I am currently working on the digitization of German city directories (Adressbücher) using simple, but highly optimized Optical Character Recognition (OCR) to speed this process up. See Albers and Kappner (2022) for a discussion. 

Soon, I am going to digitize the entire city directory for the largest German cities (50-100 books per city), for which I currently need about 1 week. Currently, the character error rate achieved for Fraktur is approx. 1 error per 3 pages (approx. 0.025% CER), or about 0.01% with some post-processing. This is way below any AI or VLM. I expect further, tailored post-processing algorithms can bring it down to the ballpark of 0.001%. This was done without any AI, and I am planning to not use any AI/LLM in the future. There are no additional variable cost for tokens (except for my time).

City directories provide us with personal information (name, occupation, exact address, telephone number, floor number for a few city directories) about the household head, as well as business/amenity locations within the city over 200 years at a nearly annual frequency. Occupations can be mapped to income via HISCAM scores (Lambert et al (2013)). 

You can find the excel file for Düsseldorf 1891, which apart from the current OCR/clean-up pipeline, took about 30 minutes of manual corrections of flagged errors here (0.9mb, done in May 2025) not counting here the 20 minutes of OCR that can be done in the background. This is lower than some reliable, not overexaggerated estimates for manual entry of 0.13% of non-Fraktur sources (see the great work/slides by Lin, Moulton, Rand and Smith page 48). However, note that the error rates for Fraktur typeset is about a magnitude higher due to common issues by OCR to distinguish ligatures (u vs n, f vs s, c vs e) which can be partially corrected via post-processing.