For budget and programme analysts in government
No obligation, no commitment, no disbursement, no reprogramming, no apportionment judgement, and no determination about what a fund may lawfully pay for. What can be carried is the assembly and the reconciliation reporting — the work that turns several systems with different clocks into a single current picture somebody can act on.
A single programme is rarely funded from one place. There is a base appropriation, a supplemental, a pass-through from another level of government, a grant with its own period and its own allowable costs, a fee account, and sometimes a carryover balance whose rules differ from all of them.
Each carries its own calendar. One expires at the end of the year, one is available for several, one has a period of performance that started before the fiscal year and ends inside it, one has a liquidation window that outlasts the availability window. The project has one plan and the money has six.
The rules about what each pot may lawfully pay for are not interchangeable, and the consequence of getting it wrong is not a rounding error — it is an improper payment and a finding. So the analyst carries a mental map that exists nowhere else, and everybody knows the map lives in one person.
Reconciliation happens by hand, usually in a spreadsheet the analyst built, because the financial system, the grants system and the programme’s own tracker each hold a piece and none of them holds the whole. The spreadsheet is more current than any of the systems and it goes home on a laptop.
And the year-end is a cliff. Funds expiring on a date create a compressed period where obligations must be made correctly and fast, which are the two things that do not go together — and the errors made in that window are found the following spring by somebody who was not there.
No system holds the whole position, so producing an accurate current balance across sources is manual work — which means decisions get made against a stale figure or against somebody’s recollection.
That is the actual cost, and it is invisible in every budget report. Not the hours the analyst spends, but the decisions made without the number because getting the number was too slow to be worth it.
The first act is therefore to assemble rather than to automate: for one programme, every funding source with its authority, its availability window, its performance window, its liquidation window, its allowable-cost rules as your own documents state them, and its current position at each of the eight states. Written down once, in one place, for the first time.
What can then be operated is the reconciliation reporting. Pull the current figures from each system on a cadence. Compare them. Where two of your own systems disagree, report the difference with both figures and both sources rather than choosing one. Surface what is approaching expiry with enough lead time to matter. Assemble the package a reviewer or an auditor asks for.
What never moves is every act that touches money or judgement: committing, obligating, disbursing, reprogramming, apportionment, determining allowability, or deciding which pot pays for something. Those are exercises of authority held by named officials under law, and a system that made one would be creating an improper payment with a clean audit trail, which is worse than one without.
Time to produce an accurate current position across sources — measured by elapsed time from request to a defensible figure, timed on a real request rather than estimated.
Disagreements between your own systems — measured by count and size of differences surfaced, which is generally unmeasured until an audit finds them.
Lead time before funds expire — measured by days between an expiry being surfaced and the expiry date, compared with the previous year.
Decisions made against a stale figure — measured by count of programme decisions where the position used was more than a stated age.
The reconciliation living in one person — measured by whether a second analyst can produce the same position from the written source map without asking the first.
Year-end error rate — measured by count of corrections required after year-end close, against the prior year.
Time to assemble an audit package — measured by elapsed hours from an auditor request to a complete package with support attached.
any commitment, obligation, disbursement, reprogramming, apportionment or allotment; any determination of allowability, availability, or purpose; and any decision about which fund pays for what. Those are exercises of authority held by named officials under fiscal law. Nothing here writes to a financial system at all.
Everything is read. The financial system, the grants system, the programme tracker and the procurement record are sources; none of them is written to. A system that can write to a financial record is a system that can create an obligation, and that is not a capability worth having here at any level of care.
The source map — authority, windows, allowable-cost rules — is built from your own adopted documents and your own guidance. Nothing here interprets fiscal law. Where two of your own documents disagree, that is reported as a question for your budget officer or counsel.
The reconciliation output is a report, not a ledger. It is explicitly not authoritative and says so on its face, because a convenient summary that starts being treated as the record is exactly how a shadow ledger is born.
Where an interface does not exist, that is a limitation and the source stays manual. Screen automation against a financial system is refused for the same reason writing to one is.
The read-only property is the whole trust argument here and it is structural rather than configured. There is no tier at which this writes to a financial system, and no permission an administrator could grant that would create one.
The reconciliation report carries its own status on its face: derived, as-of a timestamp, from named sources, not authoritative. A report that does not say what it is becomes the record within a year.
Financial data stays inside your tenancy, on your retention schedule, exportable by you, and is not used to train anything serving another organisation.
And every action leaves a record naming actor, authority, time, source and result — which for a read-only arrangement means an auditor can see exactly which figures were read, from where, and when.
The chief financial officer’s question is whether anything can touch the books. Nothing can, and it is worth testing rather than accepting: ask for the access model and confirm that the credential used is read-only at the system level rather than by convention.
Internal audit will want the source map and the report’s own status line. A derived report that presents itself as authoritative is the finding they are looking for, and this one is built to fail that test deliberately.
Counsel’s question is whether anything interprets fiscal law. Nothing does. The rules in the map are transcribed from your own adopted guidance, and a disagreement between two of your documents is escalated rather than resolved.
The financial system owner should confirm the read pattern and its load. A reconciliation cadence that hammers a production financial system at close is a real operational risk and it is sized with them rather than around them.
One programme’s funding sources mapped to a single document — authority, availability, performance and liquidation windows, allowable-cost rules — with read-only access and no reconciliation running yet.
The mapping phase is frequently the entire value. Most programmes have never had their funding structure written down in one place, and the act of doing it surfaces windows nobody was tracking and rules that exist only in one analyst’s memory.
Stopping there is a legitimate outcome. A written source map is an asset the agency keeps, it survives the analyst, and it is the thing that makes a second analyst possible. If it turns out the programme has one fund with one window, this page is not about your problem.
If it continues, the first reconciliation runs read-only on a stated cadence, reporting differences rather than resolving them, with the output labelled derived from the first day.
That is a defensible position and it may end the conversation, which is fine. If it does not, the thing to test is the access model rather than the promise: the credential should be read-only at the system level, granted by your system owner, revocable by them, and scoped to the specific objects a reconciliation needs. If your policy does not permit even that, the mapping phase can still be done entirely from documents your budget office already holds, with no system access at all.
Very likely true, and the risk is not correctness — it is that the reconciliation exists in one person and in a spreadsheet on their laptop. The mapping phase addresses exactly that and delegates nothing: the source map is a document your office keeps, and its value is that a second analyst can produce the same position without asking the first. If your programme already has that document, you do not need this.
It will, unless it is built to resist it, which is why the output declares itself derived with an as-of timestamp and named sources on its face rather than in a footnote. That is a real risk you are right to name — a convenient number becomes the number. If your internal audit reviews the report and thinks it could still be mistaken for authoritative, the label is the thing to change before anything else.
Which is why nothing here interprets them. The map transcribes your own adopted guidance, and the question "may this fund pay for this" is never answered — it is routed to your budget officer with the rule at issue named. If a transcription is wrong, that is a document error your office can see and correct in the map, which is a different and much smaller failure than a system that reasoned its way to an improper payment.
Agreed, and the mapping phase is the only part worth doing near a close — it needs no system load and no reconciliation cadence. Starting a read pattern against a production financial system during year-end close is an operational risk to the close itself, and the system owner would be right to refuse it. Map now if it helps; run nothing until the books are closed.