Home · Solutions · Legal & compliance
Solution · Legal & complianceRetention is arithmetic; a robot keeps the lists, the archivist keeps the decision
The archive that knows what it may destroy and when
Closed cases are checked against the file plan every night, retention is calculated continuously, and transfer and appraisal lists reach the archivist ready for a decision.
Executive summary
Cases close, files do not: categories unchecked, transfer lists retyped, appraisal postponed another year.
The design starts from a constraint rather than a tool: the robot never changes a case file and never destroys a record.
The archive learns that a case has closed on the night it closes, so the work spreads across the year instead of arriving in one fortnight.
the archive register and both eligibility registers on SharePoint; Microsoft Purview retention labels on electronic case files; Power BI reporting
Business problem
Records management
Every case takes a reference built from the symbol of the unit that runs it and an entry in the file plan. That entry carries an archival category: A, B with a retention period, or BE. With the closing date it decides when the file moves to the archiwum zakładowe and when it may be appraised. It is set in the first minute of a case, then never looked at again.
What follows is predictable. Categories are copied from the last case that looked similar, and entries withdrawn from the plan stay in use because nobody circulated the change. A case closes in the system while its paper file stays in the unit, because transfers happen once a year and only if somebody organises the round. The archivist then receives boxes and a spreadsheet typed from their labels.
She carries that risk with the smallest team in the building, heads of units lose a fortnight before every round, and the office secretary is asked for numbers that do not exist. Records whose retention expired years ago occupy rented space, appraisal is postponed because proving which qualify means opening files, and a court asking for one file gets an answer built on somebody's recollection.
How it works today
What we find where the case system is used well and the archive still runs on paper habits.
- PersonA clerk closes a case, writes the closing date on the cover and copies the category from the last similar case
- WaitingThe file stays in the unit until the next transfer round, annual at best and slipping when staff are short
- PersonBefore the round somebody types a transfer list from the folder labels, and afterwards the archivist retypes the accepted rows into her own register
- Risk of errorFiles with no category, or an entry that no longer exists in the plan, are found at the archive counter and sent back
- WaitingRecords whose retention has run out stay on the shelf; the appraisal round is planned every year and postponed when nothing forces it
- Risk of errorRetrieving one file for a party or an inspection depends on remembering who handed over what
Why the current process costs more than it appears
The most expensive part of this process has no cost line.
- Typing the same rows twice is the visible half and the smaller one. Every returned file goes back to a unit, waits for the person who ran the case, and reappears a round or two later.
- Categories set at registration and never verified decay quietly. A file under the wrong entry is either destroyed too early, which cannot be undone, or held for decades nobody budgeted for.
- Postponing an appraisal round costs nothing in the year it is postponed, which is why it survives every budget, while shelving is rented for records the office could have appraised. The bill arrives later, as a backlog too large for one year to clear.
Cost of inaction
An appraisal round that slips by a year costs nothing anybody can point at, and that is the difficulty. The third row prices the backlog once: 15,000 files at eleven minutes each is 2,750 hours, roughly €55,000 at the same hourly cost, a year of one person reading folder covers.
What accumulates is not only cost. Shelving is rented for records that could have been appraised, retrieval depends on two colleagues' memory, and every postponed round adds a year of files the office cannot describe.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A city office of about 420 staff in fourteen organisational units, with an electronic case system in daily use, Microsoft 365 across the office and an archiwum zakładowe run by one archivist and a deputy.
9,400 cases closed a year, roughly 780 a month, behind them some 15,000 files with no confirmed category; one transfer round a year and no completed appraisal round in four.
Categories are set at registration and checked only at the archive counter. Transfer lists are typed from folder labels, then retyped into the register. Retention is worked out by hand on request.
Eleven minutes per file to open it, match its symbol against the plan, confirm the category and the date, and write its row. Across a year of closures that is a job nobody can schedule, so it happens late and under pressure.
Robots read closed cases every night, check each against the file plan, and raise a correction task in Microsoft Teams wherever an entry, a category or a date is missing. Retention is calculated from the closing date and the category, so both eligibility lists stay current and drafts reach the archivist as an approval.
The annual scramble becomes a review: lists arrive complete, corrections are cleared through the year, and the office can state on any working day what falls due. The figures are a model.
Proposed solution
The design starts from a constraint rather than a tool: the robot never changes a case file and never destroys a record. It reads, checks, calculates and drafts; anything that alters a case, a category or the fate of a record is done by the official answerable for it.
Each night it reads the cases closed since the previous run and tests each against the file plan in force: entry present and still current, category matching that entry, closing date, owning unit and location recorded. Cases that pass go into the eligibility registers; the rest become a task for the case owner in Microsoft Teams naming the missing field, and the clerk corrects the record in the case system.
