ERP AI

Open the dashboard for your role rather than a generic one: CFO, CEO, head of e-commerce, supply chain, marketing, merchandising, customer experience, fulfilment, procurement, warehouse manager or data admin.

Read the headline row first — the four to six numbers that decide whether anything else on the page needs looking at, each with its trend and its comparison window.

Get 'needs attention' as a list of breached thresholds in plain language, with the value and the band it crossed, and click straight through to the tile that explains it.

Change the window and the location scope from the top bar, on the pages that actually honour them — a picker that would change nothing is not rendered.

Read four totals before anything else: money at risk if nothing changes, money already lost in the window, capital tied up doing nothing, and how many findings are rated critical.

Work a ranked list rather than a dashboard — each finding named, priced, and carrying the horizon over which the impact lands.

Expand 'show workings' on any of them: the inputs, the counterfactual the impact is measured against, and the formula. An impact whose derivation is hidden gets believed when it should not be.

See the detectors that ran and the ones that did not, with the reason — a missing input is reported, not silently treated as a clean bill of health.

Get the same briefing as a morning email, composed from the findings rather than generated over them.

Ask in your own words and get a number, a chart and a sentence — with the certified metric definition, the window it ran over, the comparison period, the data-as-of date and a confidence score printed beside it.

Open 'resolved query' to see exactly what was run: the metrics, the dimensions, the filters. There is no hidden SQL to take on trust, because SQL was never written.

Start from the questions your role usually asks, or ask for a whole report — 'a dashboard for cash and receivables' — and get a set of widgets you can save.

Save any answer as a view, or open it in Explore to keep pulling on it.

Run it with no model API key at all: the assistant falls back to deterministic matching over the same catalog, which understands fewer phrasings and returns identical numbers.

Search 229 governed metrics by name, synonym or domain, each carrying its definition, its owner and whether it is certified.

Group by up to four of 101 dimensions, filter on declared columns, set the bucket and the comparison, and switch between chart and table.

Read the certification mark before you use a number: a certified metric has an owner and a settled definition; an uncertified one says so rather than looking identical.

Copy a link to the exact query, or save it as a view for the rest of the workspace.

Stay inside the governance boundary the whole time — this is a picker over the catalog, not a SQL console, and it is gated to analyst and above for that reason.

See revenue by category by day, week or month, with the mix stacked rather than merged into one line.

Read the top products by revenue with units alongside, so a high-revenue, low-volume line is distinguishable from the opposite.

Watch the most recent orders arrive with their status and their value.

Use it as the shared screen: it reads the same aggregates a viewer can already see, so it needs no extra permission.

Read four numbers that together say whether the position is healthy: value on hand, stockout rate, dead-stock share and the in/out ratio.

See where the value sits by warehouse and by age, and which lines have not moved inside the dead-stock window.

Work the stock list by product and warehouse — units on hand, units available, inventory value, sold per day, days of cover — as a governed report a viewer may open without ad-hoc query rights.

Get reorder points sized for a 95% service level, with the safety stock and the lead time that produced them.

Read dimensional stock on a Dynamics deployment: site, warehouse, location and inventory dimensions, rather than one flat quantity per SKU.

Read OTIF at the top and its two components under it: a supplier can be on time and short, or complete and late, and the fix is different.

See where the spend goes and how reliable each supplier is over the window, with lines, on-time and in-full rates side by side.

Track lead-time variance rather than average lead time — the average is what you plan to, the variance is what breaks the plan.

Watch supplier concentration, so the risk of a single failure is a number rather than an assumption.

Work over the right window: purchasing calendars are lumpy, so this page opens on all time rather than sampling one cluster of order dates and reporting an OTIF of zero.

Arrive at a vendor from Procurement or a customer from Stores, and have a permanent page to return to rather than a filter you have to rebuild.

Read one supplier's spend, lines, OTIF, on-order and direct position together, then open their purchase orders.

Read one customer's lifetime spend, order count, average basket and days since last order — a lifetime question, so the directory opens on all time.

Segment either side and compare, without leaving the governed report path.

Write a rule against any catalog metric: an absolute threshold, a percentage change, or a statistical anomaly detected without a model.

Choose the cadence and the delivery — email or Slack — and see the last check and its result on every rule.

Read the event history: what fired, when, against which value, and what the rule was watching at the time.

Trust that an alert and a dashboard cannot disagree, because both resolve the same metric definition through the same compiler.

Save a query from Explore or from an answer, name it, and share it with the workspace or keep it to yourself.

Annotate a view so the next reader gets the context that made the number mean something.

Schedule an export — CSV, XLSX or PDF — to arrive without anyone opening the product.

Open a governed operational report as a paginated, sortable, filterable table, where the metrics and the fact table are fixed server-side and only page, sort and declared filters come from the browser.

Connect Dynamics 365 Finance & Operations, Business Central, NetSuite, Acumatica, Shopify, Amazon SP-API, or a CSV — and add the one the ERP does not hold without a code change.

Watch sync health per connection: last run, records pulled, records written, and a degraded flag after repeated failures rather than a silent stall.

Resume rather than restart — cursors are persisted after every page, and unchanged records are skipped by content hash instead of being rewritten.

Inspect and replay the dead-letter queue: an unmappable record is quarantined and surfaced, and the sync carries on.

See freshness on the tiles that depend on it, so a stale number is labelled rather than looking current.

See every persona with the metrics it is graded on and the default window it opens over, and assign one to a member.

Preview a persona's dashboard as that role would see it before you hand it over.

Build on top of one: add a tile, write a formula over existing metrics from a template, and keep it inside the catalog rather than beside it.

Read what each dashboard is actually for — the question the role arrives with, not a list of charts.

Set roles — owner, admin, analyst, viewer — and row-level data scope on top, so a member sees only their channels or their region.

Rely on the scope being unwidenable: it is appended by the compiler after the caller's own filters, not offered as a default they can change.

Read the audit log of every mutation and every denied access, on a table the application role cannot rewrite.

Manage currency and FX, data-quality checks, people and invitations, and the workspace's plan and usage.

Rotate credentials without leaking them — the route records field names only, and redaction runs over everything on its way to a log.

Work four steps: choose a persona, bring in data, run the first sync, and see the dashboard.

Come back to it when a second source is added — the checklist is derived from workspace state, so it is right again rather than gone.

Get the answer to 'why is my dashboard empty' in the one place a viewer will look for it.

Explore the whole product with no ERP and no API key: every connector runs in mock mode, and a seeded workspace ships 24 months of synthetic data.