The same topics as the in-product Help guide — written for people who want a clear path from first login to reliable answers.
Guide / Capabilities
Capabilities
BEAS AI Report is built for business teams who need trustworthy answers from their own data —
without waiting on a custom BI project for every question. Below are the product capabilities in
everyday language (not implementation detail).
Ask in plain language
Leaders and analysts type the question they already ask in meetings.
Answers come back as text, tables, charts, or KPI summaries — in the same conversation.
Outcome: faster ad-hoc reporting without a ticket queue.
When a chart or table is right, save it as a widget or place it on a dashboard.
Shared metric definitions keep the same number consistent across boards.
Outcome: one definition of “revenue” (or any KPI) everywhere.
Assemble dashboards with a common date range, click-to-filter, drill-down, export to PDF/Excel,
and optional share links for stakeholders who only need to view.
Outcome: a living operating board instead of static slide decks.
Upload approved policy PDFs, describe what each data set means, map how tables relate,
and save successful questions as examples. The assistant prefers your vocabulary over guessing.
Outcome: answers that match your status codes, KPIs, and joins.
Load files (CSV / Excel) or connect operational databases. Prepare reporting data sets with
approval steps and privacy rules before people ask questions on them.
Outcome: reporting on governed data you already own.
Schedule monitors on the metrics that matter. Review open incidents, explain what changed,
and keep humans in the loop before any operational change is applied from chat.
Outcome: earlier awareness without another monitoring silo.
Ask for a forecast in Report. Review the plan, approve the use of sample history when required,
then see past versus predicted values in the same thread.
Outcome: directional outlooks with clear consent and audit.
Define a recurring summary (weekly KPIs, risk digest, channel performance) and deliver it
to the inbox of the people who need it — without rebuilding the report each time.
Outcome: proactive reporting on a cadence leadership already expects.
Each customer owns a workspace with its own AI connection and reporting data.
After you sign in you can ask questions in plain language, save answers as widgets, put them on dashboards,
watch for anomalies, and schedule emailed summaries.
First path (new workspace)
Register and verify — create the account, open the verification email, then sign in.
Land on Home — a new workspace shows setup incomplete until LLM and data connections pass.
Connect LLM + AI Report DB — finish the setup wizard or Workspace settings; run both connection tests.
Teach the AI — load data, then add PDFs, matching tasks, and examples under AI Context.
Ask in Report — save a good answer as a widget, then build a dashboard.
Invite the team — pick a role for each person from Workspace.
Where do I go to…
I want to…
Go to
See if everything is healthy
Home
Ask a question about my data
Analyze → Report
Reuse a chart or table
Widgets & Queries
Build, share or export a board
Dashboards
Upload CSV / Excel
Data Sources → Data Ingest
Help the AI understand my data
AI Context
Change LLM key or invite people
Workspace
SaaS sign-in uses username and password. Create a workspace on register, or join through an invitation.
Email verification is required before the first sign-in.
Registration owner becomes Tenant admin on the Free plan.
Verify via inbox link, or resend from the verify-email screen.
Forgot / reset password is available when mail is configured.
Invitation links let teammates join an existing workspace without creating a new one.
Chat and live boards unlock after the workspace connects an LLM and an AI Report DB (OpenSearch).
You can finish later with I’ll configure later, then complete under Workspace settings.
Add your LLM provider, base URL if needed, models, and access token.
Choose BEAS-managed OpenSearch or your own external cluster.
Run Test LLM and Test OpenSearch until both show OK.
Workspace (Management) is where Tenant admins manage AI connections, data connection mode,
members, and invitations after the first setup.
Update LLM provider, models, and rotate the API key.
Switch between BEAS-managed and external OpenSearch; re-run connection tests.
Invite members and assign Tenant admin, User, Reporter, or Viewer.
Package limits (storage, chat quota) surface here when applicable.
The side menu hides what you cannot use; the API still checks every change.
Platform admins operate Super Admin; everyone else works inside a workspace role.
Role
Intent
Tenant admin
Owns the workspace — setup, invites, approvals, full product
User
Analyst builder — Report through data pipeline, AI Context, logs
Consumes boards and Report history — no create, share, or export
Platform admin users land on Super Admin — not tenant Report.
