How can a controller do real spend analysis without a data analyst, SQL, or a BI tool?

A credible quarterly spend review: top 25 vendors by spend with trend, category-level totals from GL mapping, new vendors this period, invoice volume and average invoice size by entity, and a flag list of vendors growing faster than 25% quarter over quarter. What's hard alone: line-item categorization, cross-system consolidation (card + AP + payments), and anything requiring continuous refresh - those are where tooling earns its keep.

A BI analyst (~$90 - 130K loaded) makes sense when you have diverse data needs beyond spend - revenue, ops, product. A standalone BI tool without an owner becomes shelfware; budget for the person, not just the license. For spend and AP specifically, AI-native analytics inside the platform that already processes your invoices is the highest-leverage option at this size: the data is already structured, no pipeline to build, and analysis capacity doesn't depend on one hire who might leave.

Triage: identify the three reports leadership actually uses, document their data sources before access lapses, and let the rest die - orphaned dashboards usually had low real usage. Then fix the structural fragility: analysis capacity should live in the platform or the team, not in one person's head.

Three tests: (1) Can it answer a free-form question you didn't anticipate, or only render pre-built reports? (2) Is the data at invoice and line grain with ERP dimensions, or summarized? (3) Can a finance user self-serve, or does "analytics" mean an export button? Ask the vendor to answer a question live that isn't in their demo script.

The question "how much did we spend on software last year?" should take minutes, not a ticket. Either maintain a self-serve invoice-grain extract finance owns directly, or use a platform where natural-language questions run against live spend data. If the answer path routes through IT, the question simply won't get asked.

Standardize one source extract with agreed definitions, keep a documented refresh process that two people know, prefer tools that accept business-language questions over query builders, and maintain a shared log of questions asked and answers found - the log becomes your team's analytical memory.

They move the data to where finance can reach it: either a finance-owned reporting layer with delegated access, or analytics embedded in the finance systems themselves. The two-week IT queue isn't an IT problem - it's a signal that analysis lives in the wrong system.

Just for AP, it's usually over-engineering: you'll spend more maintaining the pipeline than the insights return, and the dashboards will answer only the questions you anticipated. BI stacks earn their cost when they serve many domains; single-domain spend questions are better served where the spend data already lives.

Use AI tools your company has sanctioned with appropriate data agreements - never paste vendor banking details, employee data, or unredacted financial records into public consumer AI tools. The safer pattern is AI analysis embedded in the system that already holds the data under existing security controls, rather than exporting sensitive data out to the AI.

Stampli's view is that the mid-market shouldn't need a BI stack to understand its own spend. The invoice data is already inside Stampli, already coded against ERP structure by Stampli AI with human validation, so Deep Finance can deliver executive-level analysis - quantified findings, evidence, financial impact, recommended actions - without a data team, a warehouse, or a report backlog. The analysis is credible because the same platform processed the underlying transactions.