Your day starts with a decision, not a report
My work is the first screen after login and the only place where the entire system narrows down to one person and one day. Instead of a menu with dozens of entries and a dashboard that still has to be interpreted, the screen opens with what needs a decision: exceptions, approval requests, and tasks with a deadline. Below that sits a daily brief NG One has assembled, suggested actions prepared as drafts, and an exact record of what the system completed without you since your last login. Every conclusion leads back to the data it came from — no figure is a claim without backing. Charts come last, as personal KPIs, because measurement is not the start of a working day but its summary.
- My day
- Tasks and approvals
- Exceptions
- AI daily brief
- Executed automations
- Personal KPIs
Why the home screen decides whether an ERP gets used
A system is not judged by its module count but by how long it takes from login to the first useful action.
In most business systems companies run today, the day begins with a search. The user opens a menu, picks a module, builds a filter, and only then learns whether anything is waiting. Approval requests sit in one module, overdue receivables in another, a mismatch between a purchase order and a supplier invoice in a third, and an expiring deadline nowhere at all — until it has passed. The cost is not only time: information you have to hunt for is found selectively, so decisions get made on part of the picture. Whatever is not on the home screen effectively does not exist.
NG One reverses the order. The My day screen does not ask what you are looking for — it aggregates state across every space where you hold rights and presents it by priority: decisions and exceptions first, then tasks and approvals, the daily brief, suggestions as drafts, executed automations, and only at the end the KPIs. Tasks and approvals do not come from a workflow add-on; they come from the foundation. The workflow kernel — approval limits, segregation of duties, maker-checker, delegations, escalations — sits underneath every business module rather than beside them. That is why a purchase order, a journal entry, and a leave request all land in the same inbox, under the same approval rules and with the same decision trail.
The structural advantage is that this follows from the architecture rather than from an extra screen. The space is assembled from a navigation registry in which every entry carries its space, required roles, and license entitlement — the same registry that feeds the left rail, the Atlas, and the capability catalog. Layout therefore follows role, not preference: a director, an accountant, and a sales rep get a different hierarchy of information, not the same dashboard with a different filter. Personalization exists — favorites, saved views, a chosen start page — but it operates inside the user's rights and never around them. A screen that surfaced data outside a role's reach would not be personalized; it would be an access-control failure.
What a day in My work looks like
Six steps from login to a closed decision. The order is the same every morning; the content is not.
- Step 1
Login recognizes who you are and what you do
The system does not open the same page for everyone. Role determines which spaces are visible, in what order, which workspace opens first, and which KPI cards appear — and within their rights, users choose their own start page and working environment.
- Role-aware layout defined by an administrator or consultant
- Choice of start page, language, theme, and display density
- First-login workspace setup with a live preview
- Menu state, favorites, and recently used are remembered
- Step 2
What needs a decision comes first
The top of the screen holds exceptions rather than charts: documents that deviate, receivables running late, matches that failed, deadlines about to expire. Each exception is a record with an owner and a next step, not a line in a report someone has to notice.
- Exceptions from every flow the user has rights in, in one place
- Overdue and at-risk receivables linked to open items
- Three-way match deviations across order, receipt, and invoice
- Unmatched bank statement lines and suspected duplicate invoices
- Step 3
Tasks and approvals in a single inbox
Everything awaiting your signature arrives in one place, regardless of the module it came from. An approval carries its limit, its chain, and its segregation-of-duties constraint — if an amount exceeds your limit, the system says so before you confirm, not after.
- Task inbox with deadline, priority, and the source document
- Multi-step approvals with limits, branching, and maker-checker
- Bulk approval when there are many items and one decision
- Delegation and cover for absence, with a visible trail of who actually decided
- Step 4
The AI daily brief as a summary, not an opinion
NG One assembles a short brief in a few sentences: how many decisions are waiting, what is at risk, what does not reconcile, and what the system handled since your last login. Every statement in the brief is a link — a number opens a list, a list opens a document, a document opens its journal entry.
