Your best bet is to use experimental data. This cannot be done from automatic genome annotations in databases, mostly due to historical reasons and because bacterial transcription machinery works differently from the eukaryotic one. Unlike in eukaryotes, bacterial 3'-UTR's as a regulatory entity have only recently moved into focus and little is known about their regulatory role (reviewed by Menendez-Gil & Toledo-Arana).

Prokaryotic genome annotation tools mostly feature a simplistic gene model of "1 gene == 1 CDS," often without annotating transcripts and other features. Therefore, extracting UTR candidates from public annotation will not work out of the box. The review paper lists some methods and studies concerned with detecting bacterial UTRs.

If you only have automatic annotation data, your best bet may be to detect rho-dependent termination signals and investigate the region between CDS end the signal. The result may be noisy.

With AGAT.
agat_sp_extract_sequences.pl --gff infile.gff --fasta infile.fasta -t 3'-utr

or three_prime_utr it depends how the feature is defined in your file.

To focus only on the chosen gene ids, prior, you should filter your infile.gff with agat_sp_filter_feature_from_keep_list.pl