Paul Barnhurst explains why ChatGPT and Claude can analyze financial statements at a high level but need direct access to transaction data to answer real questions. Without database connectors and schema awareness, models can't drill into general ledger records or trace what actually caused a 60% vendor spend increase.
We figured out ChatGPT Claude Gemini are pretty good at if you load in some financial statements it can look at the trends it can do ratios for you at the financial statement level it was giving you some good analysis but we didn't have the context and the ability to drill down and you didn't have the context window that's big enough. You're not going to put 10,000 GL transactions a month in there or whatever. So being able to tie directly in especially with the MCP connector being able to sort of have that schema to go across your database and all your tables know where to get the source information then it can tell you your cogs are off. But here, here's what drove it and it can look not just at this month but at historical and say hey you increased your spend with this one vendor by 60% this month. That's the big NQ in those things and that's where it becomes a valuable partner.
Trust and Leadership Lessons for FP&A From 100+ Top Leaders with CFO Jon Laudie
