Four demands. Capacity for two.

What this choice buys

Planning runs on one margin with an accountable owner in the business and a written bridge to the view Sales reports, so a difference between the two reads as a definition rather than an error. Each pipeline behind the morning report gets a deadline, a notification when it is missed, a named responder, and a fallback agreed in advance: the meeting is told the number is from yesterday, and told by whom.

When it holds

It holds when annual planning starts this quarter, so two margins would otherwise reach it, and when the late report feeds decisions where a stale number has already cost something. If planning is a year off, the margin work is early.

What you accept

A staffing view, deferred

The roster is set from the Monday extract for another quarter, on the previous week's numbers when the analyst is out. Operations carries that: a backup for the extract is named, the Mondays it ran on old numbers are counted, and the count is reviewed with the head of operations before the next quarter is planned.

Evidence on the AI assistant, deferred

The executive team goes another quarter with a demo as its only evidence, and the question returns every month. The sponsor is asked to hold the rollout decision until an evaluation is funded, and the answer is written down. If the decision will be made regardless, that is a cost of this pair, not a risk.

The chosen work, once delivered

Alert thresholds take a quarter of tuning before the responders trust them, and someone answers every one; the planning margin needs a steward when the next product line or channel arrives.

What changes the choice

Replace the margin work with the staffing view if planning does not start until next year. A roster set on stale evidence costs something every Monday, and a plan a year away can wait a quarter for its definition.

What this choice buys

Planning runs on one agreed margin with an owner and a bridge to the other view. Every Monday the roster is set with the demand forecast beside capacity while the decision is still open, and the view shows when its numbers were loaded, so a Friday number is at least labeled as one.

When it holds

It holds when the staffing view reads from a source that lands before Monday morning, not from the pipeline that has run late, and when the head of operations agrees that a labeled stale number beats no view. If the source is the late one, the view ships the problem with better typography.

What you accept

Freshness alerts, deferred

The analyst checking timestamps before eight stays the alerting system for about sixty mornings and will miss some of them. The managers who run the meeting carry that: an owner and a backup for the check are named, the check now covers the staffing source too, and the mornings it was missed are reviewed after a month. When the report is late, the fallback is still to say so in the room.

Evidence on the AI assistant, deferred

Another quarter with no evidence on the assistant. The sponsor is asked whether rollout waits for the evaluation, and if it does not, the decision is made on a demo. This pair accepted that in advance rather than discovering it.

The chosen work, once delivered

The view needs adoption as much as maintenance: a quarter of Mondays beside the spreadsheet before the spreadsheet goes, and the planning margin needs a steward from the day it is agreed.

What changes the choice

Replace the staffing view with the freshness alerts if the view would read from the pipeline that has run late three times. Evidence that arrives after the roster is set is evidence for next week.

What this choice buys

Planning runs on one agreed margin with an owner. The executive team gets an evidenced recommendation on the assistant: a question set drawn from what people actually ask, the data it may see, business reviewers for its answers, and a finding to proceed, narrow or stop. Rollout stays a separate decision the sponsor owns. Margin is in the question set, so how the assistant handles a disputed definition is reported rather than discovered in use.

When it holds

It holds when the executive team will decide on the assistant this year with or without evidence, and when reviewers in the business can give the evaluation a few hours a week. A recommendation to stop is a useful result; this pair is not a bet that the assistant ships.

What you accept

A staffing view, deferred

The roster stays on the Monday extract, set on the previous week's numbers when the analyst is out. Operations carries it with a named backup and a count of those Mondays, reviewed at quarter end. The head of operations watches two executive requests get funded and theirs wait, and is owed that reason in person rather than in a roadmap.

Freshness alerts, deferred

The morning check stays manual, with an owner, a backup and a review of missed mornings after a month. The evaluation includes the case where a table has not landed, so the assistant is scored on whether it says so, and a correct refusal scores as a correct answer.

The chosen work, once delivered

An evaluation set is not finished when the quarter is: every change to a definition or the model reruns it, and someone who knows the domain reads the failures.

What changes the choice

Replace the evaluation with the staffing view if the sponsor agrees to hold the rollout decision until next quarter. The roster is set every Monday whether or not the evidence arrives; a held decision costs a quarter of waiting.

