(This message was not created by AI. My English (words, sentence structure, grammar, punctuation) was improved by deepl, but the content is mine.)

There are many ways to improve this system. Adding a single letter, for example, shifts the problem slightly (from 676 possibilities to 17,576, to be precise, while ignoring several properties of real human names).
Using two letters from the first name and two from the last name (while falsely assuming that people have at least a first and last name, and that names have at least two letters) would result in 456,976 unique possibilities.

More robust schemes exist that do not make false assumptions about properties such as first and last names. For example, one could concatenate the full name, reduce the number of vowels, and if necessary, rank and drop letters by priority.
For some ideas, see:
A Study of Methods for Systematically Abbreviating English Words and Names [ https://dl.acm.org/doi/abs/10.1145/321088.321094 ] (1961).

Ultimately, the whole topic of names in IT is a minefield. See: Falsehoods Programmers Believe About Names – With Examples [ https://shinesolutions.com/2018/01/08/falsehoods-programmers-believe-about-names-with-examples ].

Außerdem arbeiten wir in einem Englischsprachigen Umfeld, dort sind Initialen typischerweise 3 -sellig.