The command center for decisions and automation
This space gathers what the other eight produce: every posted document, every open item, every rule that fired. The Executive Cockpit does not open with a wall of charts — it opens with a change, breaks it down into causes, states the cost of doing nothing and offers three decisions, each with a what-if projection and its evidence. The Automation Center shows what NG One handled on its own and how much time that returned to people. The AI copilot answers with the figure, the reasoning and the source documents behind it — not with a sentence you are asked to trust. Anything AI proposes is a draft a person confirms.
- Executive Cockpit
- Automation Center
- Copilot with evidence
- OLAP and spreadsheet
- Forecasts and anomalies
- Agents within limits
You have the data. You do not have the answers.
What separates a report from a decision is an explanation — and explanations do not export to Excel.
A company ten years into operation already holds every figure it needs and almost none of the answers. Reports get exported, merged by hand, and go stale before they are assembled. BI arrives as a separate product fed by a nightly transfer, so the number on the dashboard differs from the number in the general ledger by just enough that nobody trusts either. When someone asks why margin dropped in March, the answer takes three days, arrives as a hypothesis, and is never verified.
NG One has no separate analytical product because analysis reads the same rows the work is done on. Dimensions sit on every line from the first one, and each transaction keeps a reference to the version of the rules it was posted under — which is why “explain this posting” is a function rather than a reconstruction. That is what made it possible to organise the Executive Cockpit around a decision instead of a chart: a change, a breakdown of its causes, the cost of inaction, and three decisions with what-if projections and evidence. Every figure on that screen has a path back to the document it came from.
The structural advantage is that AI here is not a layer above the system but a user inside it. It passes through the same permissions, the same data scope, the same approval limits, the same segregation of duties and the same immutable audit trail as a person — so autonomy is switched on gradually, per process and per amount, instead of being accepted or rejected wholesale. That is also why the split is strict: Cmd+K stays deterministic and separate, the panel stays contextual and always carries evidence, and the Automation Center measures how much work the system actually took over. What is not measured is not sold here as automation.
How an insight becomes an executed action
Six steps that close the loop — from the posting to the measurement of what the system did by itself.
- Step 1
Data becomes fact
An insight does not start in a report; it starts in a posting. Every line carries its dimensions and a reference to the rule version it was posted under, so later it is read rather than reconstructed.
- Dimensions on every line from the first one
- Rule version referenced on the transaction itself
- One partner record and one item record for the whole system
- No nightly transfer into a separate analytical product
- Step 2
See what happened
The screen does not open with generic charts. It opens with a change: what deviated from expectation, by how much, and since when.
- Executive Cockpit with the change as the entry point
- Dashboards per role, not one view behind different filters
- Trial balance, ledger cards, stock, journal and ageing with drill-down
- OLAP and pivot across posting dimensions
- Step 3
Understand why
The change is decomposed into causes — a waterfall shows what contributed how much. Every contribution has a path to the source document, and the AI panel explains the posting instead of paraphrasing the number.
- Cause waterfall under every change
- “Explain this posting” against the rules in force at the time
- Source documents attached to every AI answer
- Drill-down and quick preview without leaving the screen
- Step 4
Decide what happens next
The cockpit offers three decisions, not twenty options. Each carries a what-if projection and the cost of doing nothing.
- Three proposed decisions with what-if simulation
- Cost of inaction stated as a figure, not an adjective
- A draft reminder, purchase order or task next to the decision
- The decision enters the workflow with limits and an audit trail
- Step 5
Execute: a rule instead of a repetition
A decision that repeats becomes a rule. The rule works inside the limits a person set and leaves the same trail as a human action.
- Event → condition → action on the workflow kernel
- Approval limits, SoD and maker–checker apply to automation too
- Autonomy level per process and per amount
- Immutable execution journal, with no exemption for the system
- Step 6
Measure whether it was worth it
Automation that is not measured is a promise. The Automation Center shows what the system handled alone, where a rule failed, and how much time went back to people.
