I Was Already Using Tools

Compliance people aren’t strangers to tools. We just use different ones.

In fraud analysis, the data is usually messy — tens of thousands of expense records, inconsistent fields, formats all over the place. My workflow went like this: pull the data out of the system with SQL, clean it up in Tableau Prep — merging tables, handling missing values, standardising field names — then throw it into Tableau for visual analysis.

Fraud hides in patterns:

  • A vendor registering and invoicing in quick succession
  • Expense claims clustering around times when the approver was on leave
  • Amounts sitting just below the approval threshold
  • Claimed locations that don’t match travel records
  • Sudden spikes in one person’s reimbursement frequency

Scatter plots, timelines, heat maps — Tableau turns these patterns into something you can actually see. Power BI does similar things; I just preferred Tableau’s interaction model.

So to be accurate, I didn’t start from zero with data tools. The tools evolved, and I followed.

What Changed When AI Came In

The shift was gradual. It wasn’t a single moment of deciding “I’m going to use AI now” — it was one concrete problem after another pushing me forward.

Claude is the one that stuck around longest. I don’t use it to write compliance policies or audit reports — those demand too much context for AI to get right. What I use it for is more specific: debugging code when something breaks, stress-testing the logic of a new compliance review process, translating cross-border regulatory documents. Treat it like a colleague, not an authority — it makes mistakes, and you need to check.

Google Gemini I use too. Particularly good with long documents — a fifty-page regulatory update or an industry report, drop it in and have it pull out the key points or compare changes between versions. Different tools are good at different things. No reason to lock yourself into one.

NotebookLM was an unexpected find. I use it to organise study materials — throw in a stack of documents, notes, PDFs, and it helps you see the structure and connections. The most interesting part is that it can turn your materials into a podcast format. Listen during the commute instead of staring at documents. In compliance, you’re constantly digesting regulatory updates and industry reports. Having a tool that converts all that into something you can listen to on the go — the time it saves is real.

The Toolkit Wasn’t Planned

Looking back, the evolution wasn’t strategic at all:

SQL + Tableau Prep → Tableau / Power BI → Claude / Gemini / NotebookLM → Python scripts for automation

Each step happened not because a technology was trending, but because the tools I had couldn’t solve what was in front of me. Tableau solved visual analysis. AI tools solved document processing and thinking efficiency. Python solved the repetitive batch work.

If you’re in compliance too, my suggestion: don’t start from the tool, start from whatever annoys you most. Find the thing that eats your time every week, then find the tool that kills it. The order doesn’t matter. Solving the problem does.

Detours

Trying to build everything at once. Initially wanted a “fully automated compliance system.” Finished nothing. Learned to start from the smallest problem. One tool, one fix.

Collecting tools. Tried thirty-plus. Kept fewer than five.

Going it alone too long. Not many people in compliance are using these tools yet. Part of why I started this blog — writing it down in case someone finds it useful.

First post on this blog. More to come.