Home · Solutions · Management & planning
Solution · Management & planningOne set of numbers for the board, refreshed overnight instead of merged over four days
The management pack in Power BI, not fourteen Excel files
Robots collect the figures from every source system on a schedule, owners write their commentary in Teams, and the board reads one governed model instead of a workbook.
Executive summary
The board pack should not depend on which analyst merged which spreadsheet on which day.
We build the layer that should have been under the pack.
The pack is ready the working day after the ledger closes rather than on the fifth, so decisions are taken a week earlier.
Power BI semantic model and workspace; the Power BI app with board, function and site audiences; SharePoint library with the extracts and the PDF
Business problem
Management reporting
The pack exists because no system in the group holds the whole picture. The ledger knows revenue and cost but not the pipeline. The CRM knows the pipeline but not what was delivered. The HR system knows headcount but not the cost centre it belongs to. Once a month someone puts the views side by side, and that someone is an analyst with a workbook.
The work also lands in the worst week. The pack is assembled while the close is still moving, so figures are pasted, re-pasted and reconciled against a version that has since changed. Commentary is requested by email from eight owners and half answer on time.
At scale the pack becomes a liability. Every new site adds files and another manual mapping. Definitions drift: an FTE in the HR extract is not an FTE in the cost report, and the difference hides in a column nobody opens. A board surprised twice commissions its own extracts, and the workload doubles.
How it works today
Before the pack is automated it is assembled file by file, and the route is much the same whatever the ERP underneath:
- PersonAn analyst downloads the trial balance and cost-centre reports into the month's folder
- WaitingTwo of the fourteen files arrive late because a site has not closed
- PersonFigures are pasted into the master workbook and last month's broken links repaired by hand
- SystemPipeline is exported from the CRM and compared with the ledger; the gap is noted, not explained
- PersonCommentary is chased by email from eight owners and edited in however it arrives
- Risk of errorHeadcount in the HR extract and in the cost report disagree, and the reconciliation sits in a hidden tab
- WaitingThe pack goes out as a PDF, and a corrected version follows within two days
- Risk of errorOnce the analyst who built a tab is on leave, nobody can reproduce the figure
Why the current process costs more than it appears
Nobody planned this work; it accumulated.
- Four days a month are spent by the people best able to interpret the numbers on moving them between files instead.
- Decisions wait for the pack. An intervention at a weak site that could start on the third working day starts on the tenth.
- Reconciliation arguments are not free. When the CRM and the ledger disagree, the meeting decides which figure is right instead of what to do.
- Every unexplained difference costs trust, and a board that stops believing the pack asks for parallel reports.
- Because definitions live in formulas rather than a model, one KPI is calculated two ways, and the gap surfaces in front of an auditor.
Cost of inaction
Standing still is comfortable, which is why the pack survives every review. The four days come round again, the corrected version follows the first, and the board settles for numbers it half believes. What grows is the gap between the speed of the business and the speed of its reporting: price, capacity and hiring decisions are taken on figures three weeks old.
The second risk is concentration. The knowledge behind the pack is undocumented: three analysts and one workbook, each a single point of failure. A resignation in the reporting team is a reporting outage, and rebuilding the logic from formulas costs more than modelling it once while its authors are still here.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A multi-site services group of roughly 1,400 people across eleven locations in four countries; SAP S/4HANA as the ledger, a CRM for pipeline, a separate HR system, an operational system for delivered hours, and Power BI Pro in finance.
Fourteen source files a month; three analysts for four working days each, twelve analyst-days in total; sixty pages, eight commentary owners, two reissues a quarter.
Extracts are downloaded by hand, pasted into a master workbook, reconciled where time allows, wrapped in commentary collected by email and published as a PDF on the sixth working day.
The assembly itself. Four days of manual consolidation sit between the ledger closing and the board seeing anything, and a late correction restarts part of the work.
Robots pull each source on a schedule, controls check completeness against the ledger, a governed Power BI semantic model holds the definitions, owners write commentary on a card in Microsoft Teams, and the pack goes out as a Power BI app with a PDF alongside.
The pack is ready the morning after the ledger closes, twelve analyst-days move from assembly to analysis, and breaches reach a named manager the day they appear. The figures are a model, not a measurement.
Proposed solution
We build the layer that should have been under the pack. Robots run on an Orchestrator schedule and take the extracts an analyst takes today: trial balance and cost centres from SAP, pipeline from the CRM, headcount from the HR system, billable hours from the operational system.
Each extract is dated, then loaded into a Power BI semantic model where the definitions live: one measure for revenue, one for FTE, one for utilisation, one period logic. Every report is a cut of that model. Control totals decide whether it publishes at all: if the ledger and an extract disagree, the run stops and a person is told what is missing.
Commentary is collected where the owners work. Power Automate sends each an Adaptive Card in Microsoft Teams with their variance and a deadline; the text is stored with a name and a timestamp. The pack is published as a Power BI app with audiences for the board, the function heads and the site managers, and as a PDF. A data alert posts threshold breaches to the management channel in Teams.