- Share of documents processed automatically, per process
- Execution history with outcome and reason
- Time saved as a KPI, not as an estimate in a proposal
- AI evals in CI: golden sets, tenant isolation, prompt injection
What this space covers
Thirty capabilities across five groups — from the Executive Cockpit to the MCP server. All of them read the same rows the postings live on, and pass through the same permissions a person does.
Decisions: Executive Cockpit and dashboards
The screen opens with a change, not a chart, and every number has a path to its source document.
6 capabilities
Executive Cockpit
A hero screen organised around decisions: what changed → waterfall of causes → consequence of no response → three decisions, each with a what-if projection and its evidence.
What-if beside the decision
Before confirming, you see the projection: what happens to cash flow, margin or lead time if the decision is executed — and what happens if it is not.
Role-aware dashboards
A director, an accountant, a sales rep and a planner receive a different hierarchy of information — not the same dashboard behind a different filter.
Operational and financial views
Trial balance, ledger cards, stock, journal and ageing — each with drill-down to the line and export to Excel.
Personal and team KPIs
Role-based KPI cards on My Work and in the space workspace; users personalise them strictly within their own permissions.
Atlas with live state
Value chains showing active document counts, delays, exceptions and the share processed automatically — the intelligence layer runs through all nine spaces.
Reporting, OLAP and spreadsheet
Analysis over the same rows the work is done on. No nightly transfer, no second version of the number.
6 capabilities
Table standard on every list
Column toggling and pinning, advanced filters with operators, saved views per user, footer totals, bulk actions, Excel export, full-screen mode.
Reporting center
A library of ready reports by module in one place, with access rights and a role-based layout — instead of a shared folder full of spreadsheets.
Report and print designer
Users and consultants build reports and document layouts without a developer; templates are versioned configuration, so a core upgrade does not overwrite the changes.
OLAP analysis across dimensions
Drill-down and drill-through across posting dimensions, plus pivot mode on any list — analysis starts from the list you already work in.
Spreadsheet over live data
Excel-like work with pivots over current state, without exports and manual merging of files that are stale the same day.
Power BI semantic model
For companies already invested in Power BI: a semantic model instead of raw tables, so an external report does not invent its own definition of margin.
Automation Center: what the ERP did by itself
Automation is visible, measured and reversible. A rule nobody can see is a rule nobody trusts.
6 capabilities
Automation Center
A hero screen: active rules, execution history with outcomes, time saved, and the list of what the system completed without a person.
Event → condition → action rules
Automation rules on the workflow kernel: expected receipts from purchase orders, replenishment proposals, dunning on due date, recurring invoices, statement matching.
Automation KPI: “what the ERP did on its own”
Share of documents processed automatically per process and the count of actions with no human touch — a measure that either rises or does not, but is never assumed.
SLAs, escalations and delegation
A deadline on the task, escalation when the deadline breaks, a stand-in when the owner is away — without a reminder living in somebody's inbox.
Time saved and correction-free auto-posting
Time saved per process and the share of documents posted automatically with no correction at all — tracked month over month at the same client.
Durable orchestration of multi-step processes
Retries and compensations when a step fails; the process does not stop halfway and leave a partial state somebody discovers a month later.
AI with evidence
Every answer carries the figure, the reasoning and the source document. Every suggestion is a draft.
6 capabilities
AI Daily Brief
On My Work, at sign-in: what changed, who is late, what is running out — with proposed actions. Every conclusion carries an explanation and drill-down to the data.
NG One AI — contextual panel
A right-hand panel on every screen, working on the current object: explain this posting, check the VAT treatment, find the related goods receipt, check for a duplicate, draft a message to the supplier.
OCR and posting proposal
PDF or scan → extraction → supplier recognised → duplicate detection → posting proposal → proposed link to the goods receipt; uncertain fields are flagged, not guessed.
Action suggestions as drafts
“Prepare reminders for customers more than 30 days late” produces drafts for review, not sent reminders. The draft enters the same workflow as a human decision.
AI evaluation framework in CI
Golden sets of queries and documents, regression tests, tenant isolation tests and prompt-injection tests; OCR is measured per field type and by the share of fields the system refuses to guess.
