FlashSystem.ai

About FlashSystem.ai

The assistant supports common tasks such as creating volumes, volume groups, hosts, host clusters, and snapshots. It can also assign snapshot or replication policies to volume groups to support disaster recovery workflows.

The assistant uses retrieval techniques to base responses from IBM FlashSystem product documentation, ensuring accurate and relevant answers to user queries.

FlashSystem.ai requires an active IBM Storage Insights connection, using a Storage Insights API key and tenant ID.

Capabilities

Natural‑language administration

FlashSystem.ai supports natural‑language interaction. You can use plain‑language instructions in the management GUI to retrieve system information and perform selected administrative tasks, including the following operations and queries.

• Creating volumes, volume groups, hosts, and host clusters
• Creating snapshots
• Assigning snapshot or replication policies to volume groups
• Renaming volumes
• Increasing volume size
• Querying system configuration
• Asking follow‑up questions to refine results

Supported administrative actions

You can use natural language instructions to initiate the following supported administrative actions:

• Create volumes
• Create hosts and host clusters
• Create volumes and attach to hosts
• Create volume groups
• Create snapshots
• Rename or expand volumes
• Move volumes between volume groups
• Modify snapshot and replication policies on volume groups
• Perform host‑cluster operations (create clusters, map volumes)

Information retrieval capabilities

You can use natural language queries to retrieve the following information from managed systems:

• System overview
• Volumes and volume groups
• Hosts and host clusters
• Host–volume mappings
• Snapshots
• Snapshot and replication policies
• Storage pools
• Partitions
• Portsets

Filtering and aggregation are supported for all query results.

Context‑aware responses

The assistant identifies the systems that you are managing and uses this context when it runs queries or other supported operations. By default, when you request information such as a list of volumes, the assistant retrieves data from the active system. You can also request information from any other system that is configured in the container by specifying the system name.

Workload placement recommendations

FlashSystem.ai integrates with IBM Storage Insights to provide workload placement recommendations. It enables natural language queries to identify optimal placement for new workloads by using fleet-wide analytics, historical performance trends, and capacity forecasts.

Systems must be registered in both FlashSystem.ai and IBM Storage Insights to use this feature. To receive workload placement recommendations in FlashSystem.ai, describe requirements in natural language. The requirements include capacity needs, performance characteristics (such as IOPS and latency), read and write patterns, block sizes, and snapshot requirements. FlashSystem.ai analyzes the storage fleet based on this input to recommend optimal workload placement options.

The time required for analysis depends on the number of systems in the fleet. Progress indicators and tool call panels display the status during execution. During this time, user interaction is unavailable. Starting a new chat session cancels any ongoing analysis.

Proactive AI-initiated notifications

FlashSystem.ai supports proactive AI-initiated notifications by initiating conversations when critical issues require attention. This capability enables early identification and resolution of potential issues before they impact the environment.

The system monitors storage and IBM Storage Insights for security, compliance, and capacity issues, presenting actionable notifications in the chat interface at sign-in.

To enable email notifications, the tenant administrator email ID must be configured in IBM Storage Insights.

At sign-in, relevant notifications appear in the chat interface. Select an action to start processing. You can mute specific notification types to suppress similar alerts. Up to three high-priority notifications display at once. Selecting an action removes other notifications from the view.

If no action is taken, the same notifications continue to appear in subsequent chat sessions until they are addressed or replaced by higher-priority notifications. Starting a new chat session displays the next set of highest-priority notifications that are pending.

Assistant behavior and approval model

For all supported create and modify actions, FlashSystem.ai requires explicit user approval and follows these rules:

• The assistant presents a human‑readable confirmation dialog before each operation it runs.
• The dialogs explain the action in plain language.
• You must explicitly approve each configuration action before it runs.
• All changes are performed through on‑premises API calls; nothing is executed in the cloud.
• The assistant is available only when the AI container is deployed and registered.

You can view results in a table format that supports filtering and aggregation for easier analysis.

Guardrails for safe operation

FlashSystem.ai enforces guardrails that block destructive, security‑related, and other high‑risk actions. The assistant does not support the following operations:

• Deleting system objects
• Reducing the size of objects
• Reversing replication relationships
• Restoring data from snapshots
• Removing snapshot or replication policies
• Performing any security‑related tasks
• Creating objects that require special user handling (for example, partitions, replication policies)

The assistant does not have access to any security-related information on the system.

Human‑in‑the‑loop administrative actions

All configuration actions run on‑premises. The assistant can perform supported actions on the user's behalf and uses human‑in‑the‑loop prompt boxes to confirm approval before each action is run.

Access, support, and licensing

User access

FlashSystem.ai uses role‑based access control to define which actions each user is permitted to perform:

• A dedicated Administrator account must be created on each system and provided to the AI container.
• This account must be provisioned manually by the customer.
• Read‑only interactions are possible with lesser‑privileged accounts, but configuration actions require Administrator access.
• A Security Administrator is required to register each system with the AI container by using the chaicontainerinfo command.

Licensing

FlashSystem.ai is included with the system at no additional cost. No WatsonX.ai subscription is required.