Dashboard
Four year-scoped tiles — saved reports, runs today, average time, success rate — with trends, the ten most recent runs and quick actions into every other screen.
Ad-Hoc Reporting: Pick the Tables, Build It or Ask It, Keep the Answer
An admin reporting workbench over the tables the platform has already published — eight data domains, a visual builder with joins and filter trees, a plain-language mode that writes one read-only statement you can inspect, and a library where every saved report keeps both its definition and its last result.
BR Ad Hoc never offers a table that has not been catalogued with labels and field metadata, runs nothing but reads, and stores a saved report as normalized rows — tables, joins, columns, filters, group-by, order-by — rather than a blob. It reports on SuccessFactors-fed records and on records that arrived from Workday, UKG, ADP, Dayforce or Paycom through HRIS Sync, in the same builder, with the same rules.
8 data domainsVisual or plain languageRead-only · SELECT onlyDefinition + result, both keptAny HRIS records
Eight Domains, Two Ways to Ask, One Read-Only Engine
An administrator chooses one or more published data domains and the tables inside them, then either assembles columns, joins and filters by hand or asks a question in English. The platform runs a read-only select against those tables. A save stores the definition as normalized rows and the result as a run with every cell — while a plain-language run stays in the browser until someone decides it is worth keeping. That is the whole app, and everything on this page is one of those steps looked at closely.
DashboardThe year's runs, reports and success rateSet upDomains and published tables, visual or AIBuild itSelect · join · where · group · orderAsk itOne question, one read-only statementKeep itResults, library, templates
8published data domains — Benefits, Intelligent Services Delivery, Payroll Core, Integrations & Payroll Reconciliation, HRIS Master Data, ACA Reporting, Finance & Controlling, BR Configuration
6builder tabs — Select, Join, Where, Group by, Order by, Advanced — each with suggestions that are rule-built and test-executed before they are shown
5,000rows at most on a visual run, whatever the header says; a plain-language run is prompted for the first 100
0statements that write — the visual path compiles a select and the AI path is rejected unless it begins with SELECT
Working Flow — Who Clicks What, in Order
The whole app in ten steps, exactly as it runs: one dashboard, one wizard, two ways to build, one read-only run, one library. The right-hand column is the rule the platform enforces at that step.
| # | Who · where | What happens | The rule at this step |
|---|---|---|---|
| 1 | Admin opens BR Ad Hoc Dashboard | KPI tiles for the year — saved reports, runs today, average time, success rate — plus the ten most recent runs and quick actions into the builder, AI mode, the library and templates | Trends compare to the prior year to date; runs today compare to yesterday |
| 2 | Starts a new report Setup wizard · step 1 | Picks one or more data domains and at least one published table inside them | Only tables in the metadata catalog are offered — never a raw database table |
| 3 | Chooses how to build Setup wizard · step 2 | Visual builder or plain-language AI; the choice can still be switched here | Opening a library or template item skips the wizard entirely |
| 4 | Builds it by hand Visual builder | Select · Join · Where · Group by · Order by · Advanced tabs; per-tab suggestions are rule-built from field metadata and test-executed before they are offered | A join must start from a table already in the chain and add a new one; aggregation forces every other selected column into the group list |
| 5 | Runs it Result panel | The definition is compiled into one read-only select; the run is stored with every result cell; the first 50 rows come back inline, the rest page from the run | Row limit is validated up to 100,000 on the header and clamped to 5,000 at execution |
| 6 | Or asks it AI query | Types a question; a relevance check runs first; the schema is built from field metadata; a commercially available language model writes one SELECT limited to 100 rows; a security check rejects anything that is not a read; the rows stay in the browser | Nothing is written to the run history until the admin saves — the save turns the executed statement into the same structured definition the visual builder uses |
| 7 | Saves it Library | The definition is validated against the catalog, stored as normalized rows — tables, joins, columns, filter tree, IN-lists, group-by, order-by — then run once and page 1 loads into the grid | If that first run fails, the report stays saved and the message says so; AI-built reports carry an AI-generated flag |
| 8 | Reads the result Table · Charts · Insights · Export | Six chart types built server-side from the stored cells; an Insights tab with average, min, max, outliers and empty-cell counts on the rows shown; CSV or JSON export of a stored successful run | Charts need two columns and at least one numeric; export is for stored runs, so an unsaved AI run is saved first |
| 9 | Comes back to it Report Library | Run, edit in builder, duplicate, favourite, delete, edit metadata, preview the presentation SQL; tabs All · Visual Builder · AI Generated · Favorites; search by name, description, tags, category and source table | Metadata edits change the header only and never re-run the report; delete cascades runs, cells and every child row |
| 10 | Starts from a template Report Templates | Run a template in place, or Customize — a duplicate that becomes a new library report on its next save; categories follow the first table's domain | A template is a saved report with a template flag; there is no second data model |
One Report, Six Rooms
Follow a benefits administrator answering “which employees waived medical this year, by location?” — from the dashboard, through the wizard, built by hand and then asked in plain language, into the library — and the same report over an entity that runs Paycom.