- A brief assembled from actual state, without generic phrasing
- An explanation of how the conclusion was derived and from which records
- Drill-down to source documents in a single click
- A contextual AI panel that works on the screen you are on
- Step 5
A suggestion arrives as a draft you confirm
When AI proposes a dunning letter, a purchase order, or a task, the result is a prepared document in draft status. Nothing is sent, nothing is posted, and nothing happens until a person confirms. A draft is edited, rejected, or accepted — and all three outcomes are recorded.
- A draft dunning letter for late customers, with the item list it came from
- A draft purchase order raised by a replenishment rule
- A draft task or partner message, ready to be edited
- A rejected suggestion stays in the trail — what was offered and what was decided
- Step 6
What the system did on its own — and only then KPIs
The day closes with two views backwards. First, the list of executed automations: which rules fired, on what, with what outcome, and how many human steps were avoided. Second, personal KPIs, deliberately last because they are a summary rather than a starting point.
- An automation strip with the execution history of every rule
- Automation KPI: the share of work the system handled without a person
- Role-bound personal KPIs with drill-down to source data
- A shortcut to the Automation Center for rules you want changed
What My work covers
Thirty capabilities across five groups — from the home screen and the task inbox, through six-layer authorization, to the AI daily brief and the executed automations strip. Each is named the way it behaves in the system: scope is the only proof worth offering here, so anything that is not on the screen is not on this list.
My day and the home screen
The first screen after login and everything that shapes it. My Day / Command Center carries the system's visual standard — display density, table behaviour, section order, and the rules for the light and dark themes — and that standard repeats across the other eight spaces.
6 capabilities
My day
A personalized home screen with at most five sections: Today, Needs attention, Core processes, Quick actions, Insights and automation. It answers six questions: what is happening, what awaits my decision, what is late, what can I start now, what did the system do on its own, and what is the next best action.
Role-aware starting layout
Role determines visible spaces, their order, shortcuts, the opening workspace, KPI cards, and the Atlas perspective. A director, an accountant, a sales rep, a planner, and a warehouse operator receive a different hierarchy of information — not the same dashboard with a changed filter. Layouts are defined by an administrator or consultant.
First-login workspace setup
Guided setup of language, theme, density, legal entity, and starting workspace, with a live preview assembled from the system's real components. Preferences are stored server-side and carry a schema version, so they survive a change of device and do not reappear at every login.
Favourites and recently used
Two lists that carry most of daily navigation: records the user marks and those they touched last. They sit on the home screen because in practice every user returns to a narrow set of partners, items, and documents, no matter how many modules the system has.
Saved views and personal shortcuts
Filters, columns, sorting, and widths in any list can be saved as a named view and launched from the home screen. The table standard applies everywhere: column toggling, reorder, pin, advanced filters with operators, footer totals, Excel export, and full screen.
Cmd+K — deterministic access to everything
Find a screen, find a document by number or partner, open an item, run a known action, create a new document. Deliberately separate from AI: the same query always returns the same result, with no guessing, optimized for keyboard and speed.
Tasks, approvals, and accountability
The workflow kernel is part of the foundation, not a module bought later. That is why any document in the system can enter approval under the same rules, and why a decision leaves the same trail everywhere.
7 capabilities
One task inbox for the whole system
Tasks and requests from every space arrive in one place, with deadline, priority, the source document, and the permitted actions. No separate inbox per module, and no tasks that are really just email.
Multi-step approvals with limits and branching
The approval chain is configured without developers: steps, branching conditions, amount limits by role and legal entity, parallel and serial approvals. Configuration is versioned and effective-dated, so every transaction records which version of the rule applied at the moment of the decision.
Approval limits and segregation of duties
Part of the six-layer authorization model: RBAC, data scope, approval limits, segregation of duties, maker-checker, and delegation. The system prevents the same person from preparing and approving the same document where that is prohibited — before confirmation, not in a later review.
Maker-checker on sensitive actions
High-risk actions require a second person: editing partner master data, a cash register entry, reopening a closed period. The rule is configuration rather than code, so the set of protected actions varies by tenant and legal entity.
