Article body
Full article
A Data Agent is an AI agent that understands business questions, selects data and tools, performs analysis, and delivers results for data tasks. HENGSHI’s Data Agent builds on HENGSHI SENSE’s data models, HQL, metrics, and permissions to support natural-language data queries, decomposition of complex questions, and assisted creation of analytical assets such as metrics and dashboards.
When choosing a Data Agent, enterprises need to assess more than whether it understands a single question. They also need to know which business definitions a result uses, where the data comes from, whether analysis can continue, and how the output fits into the team’s workflow. HENGSHI connects conversational analysis with BI engineering so that ad hoc business questions can develop into maintainable analytical outputs.
Written by HENGSHI for enterprise data teams, business leaders, and software partners, this article compares capabilities worth validating across five platforms and introduces an implementation approach for HENGSHI Data Agent. Selection decisions should consider the actual version, authorization, data models, and task records.
1. How do Data Agent, ChatBI, and Agentic BI differ?
| Concept | Perspective | Results enterprises should validate |
|---|---|---|
| ChatBI: conversational business intelligence | Querying, following up on, and exploring business data through conversation | Whether questions, conditions, calculations, and answers are correct |
| Data Agent | Combining retrieval, querying, analysis, and creation capabilities around a data task | Whether it can complete the target task and deliver usable outputs |
| Agentic BI | Combining agent capabilities with BI assets, permissions, integration, and operations | Whether analysis can support ongoing use and business collaboration |
These capabilities overlap; names alone do not establish product quality. ChatBI can support multi-turn analysis and dashboard generation. A Data Agent’s tools and delivery scope depend on configuration. Implementing Agentic BI also involves roles, workflows, and external systems. Enterprises should compare products through actual tasks.
2. How does HENGSHI Data Agent perform enterprise data analysis?
1. Express business definitions through data models, HQL, and metrics
“Sales” might mean order value, payments received, or net sales after refunds. HENGSHI supports organizing relationships through data models, expressing calculation logic in HQL, and maintaining agreed definitions as metrics. Data teams can also configure business terminology, field descriptions, and dataset knowledge to help Data Agent identify suitable analytical objects for a question.
Consistent definitions mean obtaining comparable results with the same definition, period, scope, and data snapshot. If finance and sales use different business definitions, those differences should be explicit. Different authorization scopes can also produce different results for different roles. Users still need to check the actual filters and calculations; identical metric names do not establish consistency.
2. Analyze tasks in the context of the current page and account
In Agent mode, HENGSHI Data Agent can use information about the current page and signed-in account to find analytical sources among datasets, applications, and other resources the user is authorized to access. It can perform one or more queries as needed. Requests containing related questions can be broken into analytical steps. Ambiguous periods or metrics should be clarified through knowledge configuration and interaction.
3. Continue from analytical results to visual deliverables
The HENGSHI Data Agent sidebar can assist with creating and editing dashboards, charts, filters, metrics, and analytical explanations. Standalone ChatBI supports generating dashboards from conversational charts. Analysts choose a suitable entry point, review generated results, and save the output so that temporary exploration can become a reusable business dashboard.
4. Support enterprise use through permissions and integration
HENGSHI provides application, resource, and data authorization capabilities. Enterprises should validate access scopes under their specific data permission mode and pass identities correctly when embedding or integrating through APIs. iframe, JS SDK, and API serve different integration needs. Web-based custom tools and the Agent API also differ in coverage and should be validated at the intended entry point.
