| copyright |
|
||
|---|---|---|---|
| lastupdated | 2026-07-17 | ||
| keywords | watsonx BI, release notes, what's new, watsonx BI as a Service | ||
| subcollection | watsonx-bi | ||
| content-type | release-note |
{{site.data.keyword.attribute-definition-list}}
{: #my-service-relnotes_saas}
Use these release notes to learn about the latest updates to {{site.data.keyword.wxbia_full}} that are grouped by date. Release notes are available for a minimum of three years. {: #shortdesc}
{: #subcollection-jul1626} {: release-note}
TIME data type supported in uploaded files
: The TIME data type is now a supported column data format in uploaded files. For more information, see Uploading a file{: external}.
{: #subcollection-jul0926} {: release-note}
Availability of native metadata enrichment in watsonx BI
: Watsonx BI as a Service now includes built-in metadata enrichment. The new watsonx BI native enrichment service uses generative AI to automatically add business context, semantic meaning, and descriptions to your data. Enriched metadata improves data discovery and helps deliver more accurate results.
: This capability removes dependency on IBM watsonx.data intelligence for metadata enrichment.
: If your Cloud account includes an instance of watsonx.data intelligence, watsonx.data intelligence is used as the default enrichment provider. However, Administrators can choose between watsonx BI enrichment and watsonx.data intelligence enrichment for users in the account.
: - IBM watsonx BI enrichment - A native, AI-powered enrichment service that requires minimal setup and does not depend on governance artifacts such as business terms and categories.
: - IBM watsonx.data intelligence enrichment – Governance-focused enrichment that supports classification, business terms, categories, and data classes.
: For more information, see Enriching data in watsonx BI{: external}.
IBM watsonx.data intelligence is now optional for watsonx BI as a Service
: IBM watsonx.data intelligence is now optional for watsonx BI. Previously, watsonx.data intelligence was a required dependency for metadata enrichment.
: Watsonx BI no longer requires IBM watsonx.data intelligence to enrich metadata. Watsonx BI as a Service now includes built-in metadata enrichment that uses generative AI to automatically generate business context, semantic meaning, and descriptions for your data.
: If you have an instance of watsonx.data intelligence in your Cloud account, watsonx.data intelligence is used as the default enrichment provider. Administrators can also choose between watsonx BI enrichment and watsonx.data intelligence enrichment from Configurations and settings.
: IBM watsonx.data intelligence continues to be required for watsonx BI deployments on IBM Software Hub.
: For more information, see Enriching data in watsonx BI{: external}.
Microsoft Azure Synapse Analytics available as a connector
: You can now connect to data in Microsoft Azure Synapse Analytics to create metrics and ask questions in watsonx BI.
: For more information, see Microsoft Azure Synapse Analytics connection{: external}.
{: #subcollection-jun1226} {: release-note}
Data analysts can review feedback to improve responses
: You can now provide feedback on responses in Conversations, and Data analysts can review and act on that feedback at the project level. Previously, users could only submit feedback.
: This update introduces a structured feedback loop that helps teams improve data quality, metrics, and AI responses over time.
: For more information, refer to Response feedback{: external}.
{: #subcollection-jun0426} {: release-note}
Improved response handling for no data and no asset scenarios
: This release improves how watsonx BI responds when a question cannot produce a meaningful result when no data is returned or no supporting data asset exists. These updates help users better understand data availability.
: No data
: When a query generates a valid SQL but returns no results, watsonx BI now analyzes column metadata to provide helpful context about available data.
: For example, the question, What was our revenue in 2020? now returns,There’s no data available for 2020. Data is available for 2021–2024.
: No asset
: When no SQL can be generated because the required data does not exist, watsonx BI now specifies which part of the question cannot be answered. It can also suggest a relevant alternative question or exploration path, helping users move forward without needing to contact support.
: For example, the question, How many returns were made at the Phoenix call center? now returns a clear, informative response instead of an error, I'm sorry but the data you're looking at doesn't contain any information about product returns. Would you like to explore how many total calls were handled at the Phoneix call center during a specific time period?. You can reply Yes to run the suggested question.
{: #subcollection-may2826} {: release-note}
Improved metric generation with question–SQL pairing
: You can now generate metrics by providing a business question and its corresponding SQL query. This update replaces the earlier approach, which relied on enriched data and the semantic data model to derive metrics.
: Pairing a question with SQL anchors metric generation in a real business scenario. The SQL defines the exact logic, while the question captures the analytical intent. Together, they reduce ambiguity and improve metric accuracy and relevance.
