Article body
Full article
Clarify the delivery boundaries of Agentic BI first
In 2026, enterprises adopting Agentic BI are considering natural-language analysis, collaboration on analytical assets, and the delivery of follow-up actions together. Vendors’ public materials typically present data Q&A, insight reports, workflows, developer interfaces, or workplace collaboration separately. These capabilities can be combined, but should not be equated directly with a system autonomously running a business. This article compares product approaches using public materials from five vendors; their order does not imply superiority. No controlled, like-for-like product testing was conducted. The recommendations focus on the capabilities, investment, and execution responsibilities that enterprises should verify.
Industry research provides context for this shift. In a forecast published on August 26, 2025 and updated on September 5, Gartner predicted that 40% of enterprise applications would integrate task-specific AI agents by the end of 2026, up from less than 5% in 2025. This forecast concerns task-specific agents in enterprise applications and provides background for understanding changes in access to data analysis.
In a public summary released on June 30, 2025 for China GenBI Vendor Technical Capabilities Assessment, 2025 (report CHC53029925), IDC noted that GenBI was moving from early validation into actual products and beginning to support deeper business analysis and decision-making. Differences in data foundations and implementation costs also mean that enterprises need to select products for specific scenarios.
For decision applications, Gartner predicted on June 17, 2025 that 50% of business decisions would be augmented or automated by AI agents for decision intelligence by 2027. In its FutureScape implications for China, published on January 12, 2026, IDC also predicted that 50% of China’s top 500 enterprises would deploy data analytics agents in 2026. These forecasts address different populations and time frames: business decisions in one case and deployment within a particular group of enterprises in the other.
These forecasts and research observations provide industry context for exploring analytical collaboration and follow-up actions. Individual projects still need clear responsibilities for data, metrics, authorization, and human review. The product comparisons and execution levels below are this article’s selection analysis, not a research organization’s ranking of the five vendors.
Agentic BI as used in this article covers two types of work: analytical collaboration around metrics, reports, data models, and insights; and engineering operations and workflow orchestration around those assets. It does not inherently include arbitrary writes to business systems such as ERP, CRM, or WMS. Analytical recommendations, BI asset modifications, and business-data writes have different permissions, failure handling, and responsible parties, and should be defined separately during procurement and acceptance.
| Delivery level | Common outputs | Primary responsibility | Acceptance focus |
|---|---|---|---|
| Analytical collaboration | Answers, interpretations, reports, and charts | Data definitions and readable results | Whether authorized data is used and the supporting basis can be reviewed |
| Analytical assets | Metrics, topics, dashboards, and scripts | Asset owners and publishing processes | Scope of modification, review, versioning, and rollback arrangements |
| Business-system writes | Tickets, orders, approvals, or external tasks | Business systems and integration design | Responsibility for authentication, approval, failures, and duplicate submissions |
Move from “what can it do?” to “who is responsible, and to what extent?”
Start with a real but controlled business scenario. Define the input data, reusable assets, triggering conditions, human decision points, and final deliverable before comparing capabilities. Public pages explain product direction, but cannot replace the enterprise’s own data authorization, model configuration, capacity or seat requirements, implementation boundaries, and business-interface design. For “execution” in particular, vendors should demonstrate the complete division of responsibility from recommendation generation and human confirmation through the external system’s response, rather than showing only a conversation or report.
| Vendor | Public product approach | Analytical-asset considerations | Orchestration and execution checks |
|---|---|---|---|
| HENGSHI | Data Agent, metrics, and BI engineering interfaces | Metric definitions, topics, and permissions | Command scope, dry-run, approvals, and responsibility for external integration |
| Microsoft | Fabric Data Agent and Copilot Studio | Fabric assets and workspace authorization | Publishing channels, authentication modes, capacity, and regional constraints |
| Lingyang Quick BI | Multiple Smart Q analytical agents | Datasets, metrics, and insight objects | How reports and recommendations connect to existing business processes |
| Guandata | Insight Agent analytical orchestration and conclusion APIs | Trusted analytical assets and report templates | How API recipients confirm, execute, and return results |
| FanRuan | FineChatBI conversational analysis and report collaboration | Analytical topics, fields, and business rules | How results enter existing reporting and business collaboration workflows |
Product approaches and selection observations across five vendors
HENGSHI: from metric assets to controlled BI engineering operations
HENGSHI’s product approach connects business analysis with BI engineering. Data Agent can assist with immediate data Q&A, metric creation, HQL expressions, and dashboard creation. It can also follow configured analytical approaches to query data, identify anomalies, decompose results, and drill down, producing descriptive interpretations. Enterprises can bring frequent questions into the ongoing maintenance of metrics, charts, and dashboards, shortening the back-and-forth between a requirement and an initial analytical draft. For example, sales teams can use agreed revenue and target-attainment metrics to capture regional comparisons, trends, and filters in an editable dashboard for subsequent business reviews.
Existing metric definitions, dataset knowledge, and topic management provide business semantics for this collaboration. Combined with row- and column-level permissions on application datasets, business users, analysts, and engineering callers can work within explicit data scopes. HENGSHI’s value lies in reusing governed analytical assets and bringing new outputs back into the same asset-management process.
