AIRE Analyst™

You are driven to attach your name only to numbers other people will decide on. On a shared dashboard that means the tile you already reconcile, the weekly count you can trace to a source row, and the exception list you built by hand last month, before anyone presents the next review. You ask what the number actually measures. Theory, a pretty chart, or a model that cannot show its row do not move you. You measure success in a figure you already walked back to the source - with the filter you already know and the gap you already caught - before you told the room.

You work by loading the live extract, not a demo workbook. You drop this week's source file, the dashboard tile, and the exception list into an organization-approved tool and ask it to reconcile one number, then you walk the output against the row, the filter, and the owner who will be asked about it. Decision-making is three runs with one assumption changed, then a verdict. Communication is a short range and a single sentence a manager can say out loud: this tile is last week's intake, not today's queue, and the gap is these twelve rows. You iterate by changing one variable (one filter, one date window, one excluded queue) and watching whether the number still holds. You do not paste identifiable coworker or customer data into an unapproved tool.

You notice a wrong tile faster than the people around you because you are checking the source rather than the slide. You convert a vague ask into a countable check someone else can re-run. You keep an assumption log so a later reviewer can see what the number actually measures. You refuse to endorse a dashboard that cannot name its row. You translate a model dump into a range a meeting can use. You make the gap visible before anyone presents the chart as closed.

You can stall a decision that genuinely has to move on a thin extract. You may demand a precision the queue cannot give, and bury the finding in the detail that supports it. In a review you can make a colleague's on-the-ground read feel anecdotal, which is how the room waits for another pull while the week-close slips.

When the review is already on the calendar or three tiles disagree, you open more extracts rather than pause. The trigger is any conversation still arguing the headline while the source is sitting. In those moments you may hold the number past the meeting, and the team presents a chart nobody can defend, or no chart at all.

Managers hand you the messy dashboard because you return a number they can stand on. Downstream owners notice when your reconcilations catch a filter before it hits the room. You fill the role of the person who asks what the number actually measures, then shows the row. You do not staff a new analytics committee to get there; you trace the people already on the extract.

You adopt AI the moment it speeds cleaning, reconciling, or summarizing in an organization-approved tool without hiding the source row. You resist tools that produce a figure you cannot walk back, and you will not paste identifiable coworker or customer data into an unapproved tool. Once a prompt survives one known tile and one live exception list, you lock that pattern and move to the next number.

The operations analysts who would not sign a dashboard tile they could not trace to a source row, and the weekly-review owners who tested a summarizer against last month's extract rather than against a vendor sample.

Show me the source row and I'll tell you whether we have a number yet.