: In addition, watsonx BI extracts relationships, calculations, and structure directly from the SQL. This reduces the need for manual modeling and helps data analysts generate metrics faster without using advanced modeling techniques.
: For more information, refer to Generating metrics automatically{: external}.
: This is a Technology Preview feature in watsonx BI. Technology Preview offers customers early access to product features, allowing them to explore functionality and share feedback during development. These features are provided for evaluation purposes and might not be fully functional or complete. {: important}
{: #subcollection-may1426} {: release-note}
Update to supported large language models in watsonx BI
: IBM is updating the supported large language model (LLM) options used for natural language conversations and analytics in watsonx BI. OpenAI gpt-oss-120b with chain of thought (CoT) reasoning is now the only supported model. All other LLM options are deprecated.
: No action is required from users using watsonx BI as a Service. Your watsonx BI instance continues to work as expected using the supported model.
{: #subcollection-apr2926} {: release-note}
Improvements to the AI steps panel (chain of thought only)
: The AI steps{: external} panel now offers improved usability and clarity when you use the chain of thought (CoT) reasoning option.
: The panel now displays a rephrased version of your question at the top of the screen. This update helps clarify how the AI interpreted the query.
: The Details view, typically found in step 3, now includes a simple explanation of what each query step does, along with the values used in the query and the columns returned. From the Data view, you can now download intermediate results for a step as a CSV file.
{: #subcollection-apr2326} {: release-note}
Improved experience when watsonx.data intelligence Capacity Unit Hour (CUH) limit is reached
: This release improves the watsonx BI experience when your watsonx.data intelligence Capacity Unit Hour (CUH) limit is reached.
: Previously, when CUH was exhausted, metadata import and enrichment errors blocked progress, preventing you from continuing work. Administrators had to upgrade the watsonx.data intelligence plan at the Cloud account level to unblock users.
: With this update, your work continues without interruption, even when metadata import and metadata enrichment jobs cannot run due to insufficient CUH. Watsonx BI continues to return results. However, response accuracy might be impacted because enriched business context is missing.
: When the CUH limit is reached, you can still:
: - Create and manage metrics
: - Work with semantic data models
: - Analyze existing data sources
: - Ask questions against uploaded files
: If response accuracy is critical during periods of high CUH usage, Administrators and Cloud account owners can upgrade the watsonx.data intelligence plan.
: For more information, see IBM watsonx.data intelligence requirement for {{site.data.keyword.wxbia_full_notm}}{: external}.
{: #subcollection-apr0226} {: release-note}
AWS Databricks available as a connector
: You can now connect to data in AWS Databricks to create metrics and ask questions in watsonx BI.
: For more information, see AWS Databricks connection{: external}.
{: #subcollection-mar2626} {: release-note}
Microsoft Azure Databricks available as a connector
: You can now connect to data in Microsoft Azure Databricks to create metrics and ask questions in watsonx BI.
: For more information, see Microsoft Azure Databricks connection{: external}.
{: #subcollection-mar1826} {: release-note}
Administrators can now set up watsonx BI
: Administrators can now provision and set up watsonx BI.
: IBM Cloud account owners can assign the minimum set of IBM Cloud access policies that allow Administrators to initialize and configure watsonx BI for the account. This enhancement gives account owners and Administrators more flexibility and helps teams onboard users faster.
: For more information, see Enabling administrators to set up watsonx BI{: external}.
Support for currency and percentage formatting in conversations
: Watsonx BI can now apply Currency and Percent formatting from the semantic data model to conversational responses.
: When you define format types for metric columns in a semantic data model, generated responses in Conversations display values with the correct currency symbol or percent sign, decimal places, and thousands separators. This update improves readability and reduces ambiguity in responses that include monetary or ratio-based values.
: For more information, see Formatting data{: external}.
{: #subcollection-mar1326} {: release-note}
Multi‑asset queries now supported
: You can now ask natural‑language questions that use multiple data assets in a single query. In earlier versions, watsonx BI selected one best‑fit asset to answer a question. With multi‑asset queries, watsonx BI can retrieve data from one asset, use that result to build a new query, and return a combined answer. This approach supports multi‑step questions such as comparisons and rank‑based filters.
: For example, you can ask:
: - For the product with the most revenue, how many returns did it have?
: - What is the return count for the top 10 SKUs by revenue in 2025?
: Watsonx BI automatically determines which assets are required, issues independent queries, and generates a combined answer.
: The multi-asset query capability works when chain of thought (CoT) is enabled in watsonx BI and all referenced assets are stored in the same container, such as a project or catalog.