The dashboard — the year at a glance, then straight into work.
The home screen opens on four tiles computed for the calendar year: saved reports created this year, runs started today, the average execution time of successful runs, and the success rate. Trends compare to the prior year to date, and runs today to yesterday; for time, lower is better and the arrow says so. Beneath them sit the ten most recent runs, newest first, and four quick actions — the visual builder, the plain-language mode, the library and the templates — plus the shared application launcher and the Essential Insights strip every admin app carries.
What the dashboard shows
- Saved reports this year — active, non-template definitions
- Runs today, average run time and success rate, each with a trend
- The ten most recent runs with status and row count
- Quick actions into the builder, AI mode, library and templates
Why it matters
An administrator sees whether the reporting estate is healthy before opening a single report.
Deeper dashboard charts — run trends, status mix, top tables, execution-time buckets — are computed by the service and not yet placed on this home screen; they are a roadmap item, not a claim.
The wizard — domains first, tables second, then how to build.
A new report starts with two questions. First: which data domains, and which published tables inside them — Benefits, Intelligent Services Delivery, Payroll Core, Integrations & Payroll Reconciliation, HRIS Master Data, ACA Reporting, Finance & Controlling, BR Configuration. The table list comes from the catalog for that domain, each with its label, description and category, and the builder never sees a table the catalog does not hold. Second: build it visually or ask it in plain language. The visual and AI entry points pre-select a mode, and the administrator can still switch here.
What the wizard fixes
- The domains and tables the report may read — nothing outside them
- The mode — visual builder or plain-language query
- Each table's label, description, category and row count, from the catalog
- Skipped entirely when a library or template item is opened
Why it matters
Scope is decided before the first column is chosen, so a report cannot wander into tables it was never meant to read.
Metadata administrators publish tables and fields in configuration — labelling, activating and assigning a domain — before this app can offer them.
The visual builder — six tabs that compile to one read.
Select picks columns from the field catalog with an aggregation — none, sum, count, average, min, max. Join chains tables with inner, left, right or full joins on one field pair; each join must start from a table already in the chain and add a new one. Where builds a filter tree of AND/OR groups and typed conditions. Group by is enforced the moment an aggregation appears. Order by sorts; Advanced holds distinct and the row limit. On every tab a suggestion panel offers candidates built from field metadata and test-executed first — a suggestion that fails is never shown. Run compiles the definition into one select; Save validates it, stores it as rows and runs it.
What each tab decides
- Select — columns and aggregation; count with no field is a row count
- Join — inner, left, right, full; one equality per join; no reuse of a joined table
- Where — a tree of groups and eleven operators; IN values kept as rows
- Group by, Order by, Advanced — grouping enforced, sort direction, distinct, limit
Why it matters
A report built here is always a valid, read-only statement — the builder will not let an invalid one exist.
Employee code fields — department, location, company and the rest — are joined to their text automatically, so the grid shows names; under a group-by the text is aggregated correctly.
Plain language — one question, one inspectable read, nothing stored until you say so.
The administrator types the question. A relevance check runs first: a question the selected tables cannot answer comes back with an explanation and no database read at all. Otherwise the platform builds a schema description from the field catalog — long text and binary columns omitted — adds an effective-date hint and, if display columns were chosen, requires them in the select. A commercially available language model writes one statement limited to the first 100 rows. A security check rejects anything that does not begin with SELECT. The statement runs, the rows come back to the browser, and the executed statement is shown. If it fails, a second model call explains whether the question can be answered. A best-effort pass turns the statement into the same structured definition the visual builder uses — and only then does Save appear.
What happens to a question
- Relevance check — irrelevant questions never touch business tables
- Schema from the field catalog; effective-date hint; chosen columns required
- One SELECT, first 100 rows; rejected unless it begins with SELECT
- Rows stay in the browser; Save converts the statement into a definition and flags the report AI-generated
Why it matters
You can read the statement that answered you, and nothing is kept until you choose to keep it.