Retention is then arithmetic run nightly across the whole population rather than once a year across a sample: closing date plus the rule the plan attaches to that category. Two registers stay current, one of files due for transfer to the archiwum zakładowe and one of records past retention. Ahead of a round the robot drafts each unit's list in the required form, and the archivist accepts it, returns single lines or sends it back. Where the file is electronic the same category drives a Microsoft Purview retention label.
UiPath Orchestrator queues, time triggers and audit; UiPath Action Center tasks completed inside Microsoft Teams; Microsoft Teams Approvals app; SharePoint libraries with metadata and versioning; Microsoft Purview retention labels and disposition review; Power BI
The mapping from file-plan entry to category and retention rule; the completeness checks; the retention calculation; the two registers; the list generators; the correction and approval flows; the evidence record and the report
Read access to the case system through its API, a read-only export where a departmental system has none, and a one-off load of the archive register
How the automated process works
- AutomationA nightly trigger starts the run; every case closed since the last one is read and queued in UiPath Orchestrator
- SystemEach case is checked against the plan in force: entry present and current, category matching the entry, closing date, owning unit and location recorded
- PersonAnything incomplete becomes a correction task in Microsoft Teams for the case owner, who corrects the record in the case system itself
- AutomationRetention is recalculated across the whole population from the closing date and the category, and both eligibility registers are refreshed
- AutomationAhead of a round, draft transfer and appraisal lists are generated per unit and filed in that unit's SharePoint library
- PersonThe archivist accepts a list, returns single lines or rejects it in Teams; the approval, its author and the version are written to the register, and the appraisal list goes forward under the office's own procedure
Human-in-the-loop model
Automation handles
- Reading closed cases every night and checking each against the file plan, instead of a year's worth once
- Calculating retention from the closing date and the category, and keeping both registers current
- Drafting lists in the required form, with references, dates, categories and units filled in
- Writing the evidence: what was read, which check failed, who was asked, what they answered
People decide
- The archival category wherever the plan leaves a choice, or where a case does not fit the entry it was registered under
- Whether a file is complete and may leave the unit for the archive
- Whether records past retention are appraised, held longer or reviewed; nothing is destroyed by a robot, and every disposal carries a name
- The file plan and the retention rules, which stay with the office and its archivist
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
- the case system API (closed cases, entries, categories, dates, units)
- the file plan and retention rules from the office's own instruction
- the existing archive register
- the list of file locations
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Action Center
Target systems
- the archive register and both eligibility registers on SharePoint
- Microsoft Purview retention labels on electronic case files
- Power BI reporting
Human touchpoints: correction tasks in Microsoft Teams; the archivist's approval queue in Microsoft Teams; the transfer and appraisal rounds themselves
Technologies used
read closed cases on a nightly trigger, run the checks, queue the work, retry and log every run
Acorrection tasks and the archivist's approval queue, completed without leaving Teams
Alists accepted, returned or rejected with the document attached and the action audited
Alibraries holding draft and approved lists, the evidence and the version history
Aapplies the same rule to the electronic side of a file and holds the disposition decision
Acompleteness by unit, ageing of the backlog, what falls due for transfer or appraisal
Aclosed cases, entries, categories, closing dates and units, read through a documented interface; EZD RP publishes one
AIllustrative economic model
What it is worth, with the arithmetic shown.
The file plan does this arithmetic in real life; here it is done on stated assumptions instead. 780 files a month is 9,400 closures spread evenly, eleven minutes is what checking one file and writing its row costs at a desk, €20 an hour is a fully loaded cost of an administrative post. Nothing was timed in any office, and the result is capacity returned to the units, never a saving and never a post.
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 archive learns that a case has closed on the night it closes, so the work spreads across the year instead of arriving in one fortnight
- Transfer lists arrive filled in and consistent with the plan, and the archive counter stops being the checkpoint where files are turned away
- The office can say on any working day which records it may appraise, and prove it from a register rather than by opening boxes
- Storage becomes a decision, and the capacity now spent on checking and retyping returns to the units, the only kind of return a public budget can bank
The management view
- The office secretary can see how much of the closed year is properly described, unit by unit, without commissioning a count
- Rounds stop being events: because the lists are current, a transfer or an appraisal becomes a review, not a project
- Judgement stays where it is answerable, and every decision carries a name, a date and a list version
- Equal treatment is demonstrable, because one rule runs against every case on the same schedule, whichever unit handled it
Board-level KPIs
Security and governance
The automation holds exactly the rights it needs, and not one more.