A note on platform. A scheduled import model in a Power BI Pro workspace is enough to start; Microsoft Fabric capacity matters for Direct Lake over OneLake, refreshes beyond what Pro allows, viewers without a per-user licence from F64 upwards, and Copilot in Power BI. The conversational layer on this site, the executive assistant and the morning brief, then answers from a governed model, not from fourteen spreadsheets.
UiPath Orchestrator time triggers, queues and audit; UiPath Integration Service connectors for SAP, Microsoft OneDrive & SharePoint and Microsoft Teams; Power BI semantic models, apps with audiences, subscriptions and data alerts; row-level security through Microsoft Entra ID groups; Power Automate Adaptive Cards in Teams
The extraction robots and their schedule, the controls that gate publication, the semantic model and its measures, the report pages and the paginated pack behind the PDF, the commentary workflow and the alert thresholds
SAP extraction through UiPath SAP activities (BAPI and OData, or the SAP GUI where no reporting API exists); the operational system through its API with UiPath Integration Service Connector Builder
How the automated process works
- AutomationOrchestrator starts the extraction run every night, and again early on the agreed close days, each source as its own queue item
- SystemRobots pull ledger, cost centres, pipeline, headcount and delivered hours from SAP, the CRM, the HR and operational systems into dated extracts
- AutomationControls run before publication: row counts, control totals against the ledger, missing cost centres, currency and period flags
- AutomationThe semantic model refreshes; measures, hierarchies and comparative periods are defined once there, not in a workbook
- PersonEach commentary owner gets a card in Teams with their variance, writes two or three sentences, and is reminded once if the deadline passes
- AutomationThe pack is published to the Power BI app for the board audience and delivered as a PDF by subscription
- AutomationA threshold breach raises a data alert into the management channel in Teams the moment the refresh completes
- SystemEvery run, source, control total, commentary entry and publication is logged, so any figure traces back to its extract
Human-in-the-loop model
Automation handles
- Scheduled extraction from every source, with retries and a record of what was taken and when
- Completeness and reconciliation controls, including holding back publication when a control total disagrees
- Model refresh, publication to the app audiences, the PDF subscription and the alerts
- Reminders to commentary owners and the assembly of their text into the pack
People decide
- What the KPIs mean. Definitions, hierarchies and thresholds stay owned by finance, and every change is versioned
- The commentary: why a number moved and what will be done about it
- Whether a failed control is a data problem or a real business event, before the pack is released
- Who belongs in which audience, and which figures are restricted by row-level security
Before and after
Systems and integrations
We do not add technology to make an architecture look serious. Every element below has a specific job in this process.
Inputs
- SAP S/4HANA trial balance and cost centres
- CRM pipeline and contracts
- HR headcount and absence
- operational system delivered hours
- an Excel Online template for figures held nowhere else
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- Power Automate
Target systems
- Power BI semantic model and workspace
- the Power BI app with board, function and site audiences
- SharePoint library with the extracts and the PDF
Human touchpoints: commentary cards in Microsoft Teams; threshold alerts in the management channel; the Power BI app in Teams or on a phone
Technologies used
scheduled extraction, queues, retries, run history and audit
AAPI access to the ledger and the reporting library
Athe governed model, the published pack, the PDF and the alerts
Awhere capacity is held: Direct Lake over OneLake, viewers without per-user licences from F64
Acommentary cards to owners, reminders and alert routing
Awhere commentary is written and where the pack and its alerts are read
Asource extracts and the controlled template for figures held nowhere else
Atrial balance, cost centres and period data through standard interfaces
AIllustrative economic model
What it is worth, with the arithmetic shown.
Every step of the arithmetic is shown so it can be checked, and every input illustrates the scenario above rather than a measurement from a client. Twelve analyst-days a month is three analysts for four working days, a working day is 480 minutes, and €35 an hour is a fully loaded cost for a financial analyst in Central Europe. What is shown is capacity released, not headcount removed.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- The pack is ready the working day after the ledger closes rather than on the fifth, so decisions are taken a week earlier
- Twelve analyst-days a month move from assembling numbers to explaining them, which is the work the group pays for
- One definition per KPI for every audience; site, country and board views are cuts of the same model
- Questions in the meeting are answered in the room by opening the figure, instead of becoming an action point
- A threshold breach reaches the responsible manager the day it appears in the data, not at the next review
- Adding a site or a country means mapping one more source, not two more spreadsheets and another day of merging
The management view
- Any figure traces from the visual through the model to the dated extract and the run that produced it
- Commentary becomes a record: who explained which variance, when, and what they committed to
- The reporting calendar becomes predictable and stops depending on which analyst is available
- A restatement is visible rather than silent: a number that changes after publication changes with a version and a reason
Board-level KPIs
Security and governance
Where the data sits and who can see it.
