Article body
Full article
Traditional BI, ChatBI and Agentic BI are analytical approaches that can work together. HENGSHI SENSE connects fixed dashboards, conversational exploration and task-based authoring through data models, HQL, metrics and permissions, letting teams choose an entry point for their business needs and maintain the analytical assets they produce.
A recurring business report, an ad hoc data request and a complete data-to-dashboard delivery task require different interactions and checks. Enterprises should define the task and expected result first, then decide whether analysts should configure it, business users should explore through conversation, or Data Agent should combine query and authoring steps.
1. What Does Each Approach Emphasize?
| Approach | Main focus | Typical outputs |
|---|---|---|
| Traditional BI | People organize models, metrics, queries and visualization | Recurring reports, dashboards and self-service analytics applications |
| ChatBI | Natural-language queries, follow-up questions and exploration | Answers, charts and dashboards in supported entry points |
| Agentic BI | Task understanding, tool calls and delivery within an analytical workflow | Multi-step analysis, analytical assets and integrated collaboration processes |
The names do not establish absolute boundaries. ChatBI may support multi-turn conversations and dashboard generation; traditional BI also includes scheduled updates, subscriptions and alerts. Agentic BI emphasizes how tasks advance, while the actual operations available depend on tools, entry points, permissions and configuration.
2. How Dashboards, ChatBI and Data Agent Collaborate in HENGSHI
Recurring Questions: Maintained Reports and Dashboards
Monthly business reviews need stable metrics, formats, filters and presentation. Analytics teams prepare models and HQL metrics in HENGSHI, author dashboards and publish them for business users. Subscriptions can provide regular delivery where needed. Fixed pages preserve established analytical structures for repeated review and reconciliation.
Ad Hoc Questions: Continue Exploring with ChatBI
Business users can ask questions about authorized data and explore by time, region, product and other dimensions. Standalone HENGSHI ChatBI provides a conversational entry point and can turn conversational charts into dashboards. Multi-turn analysis requires checking which conditions should carry forward or change, and clarifying ambiguous business definitions.
Authoring and Complex Analysis: Advance with Data Agent Assistance
The HENGSHI Data Agent sidebar can use account and current-page context to find authorized resources and execute one or more queries according to question complexity. It can help create and edit metrics, charts, filters, explanations and dashboards, allowing analysts to continue producing results in the same environment.
Agent mode suits tasks that require deciding the next step; Workflow mode suits relatively fixed processes. External integrations through APIs should use the modes and tools actually available. Browser capabilities should not be assumed to exist in every API.
3. Shared Foundations: Maintain Definitions, Authorization and Results
HENGSHI organizes analytical semantics through models, HQL, metric descriptions and business knowledge. Teams define concepts such as net sales and active customers, reference confirmed definitions in dashboards and conversational tasks, and check the actual period, data timestamp and scope. Metrics with identical names still require verification of their underlying definitions.
Resource authorization, data permission modes and applicable row and column rules determine access. Data Agent’s ability to find a resource does not allow it to bypass authorization. Embedded and API entry points need correct identity propagation. Newly added models and processed datasets require their own permission checks.
Analytical results also have lifecycles. Authoring-area changes and published snapshots must be managed separately; confirmed changes reach readers after republishing. Conversation history, durable knowledge, saved dashboards and external workflow state serve different purposes. Moving between pages or continuing on another day should not be assumed to restore an entire task state unconditionally.
4. Failure Handling Determines Whether Delivery Is Reliable
| Problem | Appropriate response | What people must verify |
|---|---|---|
| Unclear metric or field meaning | Add definitions, descriptions or knowledge, or ask for clarification | Actual calculations and business meaning |
| Query or tool failure | Inspect execution logs, failed requests and system logs | Specific causes in sources, models, interfaces and permissions |
| Insufficient permissions | Have authorized staff handle access under policy | User identity and legitimate access scope |
| Timeout or inadequate performance | Check queries, concurrency, model calls and infrastructure | Whether optimization preserves the question and calculation semantics |
| Failed external action | Retry, reverse or compensate through the connected system’s process | Completed work, duplicate actions and business outcomes |
Automation can reduce repeated operations, but it must not quietly change statistical granularity, expand permissions or ignore failures to declare a task complete. Changes to models, tools or knowledge should be retested with relevant business questions.
5. Open Integration: Connect More Work Through the Right Entry Points
HENGSHI offers iframe, JS SDK and API integration for delivering analytics in different product forms. HENGSHI CLI provides external agents with structured BI operations and supporting Skills for queries and engineering work on data and dashboards. Applicable commands can be rehearsed before actual writes.
Business-system calls can be added through tools in supported entry points, with implementers handling parameters, authentication, confirmation and result recording. Custom tools in the Data Agent browser interface and API mode have different coverage and require individual validation. High-impact actions such as procurement or price changes also require business systems and accountable staff.
6. Hypothetical Scenario: Three Approaches in One Business Review
The following is a hypothetical solution example. A business leader starts by reviewing monthly net sales and attainment in a HENGSHI dashboard, then asks ChatBI which East China stores have the largest gaps and follows up by product line and period. The data team has already maintained the models and calculations for stores, orders and targets.
With help from the Data Agent sidebar, an analyst organizes store comparisons, trends and explanations, verifies filters and interpretations, and saves the results. An authorized person publishes updates to the business page. If improvement tasks are needed, an integrated and authorized business API initiates them and records actual outcomes.
Fixed dashboards support routine reading, conversation supports ad hoc exploration, and the agent assists analysis and authoring. In HENGSHI, these approaches collaborate around shared data and analytical assets, with definitions, authorization and review sustaining delivery quality.