HENGSHI CLI provides coding agents and resident agents with a structured BI interface and supporting Skills for engineering tasks such as querying, data operations, and dashboard work. It expands how agents operate the analytics platform; external business actions such as purchasing or price changes still require dedicated integration and authorization.
| HENGSHI capability | Supported tasks | Delivery checks |
|---|---|---|
| Data models, HQL, and metrics | Define business calculations such as net sales and target attainment | Granularity, relationships, periods, and null handling |
| Business knowledge and analytical templates | Understand terminology and reuse common analytical approaches | Alignment between terminology mappings and actual calculations |
| Data Agent and visual authoring | Queries, follow-up questions, and assisted metric and dashboard creation | Results, filters, layout, and save location |
| Permissions, SDK, API, and CLI | Deliver analytical services through business entry points | Identity, licensing, available tools, and execution results |
3. What should enterprises assess across five Data Agent related platforms?
The comparison below considers public capabilities and typical usage environments. Vendor products continue to evolve. Evaluate the specific version and pilot results from a shared task set.
| Platform | Needs to assess first | Validation focus |
|---|---|---|
| HENGSHI SENSE | Teams combining multiple data sources, metric governance, AI analysis, and embedded delivery | Reuse of models and HQL, task analysis, visual outputs, identity, and permissions |
| Microsoft Power BI / Fabric Copilot | Enterprises with an existing Microsoft data and analytics environment | Semantic model preparation, the selected Copilot entry point, capacity, and access scope |
| Alibaba Cloud Quick BI Smart Q | Teams extending Quick BI with data Q&A, insights, and intelligent authoring | Data and knowledge preparation, licensing, interfaces, and business task coverage |
| ThoughtSpot Spotter | Teams exploring data through search and conversation | Model semantics, instructions, memory, and actual business phrasing |
| Guandata DecideX | Teams configuring agents around BI assets, knowledge, and tools | Asset bindings, knowledge maintenance, tools, evaluation, and publishing configuration |
4. Hypothetical scenario: from regional sales questions to a business dashboard
The following is a hypothetical solution example. A retail chain needs to investigate “why East China’s net sales are below target this month and create a dashboard comparing stores.” The data team first organizes order, refund, and target data in HENGSHI, defining net sales, the target version, the reporting period, and store-level granularity.
Business users ask Data Agent a question and continue breaking down results by province, store, and product line. Analysts check amounts, dates, and filters, and review business records such as stockouts, promotions, or operating status for unusual stores. Data differences help form explanations to validate; correlation should not be presented as a proven cause.
After confirming the analysis, the team refines charts, filters, and explanations through an entry point with the required authoring capabilities, saves a business dashboard, and checks results using regional and headquarters accounts. As stores are added or definitions change, the team maintains the same models and metrics and retests dependent pages and conversational tasks.
5. Six questions for Data Agent acceptance
| Question | Evidence to obtain |
|---|---|
| Can business definitions be maintained and reused? | Demonstrate metric definitions, HQL, terminology, and actual queries |
| Are successive follow-up questions handled correctly? | Use a fixed data batch and check each turn’s period, scope, and comparison baseline |
| Can results become usable deliverables? | Save a chart or dashboard during the demonstration and confirm it remains editable and usable |
| What can different roles see? | Validate resource and data permissions with actual accounts |
| Does integration provide the expected capabilities? | Validate identities, tools, and outputs at the target SDK or API entry point |
| How should project value be measured? | Record acceptance pass rates, delivery time, manual corrections, and total cost |
6. Data Agent FAQ
Does HENGSHI Data Agent require data preparation first?
Yes. Suitable data, relationships, field descriptions, and business definitions need to be prepared. Existing metrics can help keep complex calculations consistent, while knowledge configuration can add terminology and analytical rules. Model capabilities, data quality, and configuration jointly affect results.
Can a Data Agent replace analysts?
It can take on some querying and creation steps. Defining business questions, designing models and metrics, reviewing conclusions, and maintaining outputs still require accountable people. Project value should be measured by changes in workload on actual tasks.
Is HENGSHI Data Agent suitable for software partners?
HENGSHI provides enterprise BI, embedding, and multitenancy related capabilities. Software partners can evaluate placing analytical entry points in their own products. Implementation should verify user identity, tenant scope, authorization, model services, and the specific integration method.