Delegation and cover
Absence does not stall the flow. Authority transfers for a defined period and a defined scope, and the trail keeps both who decided and under whose authority. A delegation never grants more than the delegator holds.
Bulk approvals, SLA, and escalation
When the decision is the same across many items, they are approved at once while each item keeps its individual trail. Tasks carry a deadline, and a breach triggers escalation by a defined rule instead of sitting quietly in a list.
Immutable decision log
Every decision, configuration change, and rights assignment is written to a log that cannot be edited or deleted. This is a prerequisite for audit, but also for something more ordinary: when someone asks three months later why something was approved, the answer exists.
Exceptions that need attention
In NG One an exception is a first-class object with an owner and a next step, not a row in a report someone has to spot. Items originate in sales, procurement, inventory, and finance, and My day merges them into one list — the module an exception came from decides who closes it, not where it is seen.
6 capabilities
Exceptions from every flow, in one place
The Needs attention section aggregates deviations from sales, procurement, inventory, and finance — but only those the user has rights and responsibility for. Without it, an exception lives in a module that user may not open for weeks.
Overdue and at-risk receivables
Open items past their due date, breached credit limits, and customers whose lateness repeats. Leads straight to open items, the statement of account, and dunning preparation, with the 45 and 60 day payment terms the law prescribes.
Three-way match deviations
A difference between purchase order, goods receipt, and supplier invoice — in quantity, price, or terms — arrives as an exception showing all three documents and the exact gap. The decision is one step away: accept, return to the supplier, or raise a claim.
Duplicates and unmatched items
Automatic statement matching and duplicate invoice checks run in the background; only the remainder that needs a person reaches My day. Every suspicion carries its reason — same invoice number, same amount and partner, same reference in the statement.
Compliance deadlines in my day
Statutory deadlines relevant to your role — the PP PDV return, the EEO and EPP VAT records, POPDV, PPP-PD, the archive book — appear before they expire rather than after. The compliance calendar tracks each deadline per legal entity and tax period, and every one carries the live status of the channel it is filed through: SEF, e-fiscalization, CROSO.
Anomaly detection
A departure from the usual pattern — a sudden jump in purchase price, an unusual reversal, a payment outside the norm — is raised as an exception, together with the baseline the deviation was measured against.
AI daily brief and suggestions as drafts
AI in this space follows one rule: state what it found, show where it came from, and prepare a draft a person confirms. No action without confirmation, no conclusion without a source.
6 capabilities
AI daily brief
A few sentences that sum up the day: how many decisions are waiting, what is at risk, what does not reconcile, and what the system handled since the last login. The brief sits at the top of My day, directly below the exceptions, and every sentence in it is a link to the records it was assembled from — a brief with no path to the data is not shown.
Explanation and drill-down on every conclusion
No number in the brief is a closing statement. A click opens the list it came from, the list opens the document, the document shows its journal entry and the version of the rule it was posted under. If a path to the record cannot be built, the statement is not shown.
Suggested actions as drafts
A dunning letter, purchase order, task, or reply to a partner arrives as a prepared document in draft status, with the record set it was derived from. Nothing is sent or posted until a person confirms; acceptance, edit, and rejection are equally kept in the trail.
Contextual AI panel
A right-hand panel that opens on any screen and works on the object you are on, rather than an empty chat field. On an invoice it offers immediately: explain the posting, check the VAT treatment, find the linked goods receipt, check for a duplicate, draft a message to the supplier.
Copilot with evidence
An answer over your data always carries four parts: the figure, an explanation of how it was derived, the source documents, and drill-down. For OCR and field recognition the rule is that the system refuses to guess — an uncertain field is flagged, not filled.
AI agents with autonomy levels
Agents for accounts payable, sales orders, and collections, with an autonomy level set explicitly per tenant — from suggest-only to execute within defined bounds. The level is configuration with a full audit trail, never a default.
What NG One finished, and personal KPIs
The last two sections of the screen and the only place where figures appear. They sit at the bottom on purpose: the summary of a day is not its start.
