Session
Label Me If You Can: What Document Classification Really Costs
A customer asked me to classify a document library that had been growing for fifteen years. Contracts, invoices, scans, and a folder called “final_final_v3”. We built the prototype with LLM-based classification. It worked beautifully. Then I extrapolated the token cost across the whole library, put the number on the screen, and the meeting was over ten minutes early.
This is that project, told honestly. Not the polished version. Classification is the boring layer nobody puts in a keynote, and it is exactly where the money goes once you want Copilot or an agent to work on your own content. The tooling question got harder recently, because Microsoft folded Azure Document Intelligence into Content Understanding in Microsoft Foundry. Deterministic extraction and LLM analyzers now sit in the same house with very different price tags, and the docs won’t tell you which one you actually need.
I’ll show you what we built, what we threw away, and the numbers behind both. Where per-page pricing beats an LLM by an order of magnitude. Where only an LLM does the job. Which questions about security and data residency you need answered before the first API call. And how fast the bill grows when you get it wrong. If you’re a consultant, architect or admin who is about to touch real customer content, this session saves you one very uncomfortable meeting.
Ferdi Lethen-Oellers
Microsoft MVP |Increasing productivity in everyday work with Microsoft 365 | Modern Workplace enthusiast and expert in the Microsoft 365 environment #coffeefreak #gernperDu
Düsseldorf, Germany
Links
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top