What this choice buys

The roster is set every Monday with the forecast beside capacity, and the view is covered by the alerts: a deadline on its source, a notification when it is missed, a named responder, and a fallback the head of operations agreed to before the first late Monday. The morning meeting gets the same: told before the report is opened, and told what to use instead.

When it holds

It holds when annual planning does not start this quarter, or when the CFO has agreed to carry the reconciliation through it with a stated limit on how far the two numbers may diverge. It also holds when the weekly roster is where the money moves, because a view that says when it cannot be trusted is the version people keep using.

What you accept

One margin for planning, deferred

Finance keeps reconciling two margins by hand, two days a month, and both reports keep their footnotes. If planning starts this quarter, it runs on a footnoted number, and the plan is revisited by everyone who used it once the definition is agreed. The CFO carries that knowingly: the limit is written down, and each month's reconciliation is checked against it.

Evidence on the AI assistant, deferred

No evidence on the assistant for another quarter. The sponsor is asked to hold rollout until the evaluation is funded, the monthly question keeps coming, and this pair has no answer to it beyond a date.

The chosen work, once delivered

Alerts take a quarter of tuning before the responders trust them, and the view takes a quarter of Mondays beside the spreadsheet before the spreadsheet goes.

What changes the choice

Replace the alerts with the margin work if planning starts this quarter and the two numbers diverge past the limit Finance can carry. A plan committed on a number later corrected is harder to undo than a late morning report, and the check before eight can stay a person for one more quarter.

What this choice buys

The morning meeting is told when a pipeline is late, by whom, and what to use instead. The executive team gets an evidenced recommendation on the assistant, and the evaluation includes late and unsupported data as cases: whether the assistant says so when a table has not landed, and what it does with a question whose definition is disputed.

When it holds

It holds when the executive team will decide on the assistant this year with or without evidence, and when the margin question can be handled inside the evaluation: scoped out of the first question set, or kept in so the finding records that the assistant cannot answer it until the business agrees. A recommendation to narrow is a result, not a failure.

What you accept

One margin for planning, deferred

Finance keeps reconciling by hand, two days a month, and planning runs on two numbers if it starts this quarter, within a limit the CFO has written down. The evaluation surfaces margin early, because it is the question people most want to ask. The honest finding is that the assistant should not answer it yet, and it is said before the first demo rather than after.

A staffing view, deferred

The roster stays on the Monday extract with a named backup, and the Mondays it ran on old numbers are counted and reviewed at quarter end. The one person who asked for something they would use every week gets nothing new this quarter. The alerts do cover their source, so the extract is at least known to be late when it is.

The chosen work, once delivered

Alert thresholds need tuning, and an evaluation set is rerun after every change to the definitions or the model, which is someone's standing work.

What changes the choice

Replace the alerts with the margin work if planning starts this quarter and the evaluation's first question set is mostly margin and profitability questions. The planning definition is then what both other items are waiting on, and detection can stay a person one more quarter.

What this choice buys

The roster is set every Monday with the forecast beside capacity, and the view shows when its numbers were loaded. The executive team gets an evidenced recommendation on the assistant, with the disputed margin recorded as a finding rather than discovered in use, and rollout left as the sponsor's separate decision.

When it holds

It holds when the staffing view reads from a source that lands before Monday and not from the pipeline that has run late, and when the evaluation can reach a useful finding without the margin definition. Both assumptions are about sources, and both can be checked before the quarter is committed.

What you accept

One margin for planning, deferred

Finance keeps the monthly reconciliation and the footnotes, and if planning starts this quarter it runs on two numbers within a limit the CFO has written down. The staffing view never touches margin. The evaluation does, and its finding is that the assistant cannot answer margin questions until the business agrees on one, which narrows what the executive team imagined.

Freshness alerts, deferred

The morning check stays manual, with an owner, a backup and a review of missed mornings after a month, and the check now covers the staffing source. The view shows its load time; what nobody can do is tell the meeting before it opens that the time is wrong.

The chosen work, once delivered

A view takes a quarter of Mondays beside the spreadsheet before the spreadsheet goes, and an evaluation set is rerun after every change by someone who knows the domain.

What changes the choice

Replace the staffing view with the freshness alerts if its source turns out to be the pipeline that has run late. Replace the evaluation with the margin work if planning starts this quarter and the sponsor agrees to hold the rollout decision. Either check fits in the first two weeks.