Nothing here is financial advice, and none of it replaces your own checks.
These prompts help you turn actual results, the approved budget and revised operating assumptions into a reforecast bridge that executives can act on. They are for FP&A teams preparing a review pack, where a plausible narrative is not enough and every material number needs a source.
Nothing here is financial advice, and none of it replaces your own checks. Use the prompts to draft and test management-reporting work products. Your finance lead and relevant decision-makers remain responsible for the figures, assumptions and decisions.
Key point
Start with comparability
A bridge is only useful when actuals, budget and forecast cover the same period, scope, currency and sign convention.
1. Check the source documents first
Start with Reforecast source and reconciliation check. Paste the actuals, budget and assumptions as separate labelled blocks. Do not combine them into one spreadsheet extract without labels. The prompt needs to identify whether, for example, the budget is full year while actuals are year to date, or whether cash is reported at group level but operating costs are for one business unit.
Use its source-control table as the working record for the pack. Assign an owner to every unresolved item before you start explaining movement. Common problems include:
- Revenue stated net of rebates in actuals but gross in budget.
- A hiring assumption with no start month or fully loaded cost.
- Cash movements that omit capex or working capital.
- KPI definitions that changed between planning cycles.
- Forecast assumptions stated as percentages with no base volume.
Watch out
Do not force a number into an unresolved gap
An unexplained plug may make a bridge add up, but it removes the evidence executives need to trust it.
2. Build the four bridges
Run Revenue, cost and cash bridge only after the source check shows that the inputs are comparable, or clearly records the exceptions. This prompt separates revenue, operating costs, closing cash and operational drivers. Keep those bridges separate even if your final pack shows a combined headline table.
The split matters. A revenue shortfall can reduce cash, but an increase in receivables days can also reduce cash without changing revenue. Likewise, delayed hiring may improve the near-term cost forecast but create a delivery risk. The bridge should make those relationships visible rather than hiding them in one variance line.
Set a materiality threshold before running the prompt. Use the threshold already applied in your planning process. Ask the model to name material drivers individually and group only genuinely immaterial items. The output should state a start point, an end point and a residual for every financial bridge.
Check
A bridge has passed its first test when the start point plus every quantified movement equals the end point, with a zero explained residual.
If a residual remains, do not label it “other” unless you have identified its source. Keep it as unresolved and assign it for follow-up.
3. Separate evidence from interpretation
Use Driver narrative and sensitivity review to write the explanation around the numbers. It asks for evidence, confidence and what could change the outcome. This stops the common error of presenting an operational assumption as a proven cause.
Read the “Timing versus structural” section closely. Timing items should have a clear expected reversal or rescheduling point, such as a contract start date moving into the next month. Structural items should represent a more lasting change, such as a revised price list, a sustained cost increase or a changed demand assumption. If the source material does not establish the classification, keep it unresolved.
The sensitivity section is deliberately constrained. It should report only alternative cases supplied by your team. Do not ask the model to create upside cases, downside cases or probabilities from thin evidence.
4. Draft the executive memo last
Use Executive reforecast memo and decisions after the bridge and narrative are checked. Paste both, along with the meeting date and reporting period. The memo puts the decision request after the headline position, bridge summary and risks, so executives can see the supporting evidence before they are asked to act.
A decision request must be specific. “Discuss cost control” is not a decision. “Approve a pause on the named recruitment requisitions until the next forecast review” is a decision request, provided the source material supports it. Where evidence does not support a decision, the prompt instructs the memo to say so.
Note
Keep decisions separate from forecast commentary
A forecast can report an issue without implying that a particular response has been approved or is appropriate.
5. Run a release check
Finish with Board pack bridge quality check. It compares the bridge, memo and source-control record. Use its severity column to decide what must be corrected before circulation. A critical issue is normally a non-reconciling total, inconsistent period, unsupported headline claim or missing cash explanation.
How to tell the output is wrong
Check the model output against the underlying files, not against whether it sounds coherent. Pay particular attention to these failure signs:
| If you see this | Check this | Do this next |
|---|---|---|
| A variance has the right direction but wrong size | Formula, unit, period and sign | Recalculate from the named source fields |
| Cash moves without a driver | Working capital, capex and financing lines | Add the missing cash-flow source or mark it unresolved |
| A narrative claims a cause | Evidence column and source wording | Downgrade it to an assumption if causation is not evidenced |
| A decision has no owner or deadline | Decision table | Return it to the meeting sponsor for completion |
For version-dependent product behaviour or input handling, check the xAI documentation overview before using sensitive or large reporting material.
Stop
Do not circulate a polished memo with unresolved material residuals
A clear caveat is better than a false reconciliation in an executive pack.
When it does not work
If the bridge is incomplete, return to the source-control output rather than repeatedly rewriting the memo. Supply the missing period, scope, driver definition or source field. If the bridge reconciles but the narrative is weak, rerun the driver review with the relevant operating commentary. If the final quality check says “Do not release”, correct the named source issue, rebuild the affected bridge and rerun the release check.