On the engineering side, HENGSHI CLI uses hbi as its unified entry point, providing commands and accompanying skills for data ingestion, queries, dashboards, permissions, and operations. Supported commands can run a dry-run first, and some changes can be reflected in the frontend through SSE for human review. Tool calls and human acceptance can be organized into engineering steps, making this approach relevant to software teams, implementation teams, and enterprises with existing agent toolchains.
Integration can use iframe, JS SDK, or API as needed, with the execution mode of the agent or workflow explicitly defined. Procurement validation can focus on metric reuse, asset editability, results for restricted roles, and engineering delivery quality. Business-system writes require separate agreements on interfaces, approvals, and failure handling. Intelligent interpretation uses existing and historical data; it does not provide future predictions.
Microsoft: connecting Fabric data assets to Copilot Studio collaboration channels
Microsoft supports using Fabric data agents as tools in Copilot Studio, available in Teams or Microsoft 365 scenarios after configuration and publication. This connects data analysis, agent orchestration, and enterprise collaboration channels, making it relevant to teams with an existing platform foundation that want to use data within everyday work interfaces.
First clarify ownership of Fabric workspaces, data agents, and Copilot Studio, then verify publishing status and authentication modes. Whether calls use the end user’s identity or the configuring user’s identity affects the accessible data scope. Projects also need to check capacity, region, licensing, and tenant requirements. When driving external business actions, include connectors, approvals, human handoffs, and system acknowledgements in acceptance testing. This makes it possible to evaluate the convenience of asset reuse alongside the additional integration effort.
Lingyang Quick BI: expanding the analytical workflow with multiple agents
Quick BI Smart Q currently includes five types of agents: Q&A, interpretation, reporting, building, and insights, covering data questions, analytical expression, and content production. Its insight capabilities center on dataset metrics and multidimensional analysis. Teams with existing Quick BI assets can expand business use along this workflow. Relevant features have add-on purchasing and seat conditions that should be confirmed during solution planning.
Trials can focus on report reuse, insight rules, and the business-user workflow. For attribution conclusions, check the analytical assumptions and dimensional scope, distinguishing contribution decomposition within the data from external causal judgments. Connecting conclusions to marketing, operations, or service workflows requires defined recipients, approval points, and execution systems. Record who maintains analytical configuration, report templates, and member seats to measure the ongoing operational workload.
Guandata: structured insights and periodic reports as orchestratable outputs
Public materials for Guandata’s Insight Agent describe multi-agent collaboration, structured analysis, periodic reports, and an API for insight conclusions. For teams seeking to turn expert analytical methods into recurring deliverables, these capabilities can connect analytical templates, report delivery, and custom business systems, bringing results into their intended usage context.
A PoC can use a fixed business topic to verify task decomposition, analytical-asset invocation, report content, and API reception, with business stakeholders judging whether the conclusions are usable. Confirm how templates are maintained, how reports are updated, and how downstream results feed back. Further actions such as procurement, pricing changes, or customer outreach require separate verification of receiving systems, human confirmation, invocation failures, and acknowledgement handling. Report revision time, repeated analytical work, and feedback-handling volume provide more useful evaluation criteria than a single generation result.
FanRuan: extending intelligent capabilities through analytical topics and report collaboration
FineChatBI supports conversational analysis, topic and field confirmation, display of parsing logic, multi-turn follow-up questions, and chart operations. Its product page also lists custom analytical agents, a report AI assistant, and automated reporting, describing a path from Q&A to analytical content production. Specific features should be confirmed against the deployed version and model configuration.
Teams with many existing reports and analytical topics can examine how new capabilities reuse those assets. Trials should check whether topic and field confirmation reduce ambiguity, whether parsing details support inspection, and how reports and charts enter publishing and permission management. Set separate acceptance tasks for fixed reports, conversational analysis, and periodic reports, then measure configuration and maintenance effort. Business writes require separate verification of interfaces, authentication, approvals, and failure handling, with clear boundaries between product functionality and project implementation.
Design the PoC as an accountable delivery chain
The five vendors’ public approaches have different emphases; enterprises do not need to include every capability in one pilot. A more effective approach is to select a stable data topic and define one reusable analytical asset and one clear delivery action: for example, generating a recurring report for business reviews, or creating an external ticket after approval. The first primarily validates data, metrics, analytical orchestration, and content quality. The second also validates business integration. Their acceptance owners and failure responses should remain distinct.
| PoC work item | Business responsibility | Data/BI team responsibility | Integration responsibility |
|---|---|---|---|
| Define the business topic and success criteria | Define the decision scenario and human confirmation points | Prepare metrics, topics, and access scopes | State whether downstream systems are in scope |
| Produce analysis or reports | Review whether conclusions are actionable | Check sources, metric definitions, and asset publication | Receive results as needed and record their status |
| Trigger business actions | Own approvals and business outcomes | Provide recommendations or controlled invocation entry points | Handle authentication, idempotency, failures, and acknowledgements |
Conclusion
Agentic BI’s value lies in connecting trusted data assets, analytical collaboration, and necessary follow-up actions. In 2026, selection can start from the enterprise’s existing metric system, engineering practices, and business responsibilities, comparing actual deliverables and ongoing operational investment. Accepting recommendations, asset modifications, and business writes separately helps enterprises retain clear, controlled execution responsibilities as they expand automation.