AA/CEP doesn't have a native 24-bit workspace like WaveLab or Audacity. AA/CEP has two workspaces: 16-bit and 32-bit. In both workspaces, AA/CEP performs all edits at 32-bit. In the 16-bit workspace, AA/CEP performs the edits at 32-bit and then dithers back to 16-bit as part of the edit operation. In the 32-bit workspace, it simply performs the edits at 32-bit and obviously does not need to dither as part of the edit operation.
Basically, if you <1> fed AA/CEP 24-bit data (which it sounds like you did), <2> did not perform any edits (dunno the answer to this one), and <3> saved the file as 32-bit (which seems the case given the filesize), AA/CEP will have padded the last 8 bits with zeroes. Simply open the file up, and Save As, press the Options button, select 24-bit packed int (type 1, 24-bit), and untick the Enable Dithering box. This should save your file as 24-bit by simply truncating the least significant 8-bits that AA/CEP padded with zeroes.
If you performed edits, meaning the file has actual 32-bit data (as opposed to 24-bit + padded zeroes), you'll first want to dither down to 24-bit, and then Save As per the instructions above.
Edit to add: In the future, two options for saving files after transferring:
<01> Save As in 24-bit format, per the instructions above. Upside is you don't have to go back later and truncate / Save As. Downside is if you re-open in AA/CEP, it'll load the file slowly since AA/CEP doesn't build peak files for 24-bit files.
<02> Save As 32-bit and then truncate to 24-bit (as above) before tracking. Upside is if you expect to open the file in AA/CEP again, it'll open swiftly since AA/CEP builds peak files for 16- and 32-bit files, but not 24-bit files. Downside is you need to truncate / Save As 24-bit before tracking.