- Reading is the robot's only right in the case system. It cannot change a case, a category or a status, so a correction is always made by the clerk who owns the record and a disposal always carries an official's decision.
- Service accounts are separate from personal logins, hold read rights on the case system and write rights only on the registers, and are governed by group membership in Microsoft Entra ID; their secrets sit in a credential store.
- Every run leaves an inspectable record: what was read, which check failed, who was asked, what they answered, and which list version was approved by whom. The evidence is a by-product, not a reconstruction.
- Case data stays in the office's own Microsoft 365 tenant and the robots run in the EU region of UiPath Automation Cloud. Registers carry references, dates, categories and units; the names of parties stay in the case system. Retention labels and disposition review sit in the Microsoft Purview range licensed with Microsoft 365 E5 or a compliance add-on.
Why now
Offices still running EZD PUW work to a horizon somebody else set: NASK stopped distributing it on 30 April 2024 and declares support until 31 March 2031. Archive data described before a migration costs a fraction of the same data described during one.
More of the case file is born digital every year. The Ministry of Digitisation reported over 100 million items handled in e-Doręczenia by 5 August 2026, more than 50 million in 2026 alone, so much of what closes now will be appraised as electronic records.
The arithmetic runs whether or not a round happens: the modelled €2,860 a month of checking continues, and the backlog it feeds grows by another 9,400 files a year.
Relevant executive roles
Completeness stops being an annual discovery; the state of last year's closed work is a number available on any day
Corrections arrive as single tasks through the year, not a fortnight of checking before every round
Lists arrive complete and consistent with the plan, while the judgement and the right to return one stay with the archive
One read-only integration and standard Microsoft 365 surfaces, not another register in a spreadsheet
Common questions and objections
It does not. The robot checks whether a category is there, whether it matches the entry the case was registered under, and whether the closing date is recorded. Anything missing becomes a task for the person who ran the case.
That is what the first full run measures. It reports which entries are in use, which no longer exist and which units rely on them, and that report is usually what finally gets the plan updated.
The volume reaching her falls, because lists arrive complete and corrections were cleared through the year by the units that caused them. She approves a list instead of assembling one, and can return a single line.
When this is not the right solution
- Offices closing a few hundred cases a year, where a disciplined annual round in a spreadsheet costs less than any integration
- No agreed file plan, or a plan whose entries cannot be mapped to categories and retention rules; there is nothing to check against yet
- Closed cases that exist only on paper with no register a machine can read; the first step there is registering the backlog
A question for the next management meeting
9,400 cases closed in this office last year: for how many of them could the archive produce a category, a closing date and a place on a list this afternoon?
Implementation approach
What we deliver, and what we need from you to start.
We deliver
- A reading of your instruction and file plan: entries in use, categories, retention rules, and where the plan leaves a decision to a person
- The completeness checks against closed cases, agreed with the archivist and the heads of units
- The retention calculation and both registers, with the evidence recorded behind every entry
- List generators in the required form, and the return path for a list the archivist does not accept
- Correction tasks and the approval queue in Teams, Purview labels, the report, and a controlled start on one unit
We need from you
- The instruction and file plan in force, plus the entries departments actually use
- Read access for a service account, and one export of a full year of closed cases
- A named archivist and one records owner per organisational unit
- A decision on the limits: which categories always require a person, which units wait for phase two
Stages
Discovery
The file plan, the entries in use, closed-case data and the state of the archive register
Design
Checks, retention rules, list contents, the approval path and the security model
Build
Robots, registers, list generators, Teams tasks, Purview labels and the report
Validation
A year of historical closures run through the checks, compared with what the archive accepted
Go-live
One unit under the archivist's supervision, then the rest with hypercare and the backlog in waves
Departmental. Effort follows the number of systems holding closed cases, how far the plan in daily use has drifted from the one on paper, and whether the archive register is machine-readable.
The file plan says what may be destroyed and when. Nobody has the fortnight to apply it.
Send us one year of closed-case data: reference, file-plan entry, category, closing date and unit. You get back the share that already passes every check, the entries in use that are no longer in your plan, and the size of the backlog behind them.
Check one year of your closed casesThe neighbouring process usually has the same problem
Your documents sit on a network drive with no metadata, under retention rules nobody applies.
View solution IT & servicese-Doręczenia handled like post, not like a second inboxFour intake channels, one register typed by hand, and delivery evidence filed after the fact.
View solution Legal & complianceInformation requests answered inside the statutory clockRequests arrive on four channels, the clock starts on all of them, and the register is written afterwards.
View solutionIndustries we deliver this in most oftenPublic sector