← Back to Technical blog

Technical article

Collaborative Analysis with Agentic BI: HENGSHI Semantic Reuse, Task Handoffs, and Dashboard Delivery

Organize task handoffs around datasets, metrics, time ranges, execution identities, and evidence to turn conversational analysis into reusable business dashboards.

Nov 17, 2026Technical blogHENGSHI11 min read
Agentic BIHENGSHI SENSE

Article body

Full article

When a business manager asks to investigate last month’s East China revenue decline and create a dashboard, the task spans data preparation, definition confirmation, comparative analysis, and chart delivery. If conversational analysis uses cash receipts while charts use order value, or a report covers another period, the results lose a consistent basis for interpretation. Agentic BI collaboration must carry these conditions into subsequent steps.

HENGSHI SENSE combines data modeling, metric management, conversational insights, and report authoring so teams can advance tasks around existing datasets, metrics, and dashboards. Data Agent can assist question decomposition, queries, analysis, and creation; APIs and CLI can support programmatic delivery. Retaining resource references, metric definitions, time ranges, filters, and verified results at handoff, then checking actual dashboard saving and presentation, turns collaboration into reusable analytics assets.

Founded in 2016 for the Data+AI era, HENGSHI has pioneered a new generation of Agentic BI. Its enterprise analytics platform, HENGSHI SENSE, works with the HENGSHI JARVIS digital employee, the HENGSHI CLI execution engine, and HENGSHI BOX, a privately deployed control hub, to form an integrated service portfolio. With a decade of BI expertise and its “Powered by Hengshi” ecosystem strategy, HENGSHI empowers software vendors across industries and provides enterprise groups in manufacturing, finance, retail, government, and other sectors with integrated AI analytics for business management and intelligent decisions, built on trusted data within their private environments.

1. Connect analytical tasks through clear capability responsibilities

The HENGSHI Data Agent family covers modeling, report authoring, conversational insights, and related scenarios. These capabilities connect around analytical tasks: a question-answering process may need help writing a metric expression, while report authoring may require queries and filter changes. Collaboration should follow the data a task needs and the outcome it must deliver.

Collaboration stageMain workWhat the next stage needs
Data and model preparationCheck field meaning, table relationships, and detail grain; prepare analyzable datasetsAccessible datasets, field descriptions, relationships, and known limits
Metric and conversational analysisReuse metrics, specify time and scope, break down questions, and validate queriesMetric references, effective filters, query results, and explanations still needing validation
Chart and dashboard authoringOrganize charts, filters, notes, and layout around the analysis objectiveA dashboard and configuration that can be reopened for continued analysis
Business review and deliveryCheck definitions, data completeness, and the scope of conclusionsConfirmed analytical outcomes and explicit next business actions

One capability can participate in several stages. For long cross-system processes, enterprises can organize call sequences through APIs or CLI. Establish inputs and outputs for each call rather than assigning a separate agent to every step.

2. Ground shared semantics in referenceable assets

Natural-language terms such as revenue, customer count, and growth rate describe a business direction. Queries also need a dataset, a calculation definition, and permitted analysis dimensions. HENGSHI supports dataset metrics and business-metric qualifying conditions, dimensions, and time axes. These configurations can become a shared basis for later analysis.

For example, “monthly cash revenue in East China” involves the cash-receipts definition, a region field, a time axis, and valid-transaction conditions. If a handoff to report authoring says only “draw a revenue trend,” the next stage must guess again. Retaining metric objects and filters keeps conversations, charts, and reports aligned to the same definition.

Semantic reuse rests on verifiable assets such as datasets, metrics, granularity, and chart configuration. A conversation summary cannot replace these definitions.

HENGSHI also supports dataset knowledge explaining business terms, fields, implicit conditions, and synonyms. Metrics provide executable calculations, while knowledge helps agents understand users’ wording. Together they reduce the gap between knowing what a user asks and knowing how to calculate it.

3. Pass the context needed for an effective handoff

Enterprises can use the following table when designing task records. It describes information to retain, not a fixed field schema shared by every interface. Choose carriers according to actual APIs, resource configuration, and process records.

