Nexourz BIThe BI platform that does the analyst’s job
Brief it like an analyst — in plain words, even out loud — and Nexourz models the data, builds the dashboards, mails the Monday report, watches the numbers, and answers the questions. The work you were about to hire for, with every number arriving provenance attached — and a nightly self-audit to prove it.
Runs on AWS · your data stays in your account on Enterprise
The problem
BI tools don't do the work. People do — expensively.
The tool is the small half of the cost.
Tableau and Power BI licences are cheap next to the analysts and BI developers who feed them. You are not buying software, you are staffing a function.
Dashboards rot silently.
Someone upstream renames a column. Nothing errors — the number reads as empty, aggregates away, and the dashboard renders a confident zero until it comes up in a meeting.
Numbers arrive without receipts.
Two dashboards disagree, nobody can say which definition of margin is the real one, and the argument moves from the data to the people who produced it.
Joins fail invisibly.
One duplicated key multiplies every matching row. No error, no warning — just a bigger number. We measured a real case at fourteen times.
The typical fully-loaded cost of one BI developer — modelling the data, maintaining it as sources change, watching the numbers, and answering “what happened?”. That job is the layer we automate.
Verified, not asserted
Numbers we can defend in a meeting
- 2,000,000
- rows in the live parity test — warehouse SQL against our reference engine, zero mismatches
- 14×
- revenue overstatement our join guard refused, measured on real data rather than imagined
- 1,322
- automated tests, run on every change before anything ships
- nightly
- the platform audits itself: every metric recomputed two ways, and the AI's answers marked against the engine — disagreements published, not averaged away
The core idea
We automated the job description, not just the charts.
State the problem
Brief it like you'd brief an analyst. In words.
Type — or say — "we need to watch dealer margin erosion and catch discount-driven losses." Nexourz restates the goal, proposes the measures with the reason each one answers your problem, designs the dashboards, and lists what your data cannot answer. One admin signature builds all of it.
The open-questions list is the point: on a real client workbook it asked whether discount is stored as a fraction or a percent, and flagged that true gross margin was not derivable without a cost column.
Model the data
Your data is not one flat table. Ours knows that.
Point Nexourz at a warehouse and it reads the foreign keys your database already declares, then falls back to naming conventions, then asks the model for whatever is left. Every proposal states its evidence, and nothing is applied until a human clicks.
Every join key is measured for duplicates before the model can be used. Unverified is treated as unsafe.
Keep reports alive
A renamed column raises an error, not a zero.
We fingerprint the columns and types you approved and compare them on every load. When a used column is renamed, retyped or dropped, the report says so and names it — instead of rendering the silent zero every other tool renders.
Renamed join keys count as breaking, because a broken key unmatches every row while the totals stay put.
Watch the numbers
Silence, until something breaks character.
A digest arrives every Monday whether or not anything happened, which teaches people to skim it. A watch checks on a cadence and stays quiet — then speaks with the driver named, worked out arithmetically rather than narrated.
Biased toward silence on purpose: a watch that cries wolf is a watch nobody reads.
Answer the questions
An analyst that cannot make numbers up — and now it talks.
Ask in plain language, by keyboard or out loud — say “Hey Jimmy” and the answer comes back spoken, like Alexa for your numbers. Every reply is grounded in your certified definitions, highlights the chart that backs it, and carries its receipts: which certified metrics were read, over how many rows, computed where, data how fresh.
It generates no SQL at question time, has no write path, and reads only the pre-computed, access-scoped aggregates you can see on screen. That is the whole attack surface.
Prove the numbers
Provenance travels with the figure. And the platform grades itself.
Certified or ad-hoc. Computed in your warehouse or in the app. Data this fresh. Row-scoped, or complete. Every KPI — and every AI answer — carries all of it. Then, nightly, the platform recomputes every metric two independent ways and asks its own analyst exam questions, marking the answers against the engine.
The accuracy page lists WHICH check disagreed and by how much. A page that only shows a percentage is a marketing page.
The join guard, in motion
One duplicated key, 14× the revenue.
Scroll. Six order rows join six customer rows. One customer key appears twice — so every order matching it counts twice, and the revenue total quietly runs away.
No other BI tool tells you. Nexourz measures every join key for duplicates on real data before the model can be used, and refuses the join when the count says it is unsafe. Unverified is treated as unsafe.
Measured, not hypothetical: this exact guard refused a 14× overstatement on our 2,000,000-row test.
- 1 · Six orders match six customers, one to one. Revenue reads true.
- 2 · Customer key 4 turns out to exist twice. The join fans out.
- 3 · Revenue inflates to 14× and nothing in the dashboard objects.
- 4 · The fan-out guard counts the key, fails the model, and names the duplicate.
Prove the numbers
Every number comes with its receipts.
Two dashboards disagree and nobody can say which “margin” is the real one. That is not a reporting problem, it is a provenance problem — and it is the one we set out to solve.
