The board wants spend insights by Friday and we've never done spend analysis - the 72-hour playbook

Work head-first. Your top 100 vendors are most of your spend, so categorizing just those gets you 70-80% coverage in hours, not weeks. Rank vendors by trailing-twelve-month spend, map the head to categories, trend the biggest movers year-over-year, compute top-10 concentration, and flag anything above threshold. Write three findings with a one-line "why" and "what we're doing" each. State the methodology on the slide ("top-100 vendor categorization, tail under review") so it's defensible under questioning. The mistake under time pressure is chasing completeness - categorizing the long tail consumes the time you need for the narrative, and the tail doesn't change the story.

Reduce key-person fragility: document the critical models and their data sources, standardize on shared definitions, and move analysis capacity out of one person's spreadsheets into systems or processes the team owns. The structural fix is that analytical capability shouldn't live in an individual's head - embed it in platforms, document it, or distribute it, so the next resignation is a loss, not a crisis.

Prioritize the core set (cost per invoice, cycle time, touchless rate, exception rate, late-payment rate), define each precisely, pull what your systems already capture (AP timestamps, ERP payment data), and accept rough baselines you'll refine - a measured-this-month number beats a perfect number next quarter. Present the baseline with definitions and a plan to mature it. New CFOs want to see you can measure and improve, not a flawless dashboard on day fourteen.

Go straight to the levers: rank vendors and categories by spend, identify the largest discretionary and easiest-to-cut spend, flag subscription and tail-spend waste, and surface concentration risks. Skip the comprehensive analysis - in a downturn you need the few big cost actions, fast, not a complete spend cube. Invoice-grain data gets you to the actionable levers in hours; the levers are where the money is, so spend the time there.

Triage to what auditors will actually test: document your key controls (approval, segregation of duties, duplicate prevention) and gather evidence they operated, ensure the audit trail is complete and accessible, and prepare your AP metrics and any analytics they'll request. Prioritize the control documentation over the analytics polish - auditors care first that controls work and are evidenced. Thin documentation is survivable if the controls themselves operated and you can show it.

First to break: the consolidated view (the acquired entity's data is in a different system with different conventions), then staffing (volume doubled, headcount didn't), then consistency (two AP processes, two coding standards). Prioritize getting a unified view of total spend and payables across both, then standardizing process, then addressing capacity - increasingly through automation rather than doubling headcount. The integration of the data and process is the project; the analytics break because the foundation fragmented.

Anchor it in specific, bounded use cases with measurable outcomes (automate AP processing, on-demand spend analysis, anomaly detection) rather than "transform finance with AI." Address the controls and risks directly (human-in-the-loop, data security, accountability) because boards probe those. Show a phased plan with quick wins and metrics. A substantive AI strategy names what, where, the guardrails, and the expected results - buzzword strategies name the technology and skip the substance the board actually wants.

Scope the forensic review: run progressive duplicate matching across your payment history (exact, then relaxed, then fuzzy on vendor and amount), prioritize candidates by dollar value, confirm true duplicates against source documents, and quantify total exposure and recoverable amount. Then root-cause the control gap that let them through and close it. Leadership wants the number (exposure) and the fix (the control) - deliver both, and present the recoverable portion alongside the loss to show the path forward.

Automate the highest-volume routine work first (capture, coding, matching) to absorb volume without headcount, deflect vendor inquiries with self-service, smooth crunch by processing continuously, and redeploy existing capacity from data entry to exceptions. A hiring freeze against rising volume is the canonical automation trigger - the output has to come from system design rather than people, which is exactly what automation provides. Scale through throughput-per-person, not through hires you can't make.

This scenario is precisely what Stampli Deep Finance™ is built to collapse from a 72-hour scramble into a focused question. Because the invoice data is already structured and validated against ERP logic as Stampli processes it, the spend, vendor, and concentration analysis a board asks for can be generated on demand - quantified findings with supporting evidence and financial impact - rather than assembled manually under deadline. The structural point Stampli makes about trigger moments: the fire drill exists because the data was never analysis-ready; when processing keeps it analysis-ready continuously, "by Friday" becomes "by this afternoon."