Question suggestions in the side panel are templates built from amount fields, key fields and categories, then test-run — this is a query mode, not the employee Athena chatbot, and it is deliberately isolated from it.
Results, library, templates — what you keep, and how you come back to it.
A run opens in three tabs. Table pages the stored cells — the first 50 rows arrive with the run, the rest are fetched as you page. Charts builds column, bar, line, pie, scatter or area from the stored cells on the server, first column as the category, later numeric columns as series. Insights shows quick statistics on the rows on screen — average, minimum, maximum, a count of outliers beyond two-and-a-half times the average, and empty cells; it is arithmetic, not the BR Insights explanation engine. Export writes CSV or JSON from a stored successful run. The library lists every active saved report with run, edit, duplicate, favourite, delete, metadata and SQL preview; templates are saved reports with a template flag that Customize duplicates into a new report.
What the panel and library give you
- Table, Charts (six types) and an on-screen Insights tab of quick statistics
- Export CSV or JSON of any stored successful run, walked from the cells in batches
- Library tabs All · Visual Builder · AI Generated · Favorites; search by name, tags, category, source table; sort by popular, fastest, recent, records, favourites
- Templates by category — enrollment, compliance, finance, HR, payroll, service delivery — run in place or customized into a new report
Why it matters
A report is something a colleague can open next quarter and get the same definition and the last kept result — not a spreadsheet on a laptop.
The SQL preview in the library is a readable presentation of the saved definition; the run log itself stores a summary of what ran, not a statement to paste elsewhere.
The same builder over SuccessFactors records and every other HRIS.
The tables in the catalog are the platform's own — elections, dependants, deductions, invoices, employee-months, employees and organizational units — filled from SAP SuccessFactors in real time through OData and Intelligent Services Center, and from Workday, UKG, ADP, Dayforce or Paycom through HRIS Sync with the same fields and provenance. An administrator reporting on a mixed estate picks the same domains, joins the same tables and asks the same questions; the entity is a column and a filter, not a different tool. Employee code fields resolve to text from the same master whichever HRIS supplied the code.
How sources appear in a report
- One employee table, one elections table — for SuccessFactors and every other source
- Entity, legal entity and HRIS source available as columns and filters
- Code-to-text for department, location and company regardless of the feeding system
- Role-based permissions on the app and its data, resolved through SSO from each HRIS
Why it matters
A group that runs three HRIS platforms builds one report, not three.
Rows from a feed that has not confirmed yet carry their sync state, so a report can show — or exclude — what HRIS Sync has not yet reconciled.
Rules of the Builder — What the Platform Enforces
Ad-hoc does not mean unguarded. Every report, visual or plain-language, passes the same set of rules before a single row is read.
| Rule | How it works | Why it is there |
|---|---|---|
| Which tables can be queried | Only rows in the published table catalog, each with a friendly label, technical name, schema and domain; fields come from the field catalog with type, key and required flags | The builder cannot reach a table nobody has labelled and assigned to a domain |
| Joins | Inner, left, right or full; each join starts from a table already in the chain and must add a new table — reusing one is rejected | No accidental self-joins or loops; every chain is a path |
| Filters | A tree: groups combine with AND or OR; conditions use =, ≠, <, >, ≤, ≥, LIKE, BETWEEN, IN, IS NULL, IS NOT NULL; IN-list values are stored one per row | A filter is data, not a string — it can be edited, previewed and audited |
| Group by | If any selected column is aggregated (SUM, COUNT, AVG, MIN, MAX), every non-aggregated selected column must be in the group list | A visual report never compiles into an invalid statement |
| Row limit | Header limit validated from 1 to 100,000; execution clamps to 1–5,000; the AI path is prompted for the first 100 rows | A report is a working set, not a bulk extract |
| Reads only | The visual path compiles a select; the AI path must begin with SELECT — a leading UPDATE, DELETE, DROP, INSERT, ALTER, CREATE, TRUNCATE, EXEC, MERGE, GRANT or REVOKE is rejected before anything runs | Nothing in this app can change a benefits record |
| Employee codes as text | Department, location, company and the other employee code fields are joined to the code-to-text master automatically and shown as text when it exists | A report reads as words, not codes — and groups by the text correctly |
| Effective dates in AI mode | Words like history, past, prior, last year, expired or archive ask for expired rows; every other question asks for current rows — only on tables that carry begin and end dates | A plain-language question about “current” enrollments does not silently return history |
| Save and re-save | A definition is validated against the catalog before it is stored; on update every child row is replaced and the header keeps its id; saving never deletes previous runs | History survives edits; an unresolvable definition is refused, not half-saved |
| Run history | Runs of a saved report accumulate; an unsaved visual run replaces the previous unsaved run for that person; an AI run is browser-only until saved | The run log holds what people chose to keep, not every experiment |
The Six Screens of BR Ad Hoc
One admin app, six screens, one read-only engine underneath — described exactly as they run.