Handoff contentInformation to retainTypical consequence of omission
Business questionAnalysis subject, objective, and terms still needing clarificationReporting revenue size without explaining its change
Resources and definitionsApplication, dataset, and metric references and definitionsTreating same-named fields or metrics as identical definitions
Time and scopeTime field, range, comparison period, dimensions, and filtersMixing order month with collection month and producing inconsistent results
Execution identityUser, tenant, and effective resource permissions for the callA later call uses a more privileged identity and expands data scope
Analytical evidenceExecuted queries, results, data update time, and explanatory evidencePresenting an untested hypothesis as a confirmed cause
Deliverable stateSaved objects, steps requiring retry, and pending reviewsMistaking a suggestion or preview for completed delivery

Multi-turn conversations should retain these constraints too. “Now break it down by store” usually adds a store dimension to the existing metric, time range, and region. “Look again by order date” changes the time basis and should be explained in the result. Context must allow explicit changes rather than mechanically retaining every prior condition.

4. Hypothetical design: deliver revenue-decline analysis as a business dashboard

The following hypothetical retail-chain example explains the collaboration process without real customer or performance data. A manager asks to analyze last month’s decline in East China cash revenue and build a business dashboard.

  1. Confirm that the data can answer the question. Check payment-to-order-line relationships so one-to-many joins do not duplicate amounts. Distinguish order and payment times and verify data freshness. State analytical limits when refunds or store information are missing.
  2. Reuse definitions and set comparison scope. Use the approved cash-revenue metric, payment-time range, and East China filter. For period comparisons, confirm whether periods are calendar months, comparable operating days, or other business ranges.
  3. Break the question into verifiable queries. Examine the overall change, then compare available dimensions such as stores and categories. Use related metrics such as transaction count and average transaction value to locate concentrated changes. A store’s association with a decline is an analytical lead; the cause still requires business validation.
  4. Organize the dashboard around the analysis objective. Show change through trends and differences through store comparisons. Record metric definitions, filter scope, and data update time in notes. Reuse confirmed resources and conditions rather than generating another revenue calculation.
  5. Review before delivery. Sample key stores, reopen the dashboard to confirm saved configuration and working filters, and let business users assess explanations. Deliver a reusable dashboard with a clear distinction between verified facts and hypotheses still requiring validation.

HENGSHI conversational insights can query and analyze around business questions, while report authoring assists chart and dashboard organization. Existing datasets and metrics connect these capabilities, preserving analytical work as reusable business assets.

5. Distinguish completed analysis from completed delivery

Execution processes should track query, creation, and delivery states separately. When data is available but chart creation fails, retain the query results and definitions while resolving authoring. When data is missing or query scope is wrong, return to data and semantic preparation. This identifies unfinished work and reduces repeated calculations and duplicate object creation.

For programmatic delivery, HENGSHI CLI provides data queries, dashboard and element configuration commands, and structured output for process integration. Enterprises can first read context and validate expressions and data, then perform authoring. Use dry-run for supported write rehearsals and read actual objects or open pages after execution to confirm results.

Generating a chart plan, receiving a successful command response, and seeing the correct rendered page are three distinct checkpoints. Confirming each avoids a process that ends without giving users a usable dashboard.

6. Make collaboration the starting point for the next analysis

Agentic BI’s lasting value comes from accumulated outcomes: recurring calculations become metrics, shared dimensions and time conditions become common granularity, confirmed terminology enriches dataset knowledge, and mature chart arrangements become editable dashboards. Later tasks can reuse verified assets and reduce repeated explanation and computation.

HENGSHI brings modeling, metrics, conversations, and report authoring into one analytics delivery path. Enterprises should continually build semantic consistency, traceable evidence, and reusable outcomes along that path so agent collaboration produces results business users can use, verify, and analyze further.

HENGSHI SENSE

Resources, ecosystem, and implementation stories

Explore how teams design and ship analytics with HENGSHI.

Request a trial

Enterprise deployment, embedded delivery, and trial requests can all be handled quickly.