Model gateway and AI audit trail
Swappable provider, budget and quota per tenant, visible cost, personal-data redaction, and an append-only trail: input, model and prompt version, output, and who confirmed it.
Forecasting, agents and AI foundations
Autonomy is switched on per process and per limit — it never ships enabled.
6 capabilities
Forecasts: cash flow, demand, collection risk
Collection forecasting from each customer's payment history, demand forecasting for replenishment and MRP, and a collection-risk warning before a document is approved — not after it falls due.
Anomaly detection
Deviations in postings, duplicate and suspicious invoices, unusual margins and discounts — flagged as they occur, with the reason the record was singled out.
AI agents with autonomy levels
An AP agent, a sales-order agent and a collections agent operate in three modes: propose → act on confirmation → act autonomously within limits. The limits are the same approval limits a person has.
Local models and RAG
Hosted or local models for on-premise privacy; RAG over documents, master data and manuals, indexed per tenant under the same RLS isolation as the data itself.
MCP server
NG One resources and actions exposed as MCP tools: an external agent reads through permissions and data scope, and changes state only through state machines and workflow.
AI permissions, credits and consumption
The entitlement framework: AI credits, quotas and usage metering per tenant, with an entitlement check at the API layer alongside the permission check. AI cost is visible, not estimated at month end.
AI in this space
This is the command center for AI, not its only home — the panel works in the context of every screen in the system.
A copilot that answers with evidence
Not “this customer owes 2.4M”, but “this customer owes RSD 2.4M across seven open invoices, three of them more than 30 days late” — with those invoices listed and drill-down into each. An answer that cannot be backed by a source is not shown.
A daily brief that starts at the decision
AI Daily Brief on My Work: revenue above last month, three of your largest customers late, an item running out in six days — with two purchase orders proposed. Every conclusion carries an explanation and a path to the data it was derived from.
OCR that is allowed to say “not sure”
Extraction is measured per field type — supplier, invoice number, total, VAT and line items separately — and a field the system is unsure about is flagged rather than guessed. A confidently wrong field costs more than an empty one.
A suggestion is a draft, not an action
AI prepares reminders, purchase orders, tasks and replies as drafts that enter the same workflow as a human decision — with the same limits, approvals and audit trail. Autonomy is enabled per process and per amount, and can always be stepped back.
Anomalies and forecasts with a stated reason
NG One forecasts cash flow and demand, scores collection risk before a document is approved, and detects deviations in postings, duplicate invoices and unusual discounts. Every flag carries the reason the record was singled out — otherwise it is just an alert people mute within a week.
Why this differs from what you run today
Measured against what companies actually have: Pantheon, Business Central, Odoo — or their own reports in Excel.
AI as evidence, draft and context — not sparkle and a chat box
The usual “AI in the ERP” is a window in the corner that paraphrases what you can already see. Here AI works on the current screen and object, offers concrete actions instead of an empty prompt, and attaches source documents to every answer. What it cannot substantiate, it does not assert.
Automation that is measured
Pantheon and Business Central both have automation rules; neither shows how much those rules actually did. The Automation Center keeps active rules, execution history with outcomes, and the “what the ERP did on its own” KPI — a percentage per process, not an impression after a year of use.
Search and AI are deliberately separate
Cmd+K is deterministic: it finds a screen, a document by number, a partner or an item, and triggers a known action — the same query always returns the same result, optimised for the keyboard. AI is a distinct layer in the right-hand panel. Blend the fast and predictable with the generative and unpredictable, and users stop trusting both.
Analytics over dimensions from the first line
Dimensional posting is not something you bolt on later, so in NG One it sits in the foundation: every line carries its dimensions from the first one. Systems that retrofit dimensions can only analyse the period after the retrofit, while historical data has no breakdown at all. OLAP and pivot here read the same rows the work is done on — no nightly transfer, no separate BI product, and no second version of the same number.
The flows this space runs through
A business space is not an island. These processes touch it end to end, and where a flow leaves this space the record stays the same — the next step receives it structured rather than retyped.
Finance
Record-to-Report
The path from document to report. Postings come from the business event rather than a second round of data entry, and carry their dimensions from the first line, so period close does not begin by hunting for what is missing. POPDV, PP PDV and the APR statements come out of that same journal, with nothing reassembled afterwards.
