Introduction
Maren Lucht wanted to decipher one word in her great-grandfather’s letter. The ink had bled, the handwriting leaned sharply, and the town name looked as if someone had folded it into itself. An AI tool suggested three options. All of them were wrong, but one pointed her to the right archive shelf.
Maren lives in Lübeck and helps local associations prepare old documents with AI. She remains a genealogist at heart: patient, suspicious of certainty and fond of pencil notes.
Story of the Path into AI
For decades Maren had researched family history with church books, shoeboxes and letters. She never worked in IT and distrusted cloud services at first. The promise of automatic handwriting recognition attracted and irritated her. Old documents are not just data; they hold family secrets, private grief and names that should not be thrown carelessly into a tool.
She learned text recognition, archive workflows and privacy basics. Her first project digitized family letters with uncertain words marked rather than silently guessed. The model did well with repeated phrases and failed on old place names. Maren compared every suggestion with original sources.
A local historian once corrected a name by remembering how a street used to be pronounced. Maren added “ask the living memory” to her checklist.
Current Work
Today Maren supports small heritage groups that want to start digital archive evenings. She helps them scan, transcribe and mark uncertain readings. In one ship register, an AI system misread names because the ink had run. Instead of accepting the output, Maren organized a reading workshop with older residents who knew local naming patterns.
The work does not produce perfect databases. It produces better starting points and fewer hidden guesses. Maren finds that honest uncertainty is a form of respect for the dead and for the living people connected to them.
Personal Advice
“Old documents need modern tools and old patience,” Maren says. Her advice is to keep the original close and never let a clean transcript erase the scratch, stain or doubt that belongs to the source.
Key Facts
Age and place: 67, Lübeck.
Background: homemaking, retirement, volunteer cultural work.
Entry into AI: digitized family letters with uncertain words marked for review.
Focus today: AI in local archives.
Typical tools: text recognition, archive databases, source criticism.
Werkstattnotiz
Maren’s archive box has slips marked “do not trust yet.” They slow the work down, which is their purpose. She is currently testing how local knowledge can be included without turning memories into unverified facts. The hardest cases are names everyone thinks they recognize.