Every KPI and chart in Nexourz carries a trust ledger: whether the definition is certified or ad-hoc, where the number was computed, how fresh the data underneath it is, whether it was cross-checked against a second engine, and whether the view is row-scoped.
sum(profit) ÷ sum(sales) · certified by finance · two engines agreed to 18.635367%
Ad-hoc definition — nobody has signed off on what counts as an order.
Row-scoped to your region — a colleague’s total will differ, and both are correct.
Keep reports alive
A renamed column should be an error, not a zero.
Watch what happens to the same dashboard when someone upstream renames revenue_eur.
Every other BI tool
No error. No warning. Nobody is told.
The chart renders. The axes render. The number is zero and looks deliberate.
Nexourz BI
Fails loud, and names the column.
This report’s data source has changed
orders.revenue_eur no longer exists. It was renamed or dropped, and this report reads it — every value it feeds is now empty.
Detected by comparing the live schema against the one an admin approved. Join keys count as breaking too: a renamed key unmatches every row while the totals stay put, which is the most confusing failure of all.
One engine, two products
Author on Desktop. Publish to Server.
Nexourz BI Server
Hosts, governs and delivers what you publish
- Published reports, organised in folders your team controls
- 20+ connectors: Snowflake, Postgres, MySQL, SQL Server, Redshift, Excel, S3, Drive, Dropbox, Stripe, HubSpot, GA4, Shopify, Jira and more — or several tables modelled as one source
- An AI analyst on every report you can type to or talk to — with receipts on every answer
- Scheduled email reports with the PDF attached, Slack digests, silent watches, shareable data pulls
- AI-authored SQL transformations that run only behind a human signature, on a schedule you set
- Row-level security enforced in one place, inherited by exports, email, Slack, embeds and the AI
Nexourz BI Desktop
Where dashboards get built — or just asked for
- State a business problem in words; get the KPI plan, the dashboards, and the honest list of what your data cannot answer
- Drag-and-drop shelves with certified metrics as first-class pills
- An AI assistant that drafts a dashboard from one sentence — typed or dictated
- Insight Studio — an agent that explores your data and proposes what matters
- Calculated fields, LOD expressions, table calculations, drill-down hierarchies, what-if parameters
- One-click publish to Server, into the folder you choose
Also included
The table stakes, without the fanfare
- The Monday-morning report
- Subscribe any filtered view to email on a schedule — PDF, Excel or CSV attached, sent at YOUR 8am.
- Data pulls
- “Can you pull me Q3 West sales?” becomes a stable link that recomputes under each viewer's own access.
- Messy Excel, handled
- A 12-sheet client workbook is detected, staged into SQL tables, modelled and dashboarded — footnote definitions included.
- Certified metrics
- Define margin once, certify it, and the definition travels with every number.
- The Attention page
- Every open gate in one place: unsigned drafts, failed sends, stalled syncs — and “nothing needs you” when true.
- Voice everywhere
- Dictate a question, a dashboard, or a transformation. Answers can be read back aloud.
- Tableau migration
- Every worksheet converts and is graded exact, approximated or unsupported.
- Six export formats
- CSV, Excel, Word, PDF, Markdown, JSON — the file your stakeholder asked for.
- Embedded analytics
- Any report in an iframe inside your product, scoped per customer by signed token.
- Warehouse push-down
- Aggregation runs in your database, so dashboards stay fast as tables grow.
- Folders and roles
- Projects-as-folders, role-based access, per-report visibility in the catalog.
- Audit log
- Who viewed, exported, published, approved or changed anything — with timestamps.
How we compare
Honest, including about us.
Where we haven’t shipped something yet it says WIP — not a checkmark. We would rather you find the gaps here than in your third week.