- Document
- Posting
- Controls
- Close
- Reporting
Revenue
Lead-to-Cash
The path from first opportunity to money in the account. Each step hands the next a structured record, so a quote is never retyped into an order, nor a delivery note into an invoice. The invoice leaves for SEF from the same step that raises it.
- Opportunity
- Quote
- Order
- Delivery
- Invoice
- Collection
Procurement
Procure-to-Pay
The path from a need to a supplier payment. The invoice arrives over SEF, and purchase order, receipt and invoice reconcile themselves — a person decides only where the three documents disagree.
- Requisition
- Approval
- Purchase order
- Goods receipt
- Invoice
- Payment
Questions about this space
Scope, boundaries, and the rules this space posts by.
Do we still need a separate BI tool alongside NG One?
Not for the analysis this space covers. The reporting center, OLAP across dimensions, pivot mode on every list and the spreadsheet over live data are part of the system, and they read the same rows the postings live on — no nightly transfer, no second version of the same number. Operational reports — trial balance, ledger cards, stock, journal and ageing — come with drill-down to the line and export to Excel. For companies already invested in Power BI, NG One exposes a semantic model instead of raw tables, so external reports do not invent their own definition of margin.
How do we know the AI did not make the answer up?
Because an answer without a source is not an answer here. Every conclusion in the panel and in the daily brief carries the figure, a short explanation, the source documents and drill-down to the line. OCR is not measured as one percentage but per field type, and fields the model is unsure about are flagged rather than guessed. An evaluation framework runs in CI: golden sets of queries and documents, regression tests, tenant isolation tests and prompt-injection tests from attached documents. If that check fails, the change does not merge.
Can AI post an invoice or send a reminder on its own?
Not by default, and that is deliberate. Everything AI proposes is a draft a person confirms, and that draft passes through the same workflow, the same approval limits and the same audit trail as a decision made by a person. Autonomy levels are set per tenant and per process: propose → act on confirmation → act autonomously within limits. Autonomy is enabled in the console, per process and per amount, and can always be stepped back. The default is human-in-the-loop, not the reverse.
Can AI see data a user is not allowed to see?
No. AI passes through the same permissions, the same data scope and the same tenant isolation as a person — there is no privileged access “because it is the system”. The RAG index is built per tenant, under the same RLS isolation as the data itself. Every suggestion and every action stays in an append-only trail: input, model and prompt version, output, and who confirmed it. Isolation tests are part of the evaluation framework in CI, which makes them continuous rather than a one-off review.
What if our data must not leave the building?
The model gateway abstracts the provider, so the choice of model is not wired into the code. Local models (vLLM or Ollama) run on-premise — the same flows, the model inside your network. In hosted mode, personal data is redacted before the call, and budget and quota are set per tenant with cost visible per call. If AI has to be off for part of the process, it is off for that process — the rest of the system loses no function.
What is the automation KPI and why do you insist on it?
It answers the question “what did the ERP do on its own”: the share of documents processed automatically per process, the number of actions with no human touch, and the time saved. We insist on it because automation is the easiest promise to make in an ERP sale and the hardest to verify after the contract is signed. If the number does not rise, the rule is not working — and that shows on screen the same month, not in an annual review. The Automation Center also shows the executions that failed and why they failed.
How does this space fit into the rest of the system?
It is the ninth of nine business spaces and the only one that holds no domain of its own — it reads the other eight: some sixty technical domains across eight end-to-end processes. The Executive Cockpit decomposes a change that originated in sales, procurement or production; the Automation Center measures the rules running inside those same processes; the panel explains the posting on whichever screen you are standing on, whichever space that belongs to. Which is why there is no integration here whose job is to assemble the data: the analysis reads the same rows the work is done on.
Neighbouring spaces
See a decision with its evidence, not a slide about AI
We run the walkthrough on a demo tenant with twelve months of data: the Executive Cockpit breaks a change down into causes, the panel answers with the source document, the Automation Center shows what the system handled on its own. Pricing follows a scoping analysis.