AI Assist is the middle tab on the Budgets page, alongside Budget vs Actual and Budget list . It holds the parts of budgeting URBI prepares for you: a year-end forecast, a written explanation of this...
Last updated
AI Assist is the middle tab on the Budgets page, alongside Budget vs Actual and Budget list. It holds the parts of budgeting URBI prepares for you: a year-end forecast, a written explanation of this period's variance, a suggested month-by-month spread for next year's figures, and a one-click draft of next year's budget.
The tab appears only on properties whose plan includes AI budgeting. If you do not see it, that is why. The weekly forecast is prepared only for those properties as well, so a property without the tab is not given one.
What it answers: where this budget is likely to land by the end of the fiscal year, as a range rather than a single number.
The card shows when the forecast was generated, then Band low, Band high and Confidence, followed by the forecast's own written summary. Where the model used outside figures, it lists them underneath.
The band is a year-end range. Band low and Band high each start from what has already been spent this year and add the low or high estimate for every month still to come, for every account. Each remaining month is projected from the property's own seasonal pattern for that month of the year, so a quiet winter month is not forecast like a busy summer one. Confidence shows the weakest confidence found in any of those months, not the best one.
Which months count as spent. Only finished months count as already spent. The month in progress is forecast in full, like the months still to come, so a half-finished month is never read as a quiet month and never counted twice. Which month is in progress is worked out in the property's own time zone. Budget lines held on a Reserve account are listed by their real names.
How it is prepared. URBI builds this forecast automatically every Friday, for every Active budget on a property whose plan includes AI budgeting, using that property's own reporting basis (cash or accrual). It is a stored result, not a live calculation, and the card says so: "Forecast runs every Friday · stored model (not live)". Once the stored copy is more than a day old it adds "Forecast may be out of date. Consider updating the model.", which will be true on most days of the week and is not a fault.
Before the first one exists the card reads "No stored forecast yet. New forecasts are prepared every Friday, or use Update model when you have manage access."
When a fiscal year has no future months left, for example a year that has already finished, there is nothing to project, and the card says "Forecast unavailable because there are no future periods left in this fiscal year." It does not show a band of zero.
For a budget whose fiscal year has not started yet, the card does not project anything either. It says "The forecast starts when this fiscal year begins, in January 2027.", naming the first month of that year.
Updating it yourself. Click Update model. This queues the work in the background and confirms with "Forecast update queued, refresh in a moment". The figures on screen do not change when you click, so reload the page in a minute or two to see the new band. You can do this once every 24 hours; while you are inside that window the card tells you when it is next available, as "Update available again after ..." with the time. If the job fails you get "Forecast update failed" and your once-a-day slot is handed back, so you can try again straight away. Only the signed-in person who presses the button can start an update, and a failed start never leaves the button locked.
What it answers: if this annual figure were spread across the twelve months the way this property actually spends, what would each month look like?
The panel is labelled "Suggestion only. It does not change the budget." Click Generate. URBI reads up to three years of the property's own posted history and proposes a twelve-period spread for each line. It reports what it produced as a count and then the income and the expense separately, for example "2 allocations across 12 periods: $12,000.00 of income and $30,000.00 of expense." When URBI cannot tell every suggestion's side, the summary only reads "Review each suggested amount below." rather than showing totals it is unsure of. Each suggestion is listed with its per-period amounts and its yearly figure.
For a grouped budget this matters: the panel lists your categories first, then individual accounts, so money you hold on a category is included rather than quietly left out. Budget lines held on a Reserve account are listed too, by their real names.
This is a proposal you read. It does not change any budget figure by itself, and a successful run says so: "Suggested monthly spread ready. The budget is unchanged." To use a suggestion, enter the amounts in the budget grid yourself.
Before you have generated one, the panel explains that Generate suggests how each account's annual budget could be spread across the 12 months, based on past spending, and that nothing in the budget changes.
How long it takes. The button spins while the work runs and usually returns within a few seconds. If it is still going after about thirty seconds you get "Allocation schedule generation is still in progress. Check again shortly." The job normally finishes anyway, so reload rather than pressing Generate again. If it fails outright you get "Schedule generation failed" and the previous proposal stays on screen.
What it answers: why this period is off plan, in sentences you can paste into a report.
Click Generate narrative (the button reads Regenerate once one exists) and URBI writes a short explanation of the current period's variance from the budget's own group summaries and the figures that stand out. You get one piece of prose, not a scorecard.