They manage packages, mail, every workspace’s BYO settings, users, invitations, and operational logs.
Platform dashboard — users by package and expected MRR
Assign Free / Enterprise; record manual payments
Mail settings, workspaces, users, invitations, audit logs, business errors
Everyone lands on Home after login. It shows workspace health, items that need attention,
and shortcuts into the product.
Critical banners when setup is incomplete or LLM / OpenSearch checks fail.
Quota near-limit warnings for Free plan workspaces.
Quick links into Report, dashboards, and documentation.
Report (Analyze → Report) is the natural-language workspace.
Ask in plain language; answers return as text, tables, charts, HTML, JSON, or drafts you can save.
Asking well
Pattern
Example
Metric + time
Total billed amount in August 2025
Breakdown
Top 10 products by revenue last 90 days
Compare
This month vs last month order count by channel
Forced table
/table Refund rate by reason last 30 days
Scoping pickers
Above the input, optional pickers narrow what the assistant may use: indices,
data adaptors, and OpenAPI catalogs.
Leave them empty to let the assistant choose; chips show what won for that turn.
On the screen
New chat / Recent history with search and filters
Microphone for speech-to-text where the browser supports it
Type / for output format shortcuts (table, chart, KPI text, …)
Save strong answers as widgets or as Chat Query Examples (AI Context)
Dashboards combine widgets into boards with date range, cross-filter, drill-down, export, and share links.
Create from templates (Blank, Executive summary, Classic, Wide rows) or free-form.
Viewers can open boards and change the date range; they cannot create, share, or export.
Export menu covers PDF, Excel, print, and scheduled delivery when mail is configured.
Public share links can open without login when you enable Share.
Widgets & Queries stores reusable visualizations and shared metric definitions (Dashboard Queries)
that dashboards and Report drafts can reuse.
Widgets — charts, tables, KPIs bound to an index or query.
Dashboard Queries — shared DSL / metric definitions across widgets.
Save from Report chat when an answer looks right.
Index policy may require approval before a widget query can run in production.
Anomalies watch your data on a schedule, raise incidents, and can propose remediation drafts for review.
Tabs: Definitions, Incidents, Runs, and optional OpenSearch AD detector wiring.
Available to Tenant admins and Users — not Reporters or Viewers.
Ask Report to list open anomalies or explain an incident by id.
Insight Jobs schedule AI-written summaries and email them to stakeholders.
Define cadence, audience, and the question / board context to summarize.
Requires working mail settings on the platform and a healthy LLM connection.
Tenant admin and User can manage jobs; recipients only need inbox access.
Data Sources is the hub for how data enters the workspace.
Tab
Purpose
Data Adaptors
PostgreSQL, MySQL, MongoDB connections for prep and read-only use
Data Ingest
Upload CSV / Excel into an OpenSearch index
Ingest Settings
Index access, PII rules, schemas (Tenant admin)
New indices follow the workspace naming prefix. Deleting an ingest job removes the OpenSearch index as well.
Data Prep transforms adaptor data into reporting indices with visual jobs, runs, and index approvals.
Design a job with source adaptor nodes and transforms.
Run and inspect results; fix failures from run history.
Submit the target index for approval when governance requires it.
AI Context teaches the assistant your business language, field meanings, joins, APIs, and proven queries.
Without it, the AI only sees raw field names and the prompt.
Tab
You provide
Business rule PDFs
Glossaries, status codes, KPI definitions
Matching tasks
Index field meaning for reporting
Adaptor discovery
SQL / MongoDB object meaning
OpenAPI Catalogs
API operations chat may call
Schema relations
Join / semantic graph — Analyze, edit, Publish
Chat Query Examples
Question → query pairs that worked
Upload vocabulary PDFs.
Run matching / adaptor discovery for every source people will ask about.
Register APIs if chat should read or update records.
Analyze schema relations, fix the graph, then Publish to RAG.
Save strong chat answers as examples over time.
Logs help you answer “who changed what?” and “why did this fail?”
Audit logs — security-relevant actions across the workspace.
Business errors — failed operations with detail for support and debugging.
Tenant admins and Users can open Logs; Reporters and Viewers cannot.