: For more information, see Asking questions across multiple data assets.
{: #subcollection-mar0526} {: release-note}
IBM Granite 4 model now available in watsonx BI as a Service
: IBM Granite 3 models are being discontinued and replaced with IBM Granite 4{: external} in watsonx BI as a Service.
: You do not need to take any action. This upgrade is seamless and does not affect your watsonx BI experience.
: For more information, see Understanding the large language model in your account{: external}.
{: #subcollection-feb2626} {: release-note}
Apply user filters for row-level security
: User filters are now available to help protect sensitive data in metrics. Administrators and Data analysts can define conditions on a metric definition in the semantic data model to ensure that users see only the data they have permissions to access.
: When a user views a metric, the values are calculated only from the rows they’re allowed to see, based on the filters applied.
: For more information, see Manage access to data with User filters{: external}.
Cognos Analytics data modules now supported as a data source and improved connection process
: You can now use Cognos Analytics data modules as a data source in {{site.data.keyword.wxbia_short}}. This enhancement expands support beyond Framework Manager (FM) packages and lets you create and semantic data models metrics using your existing data modules.
: Data modules that use uploaded files and database tables as data source are supported. Data modules that use FM packages as a source are not supported.
: The experience for connecting to Cognos Analytics and importing a package has also been improved. You can now:
: - Create and test a Cognos Analytics connection directly during data selection
: - Choose either an FM package or a data module from your Cognos environment
: - Automatically populate your semantic data model with Cognos metadata and governance rules
: In addition, sync behavior has been improved to help semantic data models stay aligned with source changes. Updates to FM packages are applied automatically, while data modules prompt you to resync whenever changes are detected in the underlying source.
: For more information, refer to IBM Cognos Analytics{: external}.
: Integration with Cognos Analytics, is a Technology Preview feature in watsonx BI. Technology Preview offers customers early access to product features, allowing them to explore functionality and share feedback during development. These features are provided for evaluation purposes and might not be fully functional or complete. {: important}
Improvements to Conversation history
: You can now delete multiple conversations at once using the new Toggle selection mode button in the Conversations panel.
: To multi‑select conversations, expand the panel, select Toggle selection mode, choose the conversations you want to remove, and delete them in one action.
: A new search feature also lets you quickly find conversations. Simply type a keyword into the search box to locate conversations with that keyword in the title.
{: #subcollection-feb526} {: release-note}
Model update for chain of thought (CoT) reasoning
: Chain of thought (CoT) reasoning will now use OpenAI gpt‑oss‑120b instead of Meta Llama 4 large language model (LLM). This update ensures improved quality and consistency in CoT reasoning.
: Administrators must reselect CoT with the new gpt‑oss‑120b model in Configuration and settings > Model settings to continue receiving accurate and supported responses.
: Users who previously enabled CoT with the Llama model continue to receive responses for now. However, after Llama is fully discontinued, answer accuracy might be impacted until the model selection is updated.
: If Administrators do not switch to CoT with GPT, users will automatically be moved to an LLM that does not support CoT when the current model is discontinued.
Provide custom instructions and context to guide AI behavior
: A new AI instructions and context field is available in the modeling experience, which allows you to add business logic and rules that the large language model (LLM) follows when interpreting and generating responses that are related to a metric.
: Use AI instructions in scenarios where consistent interpretation depends on rules, such as when:
: - Your dataset uses business‑specific terminology that differs from column names
: - You use standard calculation methods or rules, such as weighted averages or variance formulas
: - Certain columns must always appear together for context
: - You rely on default filters or display preferences
: - You need to enforce organizational reporting standards, such as fiscal or seasonal time logic
: For more information, see Adding instructions and context for AI{: external}.
Download responses and related data from a conversation
: You can now download responses and their associated components from a conversation, making it easy to save and share insights. Available download options include:
: - Text response – The generated answer in .txt format
: - Visualizations – Charts downloaded in .png format.
: - Tabular data – Data tables downloaded in .csv format.
: - Diagnostic information – Details such as AI steps, generated SQL, and response ID in .json format (available to Data analysts only).
: For more information, see Downloading responses and data from a conversation{: external}.
Expand and collapse the Conversations panel
: You can now expand or collapse the Conversations panel to create more space for your active chat. Previously, this panel was fixed and could not be collapsed.
Conversation scope list improved with descriptions
: The conversation scope list next to the input box in Conversations now includes descriptions (if applicable) for each item, such as projects, the Metrics catalog, and any files that you have uploaded.
