I would be careful with low/high pass filters. The human voice is 'intelligible' at around 1.5K to 5K, in most cases. According to this paper (http://www.kyu.edu.tw/93/epaperv7/089.pdf) the range of an ariconditioner is hits most of that range. A low pass filter would cut out the lower ranges of the office noise, but you will still have the higher ranges, which seem to be rated at the highest dB on the chart.
I found the best option, it has been awhile, was to use a separate microphone, I liked Louroe hardware (www.louroe.com) and adjust the dB rating that activates the mic. Depending on the VMS, you may have some adjustment levels there as well.
It may be best to use a mix of options, adjust dB threshold along with an audio filter that allows for some level of control, like adjusting specific frequencies. For example, something that allows you to increase 2-5k range, while lowering everything else. If I recall, this is called, 'pocketing' in the audio world, creating space for the frequencies you want.
Good luck.