5 capabilities
Executed automations strip
A list of what the system handled since the last login: documents posted, statement lines matched, recurring invoices issued, dunning letters raised. Every entry leads to the record it acted on and to the rule that triggered it.
Automatic rules (event → condition → action)
Rules are defined in the console, without developers: the event that fires them, the condition that must hold, and the action that executes. Part of the workflow kernel in the foundation, so they apply to every document type in the registry.
Automation KPI — what the system did alone
The share of work completed without a human step, broken down by document type and by rule. The measure exists to show the effect, but also to expose a rule that creates more exceptions than value.
Personal KPIs with drill-down
KPI cards bound to a role — collections, margin, ageing, overdue tasks — rather than one set for everyone. Each card opens to its source data; a KPI that cannot be decomposed down to documents does not belong on this screen.
Notifications on the channel you use
A task, escalation, or exception can also reach you outside the system — email, SMS, Viber, or push — under rules the user and administrator set together. The channel is configuration; the content and the accountability stay the same.
What AI does here specifically
In My work, AI has one job: to shorten the path from the state of the system to your decision — and to show, for every sentence it writes, where it came from.
A brief assembled from actual state
“Good morning. Today you have 6 decisions, 3 at-risk collections, and 2 documents that do not reconcile. Since your last login NG One has posted 47 documents automatically and matched 83% of the bank statement.” That is the brief on the demo tenant — the numbers come from it, and they are not a promise of performance. The value is not in the phrasing but in the fact that each of those figures is derived from records and leads back to them.
A conclusion without a source is not shown
Every AI conclusion carries an explanation of how it was derived and a path to the data: a number opens the list, the list opens the document, the document opens the journal entry and the rule version in force that day. This is a technical constraint, not only a principle — a claim for which the system cannot assemble a path to records never reaches the screen.
A suggestion is a draft, never an executed action
When AI concludes that nine customers are more than thirty days late, the result is a prepared dunning letter in draft status, with the item list it came from and the rule it was built under made visible. The draft is edited, rejected, or confirmed. A rejected suggestion stays in the trail — what the system offered and what the person decided are both known.
A panel that knows which screen you are on
AI opens as a right-hand panel over the current object, offering contextual actions instead of an empty question field. On a supplier invoice those are: explain the posting, check the VAT treatment, find the linked goods receipt, check for a duplicate, draft a message to the supplier. A free-form question is still possible, but the result is again a draft or a preview.
Automation that reports on itself
A system working in the background has to say what it did, otherwise it is either trusted blindly or shadowed by duplicate controls. The executed automations strip names every rule, the record it acted on, and the outcome, while the automation KPI measures the share of work done without a person — including rules that produce more exceptions than value.
Why this differs from the system you run today
Measured against what companies in Serbia actually have: Pantheon, Business Central, Odoo, or an in-house system grown over years of extensions.
The day starts with a decision, not a search
Pantheon opens a menu tree, Business Central a role center of tiles that still need interpreting, Odoo a grid of apps. In all three the user first chooses where to look. My day starts with what needs a decision — exceptions, approvals, deadlines — and puts KPIs at the bottom because they summarize the day rather than begin it. The order is an architectural rule, not a preference.
One inbox, because workflow is in the foundation
In most systems approval is an add-on each module solves its own way, so procurement has one chain, finance another, and HR uses email. In NG One the workflow kernel — limits, segregation of duties, maker-checker, delegation, escalation — sits in the foundation, underneath every business module rather than beside them. That is why a purchase order and a leave request land in the same inbox, under the same rules and with the same trail.
AI with evidence instead of a chat window
Adding a chat field to an ERP is the easiest way to claim the system has AI, and the fastest way to lose trust: an answer without a source cannot be used in finance. NG One binds every AI conclusion to the records it was derived from and delivers every recommendation as a draft. A system that refuses to guess is less impressive in a demo and more usable in March, when the VAT return is due.