| Capability | Nexourz BI | Tableau | Power BI | Looker |
|---|---|---|---|---|
| AI analyst grounded in your own definitionsReads pre-computed metrics only — it cannot hallucinate SQL | ✓available | ◐partial | ◐partial | ◐partial |
| State a business problem, get the plan for signatureKPI tree with reasons, dashboard designs, and what your data cannot answer — built on one admin click | ✓available | —not offered | —not offered | —not offered |
| Talk to it: wake word, dictation, spoken answers“Hey Jimmy” hands-free ask-and-answer; the mic is on every prompt box | ✓available | —not offered | —not offered | —not offered |
| Nightly accuracy self-audit, published in-productMetrics recomputed two ways and AI answers marked against the engine — disagreements listed by name | ✓available | —not offered | —not offered | —not offered |
| AI-authored SQL transformations behind a human signatureSELECT-only by construction; the signature pins the exact text, any edit revokes it; runs on your schedule | ✓available | —not offered | ◐partial | ◐partial |
| Build a dashboard from a sentence | ✓available | —not offered | ◐partial | —not offered |
| An agent that explores your data and proposes the insights | ✓available | —not offered | —not offered | —not offered |
| Provenance on every KPI, not just dataset lineageCertified · computed where · freshness · row-scoped — on the number, with no add-on SKU | ✓available | ◐partial | ◐partial | ◐partial |
| Cross-checked against a second engineThe same figure recomputed by an independent engine, and claimed only where they agreed | ✓available | —not offered | —not offered | —not offered |
| Certified metric definitions (semantic layer) | ✓available | ◐partial | ◐partial | ✓available |
| Join keys measured for duplicates on real dataWe count them and refuse an unsafe model — Tableau and Power BI instead aggregate before joining, which avoids the problem differently | ✓available | ◐partial | ◐partial | ◐partial |
| Schema-drift detection that fails loudNo BI tool we found does this natively; data-observability tools do, if you already run one | ✓available | —not offered | —not offered | ◐partial |
| Anomaly watches biased toward silence | ✓available | ◐partial | ◐partial | —not offered |
| Slack delivery carrying certified provenance | ✓available | ◐partial | ◐partial | ◐partial |
| Row-level security inherited by exports, AI and Slack | ✓available | ◐partial | ◐partial | ◐partial |
| Embedded analytics with per-viewer row scoping | ✓available | ✓available | ✓available | ✓available |
| Migrate Tableau workbooks, graded honestly | ✓available | —not offered | ◐partial | —not offered |
| Warehouse-native push-downPostgres proven live; the Snowflake dialect is written but unverified | ◐partial | ✓available | ✓available | ✓available |
| Geographic mapsUS state choropleths, and it names any region it could not place. World countries and point maps not yet | ◐partial | ✓available | ✓available | ✓available |
| Native mobile apps | WIPwork in progress | ✓available | ✓available | ✓available |
| BigQuery connectorBuilt; ships once verified against a live warehouse | WIPwork in progress | ✓available | ✓available | ✓available |
| Joins across two different connectionsA model's tables must live on one connection today | WIPwork in progress | ✓available | ✓available | ◐partial |
| Power BI / Looker migrationThe framework exists; the extractors do not yet | WIPwork in progress | —not offered | —not offered | —not offered |
✓ available · ◐ partial or a paid add-on · — not offered · WIP we are building it and will say so until it ships. Competitor capabilities summarised from public documentation, reviewed 3 August 2026. We corrected two rows in this table after research showed our claims about Tableau and Power BI were too generous to us — tell us if we got another one wrong and we will fix it too.
Security & deployment
Built like infrastructure, not like a demo.
Your AWS account on Enterprise
The entire stack is Terraform and deploys into your VPC. Your data never reaches our infrastructure. We run the same stack ourselves.
One choke point for row-level security
Dashboards, exports, AI answers, Slack deliveries and embeds all inherit the same scope. A policy that cannot be resolved yields no rows, never all rows.
The AI is read-only where it answers, gated where it writes
At question time: no SQL, no write path, no raw-row access — it reads the same pre-computed aggregates you see. On the transformations surface it may DRAFT SQL, which is SELECT-only by construction, runs on read-only time-boxed sessions, and executes only the exact text a human signed — any edit revokes the signature. Your data trains no model.
Signed, expiring embed tokens
HMAC-SHA256, with the report and the row scope inside the signature. Editing the URL changes nothing.
Secrets never touch the repo
Source credentials live in AWS Systems Manager as encrypted parameters and are injected at run time — never in git, the image, or our storage.
An audit log that answers questions
Who viewed, exported, published, or changed a join, with timestamps — exportable for your own compliance needs.
Pricing
Per seat. AI included. No viewer tax.
No metered AI add-ons, no hidden viewer fees. Annual billing; monthly on request.
Starter
For teams who read reports
- View and explore published reports
- Every export format, plus scheduled email reports
- The AI analyst — typed or spoken — included
- Slack digests, alerts and watches
- Shareable data pulls
Pro
For the people who build them
- Everything in Starter
- Problem-statement intake: brief it, sign it, built
- Desktop Studio and Insight Studio
- Multi-table data models and Excel staging
- AI-authored SQL transformations, human-signed
- Certify metric definitions
- Tableau migration
- Embed reports in your product
Enterprise
For organisations with requirements
- Everything in Pro
- SSO and role-based access control
- Row-level security policies
- Deploy in your own AWS account
- Migration services and priority support
AI usage on Starter and Pro is governed by fair-use budgets sized so a whole team asking questions all day stays inside them; Enterprise sets its own. Nothing is charged per query, per row, or per viewer.
The ask
One data source, two weeks.
In the first onboarding session we model your schema live — relationships proposed, join keys measured, the first dashboards drafted by the AI. We onboard every workspace personally, and you will be talking to the people who built it.
If the numbers aren't right and provable, walk away.
We use what you send us only to reply about Nexourz BI. See our privacy policy.