- Robots read the source systems with dedicated accounts holding reporting rights and nothing more; Orchestrator resolves their passwords from the tenant's key vault at run time
- Access follows Microsoft Entra ID security groups: audiences in the Power BI app and row-level security in the model, so a site manager opens the same report and sees their own site. Row-level security applies to viewers, so the board and site audiences are viewer audiences
- Extracts, the model and the pack stay in your Microsoft 365 tenant; the automation runs in the UiPath Automation Cloud EU region where European residency is required
- A change to a KPI definition is proposed, approved by its owner, versioned and visible in the model's history
- The audit record covers the run, the source, the control totals, the commentary author and the publication, so a restated figure can be explained later
Why now
Power BI Premium per-capacity SKUs have been retired in favour of Fabric capacities, so most organisations are revisiting their reporting platform anyway; building the pack on a governed model during that move avoids paying twice
Twelve analyst-days leave the business every month whether or not the pack is believed, which on the modelled numbers is about €3,360 of finance capacity spent on copying
Scheduled extraction through connectors, semantic models that refresh on their own and Adaptive Cards in Teams are ordinary configuration today; the effort has moved from building a platform to agreeing definitions
Relevant executive roles
The pack stops being an artefact three people assemble and becomes a model finance owns, published on a date you can put in the calendar
Decisions on price, capacity and hiring move into the first week of the month, taken on numbers nobody argues about
Site, utilisation and delivery metrics sit in the same model as the financials, so an operational problem and its margin effect are one view
Management reporting leaves fourteen spreadsheets on personal drives for a documented refresh, an access model and an audit trail
Common questions and objections
Most deployments put reports on top of the same manual extracts, so the spreadsheet work moves behind a dashboard. What changes here is the layer underneath: scheduled robot extraction, control totals that can hold back a publication, and one model where the definitions live.
Then the robot uses the interface a person would use, unattended and on a schedule, producing the same dated extract. That is the normal case for older systems, and it is why we begin with an extraction inventory rather than a model.
Not to start. A Pro workspace runs a scheduled import model and the app perfectly well. Capacity matters for Direct Lake over OneLake, more frequent refreshes, viewers without a per-user licence from F64 upwards, or Copilot in Power BI. We size it once the model exists.
When this is not the right solution
- If the pack is genuinely one report from one ERP, a scheduled export and a single report are cheaper than a model
- If the KPI definitions are still disputed between finance and the business, settle that first; automation makes a contested number arrive faster, not become right
- If an ERP or HR replacement is scheduled within months, we map definitions now and build the extraction after the cutover
A question for the next management meeting
On which working day of the month does this board see final numbers, and how many people-days does the company spend to get them there?
Implementation approach
What we deliver, and what we need from you to start.
We deliver
- A definition workshop: the KPI list, an owner per definition, hierarchies and board thresholds
- Extraction robots for each source system, with schedules, retries and a record of every run
- Completeness and reconciliation controls that decide whether the pack may publish
- The Power BI semantic model, the report pages and the paginated pack behind the PDF
- The commentary workflow in Teams: owners, deadlines, reminders and an archive
- Threshold alerts, the app audiences and the row-level access model
- One month in parallel with your existing pack, then the switch with hypercare
We need from you
- The last two management packs and the fourteen source files behind them
- A named owner for every KPI definition and every commentary section
- Read-only accounts for the source systems and a workspace in your Power BI tenant
- Your distribution list and who is allowed to see what
Stages
Definitions
The KPI list, owners, hierarchies and thresholds; where each figure comes from
Extraction
Robots, schedules and control totals for every source
Model
The semantic model, measures, security and the report pages of the pack
Commentary and alerts
Teams cards, deadlines, reminders and the threshold rules
Parallel run
One month against the existing pack, reconciled line by line
Handover
Publication to the board audience, hypercare and the runbook
Departmental. Effort is driven by the number of source systems, the quality of their exports, and how far the KPI definitions differ today.
How many of the fourteen files behind your board pack could a robot read on its own?
Send us the list of source files behind your last management pack, with the owner and the report each comes from. We come back with an extraction inventory and the KPI definitions that would have to be agreed first.
Show us your board pack sourcesThe neighbouring process usually has the same problem
Stop building the daily cash number by hand from four bank portals; it arrives late and nobody quite trusts it.
View solution Management & planningOne morning view of every brand, site and workshopFour DMS, four importer portals and nine sites, and the board sees last month in the third week of this one.
View solution Finance & accountingAccount reconciliations and journals that close themselvesStop closing the month by downloading balances into Excel and chasing sign-offs by email.
View solution Management & planningBudget and forecast collection without version chaosStop spending a quarter of an analyst's year collecting spreadsheets from cost-centre owners.
View solution Case studyExecutive assistant on TeamsAn AI agent available on Microsoft Teams answers board questions about finance, sales and operations in seconds.
View case study Case studyThe morning executive brief, daily at 7:00The board starts the day with the full picture — before anyone has to ask for it.
View case studyIndustries we deliver this in most oftenManufacturing & industryRetail & e‑commerceServices & IT