Reserve is left out. Reserve is excluded from budgets entirely for now, so the narrative explains Operating spending only and never counts Reserve spending as a variance. See Reserve in Budgets and Reports.
URBI also refreshes it on its own on the 5th of each month for every Active budget.
Before one exists the panel explains itself in the property's own wording for the people who read it, along the lines of a variance explanation for the current period that refreshes monthly or on demand.
How long it takes. The button waits for the answer, normally a few seconds. If it runs past about thirty seconds you get "Variance narrative generation is still in progress. Check again shortly." and the work carries on in the background. A success shows "Variance narrative updated"; a failure shows "Narrative generation failed" and leaves the previous narrative in place. Pressing the button again when nothing has changed returns the stored answer immediately rather than paying for a new one.
What it answers: which open work the forecast could not put a number against.
This panel has no buttons. It lists open service tickets that the forecast model saw but could not commit money to, each with the reason it was skipped, which today is always that the ticket carries neither a cost estimate nor a schedule. Adding either one to the ticket brings it into the next forecast.
When the list is empty it reads "Open service tickets without committed forecast amounts appear here when the forecast model runs."
What it answers: give me a starting point for next year that reflects how this year actually went.
The panel is headed Next-Year Draft and the button reads Generate FY (next year) Draft. It also shows how many completed months you have, out of twelve.
When you can use it. Nine completed months of actuals are needed. Until then the button is disabled and reads "Available after nine months of actuals are recorded in this fiscal year."
How each line is drafted. The draft is a carry-forward of the current budget, adjusted by what the year has shown. For an account that has both a budget and spending so far, URBI starts from what was actually spent in the completed months against what was planned for those same months, and carries that relationship onto the whole year. An account that has not spent anything yet, such as an annual insurance premium or a seasonal contract, is simply carried forward at its current amount rather than drafted at nothing. URBI also adjusts for the reallocations the property made during the year, so an account that repeatedly borrowed budget is nudged up and one that repeatedly gave budget away is nudged down. The result is spread across the months using the property's own seasonal pattern, and each line's baseline is set to the current budget's annual figure so the percentage columns have something to measure against. The draft carries the current budget forward and does not try to forecast growth. Budget lines held on a Reserve account are carried into the draft under their real names.
What you get. A new Draft budget for next fiscal year, confirmed with "FY (year) draft created", for example "FY 2027 draft created". The screen does not take you there, but the new draft is added to the budget list, so open the Budget list tab to find it and edit it like any other draft.
If it takes too long. You get a blue message, "The FY 2027 draft is still being prepared. It will appear in the budget list when it is ready." (with the year you are drafting for), not a red error. The draft is still created in the background, so check the Budget list before trying again. A second attempt on a year that already has one returns "A draft for next year already exists".
At the bottom of the tab, Hero sits with this budget already in context, so you can ask about variance, the forecast, or reallocations in your own words instead of reading them off a table. On a narrow screen it collapses to an Open Hero Budget Chat button that opens full screen. Opening the tab no longer adds a new empty conversation to your Hero history each time.
Who Hero will answer. Hero is stricter than the Budgets screens. It answers budget questions only for people whose Budget permission is explicitly turned on for the property. If your permission is off, or your administrator has never set it for you, Hero does not answer and tells you budget access is not enabled for your account. On the Budgets screens themselves, a permission that was never set still counts as allowed. Ask your administrator to switch the Budget permission on if Hero will not answer you.
What Hero says. Questions such as "which accounts are over budget?" and general questions about the budget, without naming a year, now get an answer. Hero looks at every line of the budget, however large, and gives you one row per line with its year-to-date budget, actual and variance, plus monthly totals. Income and spending are kept apart in those totals, so a budget with $500,000.00 of income and $500,000.00 of spending is never read as one $1,000,000.00 total, and each line says whether it is income or an expense.
Which year Hero reads. Hero uses your property's own fiscal year, not the calendar year. A question that names no year is answered from the current fiscal year's Active budget. If Hero cannot read your property's accounting settings, it says so rather than guessing the year.
Reserve lines. Hero lists Reserve lines labelled Reserve and reports their spending on its own. It never counts them in a budget total, because Reserve is excluded from budgets entirely for now.
If the stored Schedule Allocation or Variance Narrative cannot be read, that panel is replaced by a card saying it could not be loaded and to refresh the page. Nothing is lost: both are stored results, and reloading brings back whatever was last generated.