Personalization that does not bypass authorization
Favourites, saved views, a chosen start page, and personal KPI cards all exist, but they always execute inside the user's rights. Six-layer authorization — roles, data scope, approval limits, segregation of duties, maker-checker, delegation — is not a filter applied to the presentation but a layer the data must pass through before it reaches the screen. A screen that shows data outside a role's reach is not personalized; it is defective.
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.
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
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
Questions about this space
Scope, boundaries, and the rules this space posts by.
Is My work just another dashboard?
No, and the difference is the order. A dashboard shows measurements and leaves the user to infer what to do. My day starts from what needs a decision — exceptions, approval requests, tasks with deadlines — and measurements sit at the bottom as a summary. The second difference is that every item has an owner and a next step: an exception is not a row in a table but a record closed by an action. The third is that the screen is not the same for everyone — layout follows role, not user preference.
Does every user see the same screen?
No. Role determines which spaces a user sees at all, in what order, which shortcuts they get, which workspace opens first, and which KPI cards appear. A sales rep sees five spaces rather than nine, and a director and an accountant get a different hierarchy of information, not the same dashboard with a different filter. Role-based layouts are defined by an administrator or consultant, and users personalize further within their rights. Personalization never widens access — what a role may not see cannot be added even as a favorite.
Does AI post or send anything without me?
Not in the part AI suggests. Recommendations arrive as drafts — a dunning letter, purchase order, task, message — and never leave draft status until a person confirms. Separately, there are deterministic automatic rules that you or your consultant define (event, condition, action) and that run without confirmation because you configured them that way: automatic statement matching, recurring invoices, notifications. Those rules are configuration, not AI, and everything they do is reported in the executed automations strip. Above that sit AI agents with an autonomy level configured explicitly per tenant — from suggest-only to execute within defined bounds. The level is the client's decision rather than a default, an agent never exceeds the rights of the user it acts for, and every change to that level enters the AI audit.
Where do the 47 documents and 83% figures come from?
From the demo tenant, and it says so next to them. The demo tenant carries twelve months of generated business so the screen can be seen at realistic data density rather than empty — a brief assembled over three documents proves nothing. A figure on this site therefore describes the scope of the system and the shape of the screen; it does not promise your result. What your own numbers turn out to be depends on master data quality, document discipline, and the rules configured during implementation. That is a question for a scoping analysis. A number without provenance is marketing, not data.
What if I do not trust an AI conclusion?
Then check it in two clicks — that is why drill-down is mandatory rather than an extra feature. Every number in the daily brief opens the list it was derived from, every list item opens the document, and the document shows the journal entry and the rule version it was posted under. If the system cannot assemble a path to the records, the statement is not displayed. This is a deliberate constraint: a system that sometimes guesses has to be checked always, and therefore never saves time.
Do we need all nine spaces before My day makes sense?
No. My day is a projection rather than a module: it has no database and no documents of its own — it aggregates what exists in the spaces you license and where the person holds rights. A company starting with finance and procurement gets a My day with three-way match deviations, overdue receivables, and purchase order approvals, and no empty sections for spaces it does not run. The navigation registry the screen is assembled from carries required roles and license entitlement on every entry, so a space switched on later appears in My day on its own, with no screen to rearrange and nothing to migrate. The task inbox, approval limits, and segregation of duties work from day one, because they sit in the foundation rather than in a module.
Does My work run on a phone and on a warehouse terminal?
It does, but not identically — and that is deliberate. The screen strategy distinguishes three modes: desktop-first for complex transactional work, a responsive view for phone and tablet, and purpose-built handheld screens for warehouse operations. On a phone, My day keeps decisions, approvals, and exceptions, because approvals happen on the road, and drops the dense tables that add nothing on a small screen. A warehouse operator does not get a shrunken desktop screen but a handheld flow written for gloves and a scanner, showing only the tasks that belong to them.
Neighbouring spaces
See your working day on your own data
The My work walkthrough runs on a demo tenant with twelve months of business, at realistic data density. To see how it looks with your roles, your documents, and your approval rules, arrange a scoping conversation.