Setup Wizard
Domains and published tables first, visual or plain-language second; skipped when a saved report or template is opened.
Visual Builder
Select, Join, Where, Group by, Order by and Advanced with rule-built, test-executed suggestions; compiles to one read; Save validates, stores and runs.
AI Query
Relevance check, schema from the catalog, one SELECT limited to 100 rows, a security check, the statement shown; browser-only until saved.
Results Panel
Table paging from stored cells, six server-built chart types, an on-screen statistics tab, CSV and JSON export of stored successful runs.
Library & Templates
Run, edit, duplicate, favourite, delete, metadata and SQL preview; tabs by mode and favourites; templates by domain category that customize into new reports.
Export-and-Spreadsheet Reporting vs. BR Ad Hoc
Most benefits reporting still means an export, a pivot table and a file nobody can reproduce. Here is what changes when the report lives on the platform.
| The moment | Export and spreadsheet | BR Ad Hoc |
|---|---|---|
| “Which tables can I use?” | Whatever the export happened to contain. | The published catalog — eight domains, labelled tables and fields, nothing raw. |
| Joining enrollment to payroll | VLOOKUP, and hope the keys match. | A typed join from a table already in the chain, one field pair, validated before it runs. |
| “Can I just ask?” | No. | One plain-language question → one read-only statement you can read, first 100 rows, saved only if you say so. |
| A filter on department | Codes, copied by hand. | Codes shown as text from the master, and grouped correctly. |
| Someone edits the sheet | The number changes and nobody knows. | The definition is rows; the run keeps every cell; a save never deletes history. |
| “Send me last quarter's version” | A file on a laptop. | The library — last result stored, run again on request, duplicate to customize. |
| An entity on another HRIS | A second export from a second system. | The same tables, with entity and source as columns, through HRIS Sync. |
AI query mode — not a chatbot
Ask the Tables and Read the Statement That Answered
The plain-language mode is a query engine with a language model in one step of it. It checks relevance, builds the schema from the catalog, writes one read-only statement, refuses anything else, runs it and shows you the statement beside the rows. It does not remember your last employee conversation, it does not act, and it is deliberately isolated from the employee Athena assistant.
- “Which employees waived medical this year, by location?” — one statement over Elections and Employee, codes shown as text
- “Show the ones who moved PPO to HDHP in July” — the effective-date hint keeps it to current rows
- “Only the ones with a dependant over 26” — a new statement, still read-only, still first 100 rows
- “Save this as a report” — the statement becomes a definition, flagged AI-generated, run properly and kept
- “What's the average deduction here?” — answered from the rows on screen by the Insights tab — arithmetic, no model
Which employees moved from PPO to HDHP in July, by location?
AI Query · relevant · one read-only statementGenerated over Elections and Employee (HRIS Master Data) and ran it: SELECT TOP 100 Employee.Location, Employee.Id, Elections.PriorPlan, Elections.NewPlan … WHERE prior plan is PPO and new plan is HDHP and the effective date is in July 2026 — current rows only. 412 employees match; the first 100 are shown, grouped by location text. The statement is visible above the grid.
Show statementSave as report
Only the ones with a dependant over 26, and save it.
AI QueryRewrote the statement with a join to Dependants and a birth-date condition: 23 employees. Saved as “PPO → HDHP movers with dependants over 26” — the statement was converted into a definition (two joins, three conditions), flagged AI-generated, and run once so the result is stored. It is now in the library under AI Generated.
Open in builderExport CSV
Ask a question about the selected tables…
Why Benefits, Payroll and Finance Teams Choose BR Ad Hoc
A Catalog, Not a Database
Only published, labelled tables and fields — eight domains an administrator can actually read.
Build It Without a Data Scientist
Joins, filter trees, grouping and sorting that the builder keeps valid, with suggestions that were tested before they were offered.
Ask It and See the Query
One plain-language question, one read-only statement shown beside the rows, saved only when you choose.