: You can now also open and preview any of these items in a new tab directly from the conversation scope. For example:
: - Launching a project opens the data platform project view, including its assets.
: - Launching an uploaded file opens the file view in a new tab for quick review.
: These enhancements make it easier to choose the data that you want to focus your conversation on.
Improved look and feel of the welcome message
: The welcome message in Conversations now has an improved look and feel. You can still click any of the suggested questions to start exploring your data, just like before.
{: #subcollection-jan2226} {: release-note}
OpenAI gpt-oss-120b now available as an independent model option
: The OpenAI gpt‑oss‑120b large language model (LLM) is now available as a stand-alone option in watsonx BI as a Service. Administrators can choose this LLM on the Configuration and settings > Model settings.
: This LLM is used for SQL generation tasks and all other language‑model–driven functions within watsonx BI.
: Selecting this LLM provides advanced question interpretation, stronger semantic understanding, and higher accuracy across a wide range of queries.
{: #subcollection-jan1526} {: release-note}
IBM watsonx.data intelligence is required for watsonx BI as a Service
: IBM watsonx BI as a Service requires an IBM watsonx.data intelligence plan to be active in the IBM Cloud account.
: As part of an initial, limited‑time introductory promotion, watsonx.data intelligence was included with watsonx BI; this promotion has since expired, and clients are now required to explicitly select and activate a watsonx.data intelligence plan. IBM Cloud accounts without an active watsonx.data intelligence plan will not be able to provision watsonx BI.
: Cloud account owners must have an active watsonx.data intelligence plan or purchase a watsonx.data intelligence plan and activate it. Cloud account owners can choose from any watsonx.data intelligence plan, including Trial, Essentials, Standard, or Premium.
: For existing watsonx BI users, the service checks whether a watsonx.data intelligence plan is available. If one is not found, users are guided to provision a plan{: external} or asked to contact the Cloud account owner for assistance before continuing to use watsonx BI.
: To check whether the Cloud account has an active watsonx.data intelligence plan:
: 1. Log into the IBM Cloud account.
: 2. Click Resource list and expand the Analytics section and check if watsonx.data intelligence is listed as an existing resource.
{: #subcollection-dec325} {: release-note}
Chain of thought reasoning
: Chain of thought (CoT) reasoning is now available with Meta Llama 4 and IBM Granite-3-8b-instruct. Available as a Technology Preview feature in watsonx BI as a Service, it delivers step-by-step reasoning for complex or multi-part questions, providing structured answers and visibility into how queries are built.
: Technology Preview offers customers early access to product features, allowing them to explore functionality and share feedback during development. These features are provided for evaluation purposes and might not be fully functional or complete. {: important}
: You can view reasoning steps, chosen data sources, intermediate calculations, and generated SQL, helping you understand the AI’s decision-making process.
: Administrators can choose Meta Llama 4 and IBM Granite-3-8b-instruct with chain of thought from Configuration and settings > Model settings page.
: For more information, see Chain of thought reasoning{: external}.
IBM watsonx BI remote Model Context Protocol (MCP) server now available for queries
: You can now use the IBM watsonx BI remote MCP server, a Model Context Protocol (MCP) compliant service that seamlessly connects AI agents with data in watsonx BI, enabling intelligent data and insight retrieval.
: For more details about remote MCP server, see IBM watsonx BI remote Model Context Protocol (MCP) server{: external}.
{: #subcollection-aug1425} {: release-note}
Introducing {{site.data.keyword.wxbia_short}} as a Service on IBM Cloud
: {{site.data.keyword.wxbia_full}} is a new service that is now available on IBM Cloud.
: {{site.data.keyword.wxbia_short_cap}} is a generative AI-powered business intelligence partner that uses natural language to analyze complex data, providing personalized insights and data-driven recommendations that are actionable and easy to understand. It works with the tools you already use and delivers intelligent, transparent answers tailored to your needs.
: Use {{site.data.keyword.wxbia_short}} to:
: - Get faster insights - Ask questions using natural language and receive answers through a simple, intuitive interface.
: - Understand data more deeply - Uncover trends, understand causes, predict future outcomes, and gain actionable strategies.
: - Act with greater confidence - Governed data and artificial intelligence (AI) help ensure trusted models, transparency for all AI workflows (powered by IBM watsonx™), and security with insights that derive from a single source of truth through a centralized Metrics catalog.
: - Perform advanced data visualization interaction - Customize the look and feel of data visualizations with conversational and graphical interactions.
: - Semantic automation - Automate data profiling, business term matching, and data modeling.