trigger. The instruction is a plain-language description of the user query cases that should route to that workflow.</p></div>

   <pre data-overflow="wrap"><code>[Action]
   {
     "title": "&#x3C;workflow_name>",
     "instruction": "&#x3C;description of when this action should trigger>",
   }
   </code></pre>
5. Delete the sample actions included in the default prompt before saving.
6. Select **Save**.

When a user query matches an instruction, the specified workflow runs and the standard fulfillment pipeline is bypassed. If no instruction matches, the pipeline continues as normal.

#### GPT 5.4 Available in Prompt Studio

You may now select **Azure Open AI GPT 5.4** as the model used in prompt studio. This can be observed by navigating to **LLMs > Prompt Studio** and creating or editing an existing prompt.

### Connectors

The following features have been added to **Connectors** in this release.

#### Event-driven RAG Indexing

Knowledge base articles ingested onto the Aisera Platform can now be indexed for RAG retrieval immediately upon ingestion, rather than waiting for a batch job to run. This reduces delay between content ingestion and availability in response.

This feature is controlled by a tenant-level feature flag and is disabled by default.

**To enable event-driven RAG indexing for your tenant:**

1. In the Aisera Admin UI navigate to **Admin UI > Configuration > Feature Flags**.
2. Enable the **Enable ConvAI RAG Event Driven Indexing** feature flag.
3. Select **OK** to save the new configuration.

## Enhancements

The following enhancements have been added to the Aisera Gen AI Platform in this release.

### Connectors

The following enhancements have been made to **Connectors** in this release.

#### SharePoint Connector Rest API v2.0 Compatibility

In response to the deprecation of the SharePoint native app model using the classic SharePoint REST APIs, the SharePoint connector has been updated to support authentication via Microsoft Entra ID and the SharePoint REST v2.0 API.&#x20;

This change primarily affects **On-Premises** SharePoint installations. If you organization uses an On-Premises SharePoint integration, you must update it to use the new authentication method. See the [SharePoint connector documentation](https://docs.aisera.com/aisera-platform/adding-data-to-your-tenant/integrations-and-data-sources/connectors/sharepoint-connector) in the Aisera Product Documentation for updated setup instructions.

### Agent Assist

The following enhancements have been made to **Agent Assist** in this release.

#### Agent Assist compatibility with Conversational AI v2.0

Agent Assist can now be configured using Conversational AI v2.0.

**To enable Conversational AI v2.0 with Agent Assist:**

1. Navigate to **Settings > AiseraGPT** in the Aisera Admin UI
2. Select **+ New AiseraGPT**.
3. Under the **Advanced** section of the **New AiseraGPT** window, ensure **Conversation AI v2.0** is enabled.
4. In the **Fulfillment Types** dropdown menu select **Private KB**.
5. Ensure **Enable Contextual Disambiguation** is <mark style="color:$danger;">disabled</mark>.
6. Select **OK** to save the new AiseraGPT.

The newly created Conversational AI v2.0 AiseraGPT will be compatible with Agent Assist.

## Issues Resolved in the Release

<table><thead><tr><th width="152.99993896484375" valign="top">Aisera ID</th><th width="202.5999755859375" valign="top">Product Area</th><th valign="top">Summary</th></tr></thead><tbody><tr><td valign="top">3.25.2026.1</td><td valign="top">Conversation</td><td valign="top">Escalation command from <code>no_answer_message</code> does not trigger Escalation Feedback flow and causes feedback loop.</td></tr><tr><td valign="top">3.25.2026.2</td><td valign="top"></td><td valign="top">Conversation message disappears for other languages when updating English entry.</td></tr><tr><td valign="top">3.25.2026.3</td><td valign="top"></td><td valign="top">Multiple requests display the same title in the requests list.</td></tr><tr><td valign="top">3.25.2026.4</td><td valign="top"></td><td valign="top"><code>live_agent_exit_hint</code> message displaying after every response.</td></tr><tr><td valign="top">3.25.2026.5</td><td valign="top"></td><td valign="top">Messages can  be significantly delayed or arrive out of order when a channel experienced API errors or rate limiting.</td></tr><tr><td valign="top">3.25.2026.6</td><td valign="top">Analytics and Testing</td><td valign="top">Flows in cloned applications are not accessible.</td></tr><tr><td valign="top">3.25.2026.7</td><td valign="top">Conversational AI</td><td valign="top">Flow Indexer Job is failing.</td></tr></tbody></table>

&#x20;