Reads Only, by Construction
The visual path compiles a select; the AI path is refused unless it is one — nothing here can change a record.
Definition and Result, Both Kept
A saved report is rows you can edit and audit; a run keeps every cell; history survives every re-save.
One Builder, Any HRIS
SuccessFactors records and Workday, UKG, ADP, Dayforce or Paycom records in the same tables, with entity as a column.
Four Reports, End to End — How the Screens Fit Together
The screens are one path. A report starts on the dashboard, is scoped in the wizard, built or asked, run and read, then kept in the library where it can be re-run, edited, duplicated or turned into a template for the next person. Four stories show the path in use.
- The Open Enrollment check — Benefits picks BF + HR, builds waived-medical-by-location visually, saves it, favourites it and re-runs it every Monday of the window — same definition, a new stored result each time
- The payroll question nobody had a report for — Payroll picks PC + BF and asks “deductions that don't match an active election this period” — reads the statement, saves it as an AI-generated report, exports CSV for the reconciliation desk
- The ACA month under threshold — Compliance opens the ACA template, runs it in place, then Customize — a duplicate that becomes their own report with one extra filter on the ALE member
- The audit request — Finance opens a favourite from last quarter: the last stored result is still there, the SQL preview shows the definition in readable form, and a fresh run is one click if they want today's rows
- The mixed-HRIS group — HR adds Entity as a column to the same waived-medical report and runs it across the SuccessFactors plants and the Paycom distribution centres in one result
Bring the Report You Build in Excel Every Month. Build It Once Here.
Tell us which export you pivot every month — waivers, deductions, invoices, ACA months. We'll pick the domains in the wizard, build it visually on your tables, ask the same question in plain language, show you the statement, save both, and export the result — in your tenant.
Published catalog · 8 domainsRead-only · SELECT onlyDefinition + result keptStatement always visible
Ad-Hoc Reporting FAQs
The administrator reporting workbench inside Benefits Reimagined. An admin picks one or more published data domains and tables, builds a report visually or asks a question in plain language, runs a read-only select against those tables, and saves the definition — as normalized rows — and the result, as a run with every cell kept. A report library and templates let the report be re-run, edited, duplicated, favourited and exported.
Eight domains: Benefits, Intelligent Services Delivery, Payroll Core, Integrations & Payroll Reconciliation, HRIS Master Data, ACA Reporting, Finance & Controlling and BR Configuration. Inside each, only tables that a metadata administrator has published — labelled, activated and assigned to a domain — with their fields catalogued. The builder never offers a raw database table.
No. The visual builder compiles a select statement. The plain-language mode runs a security check that rejects any statement that does not begin with SELECT — a leading UPDATE, DELETE, DROP, INSERT, ALTER, CREATE, TRUNCATE, EXEC, MERGE, GRANT or REVOKE is refused before anything runs.
A relevance check runs first; an unanswerable question comes back with an explanation and no database read. The platform then builds a schema description from the field catalog, adds an effective-date hint, requires any display columns you chose, and a commercially available language model writes one statement limited to the first 100 rows. The statement is checked, run, and shown beside the rows. It stays in the browser until you save it, at which point it is converted into the same structured definition a visual report uses and flagged AI-generated.
A visual run is clamped to 5,000 rows at execution whatever limit the header carries (the header accepts up to 100,000). A plain-language run is prompted for the first 100 rows. Ad Hoc is a working-set tool; bulk extracts belong to exports and integrations.
Two things, separately. The definition — tables, joins, columns, filter tree, IN-list values, group-by and order-by — as rows, validated against the catalog before it is stored; updates replace the child rows and keep the report's id. And the run — start, end, duration, record count, status and every result cell — with previous runs of a saved report kept. Saving never deletes history; deleting a report cascades everything.
No. The results-panel tab computes average, minimum, maximum, an outlier count and empty-cell counts on the rows shown — plain arithmetic, no model. BR Insights is the platform's explanation engine for computed metrics on the analytics workspaces; it is a different thing and lives on its own page.
Yes. The catalog tables are the platform's own, filled from SuccessFactors in real time and from Workday, UKG, ADP, Dayforce or Paycom through HRIS Sync with the same fields. A mixed group builds one report with entity and source as columns, and employee codes resolve to text from the same master whichever system supplied them.
Build It or Ask It. Either Way, Keep It.
Published tables, a builder that stays valid, a question mode that shows its statement, and a library that keeps both the definition and the result — over SAP SuccessFactors records and every record from another HRIS.