One of the challenges is to pack all those satellites in launch vehicle's payload adapter. The payload adapter on launch vehicle needs to be flexible enough to allow such varying number of payloads. In this mission, ISRO had to realize a new/modified payload adapter to carry those 104 satellites.

Next is to accurately inject all those satellites - either into their respective orbits if it's a multi-orbit mission, or with sufficient time gap and re-orientation between injections if it's a single-orbit mission.

This was a single orbit mission (all sats released into SSO). So, PSLV's 4th stage had to time those 104 satellite injections. It also need to re-orient itself for each injection depending on satellite's requirement. This mission had as many as 103 "separation events". All this logic is pre-programmed and executed by launch vehicle's onboard computer.

Check out QuadPacks lined up before the launch: http://blog.isilaunch.com/the-quadpack-line-up/

And, check out those QuadPacks mounted on PSLV payload adapter around primary payload: https://imgur.com/a/LM9Cp

The hardest part is in ensuring that the satellites are launched in different directions to avoid them colliding when being deployed in orbit. These satellites have to be in the same orbit but with increasing distance between them.

From what I understand, these satellites were injected into the heliosynchronous orbit (https://en.wikipedia.org/wiki/Sun-synchronous_orbit) at different slots, different angles and at different times.

However, the sole idea behind deploying 104 satellites is not to create a record of some sort but to utilize the maximum capacity of the workhorse rocket (PSLV). This results in greater ROI for the Space Agencies (Antrix / ISRO) as most of these 104 satellites (101 I think) were foreign satellites.

According to this article: http://tech.firstpost.com/news-analysis/isro-to-recover-half... ISRO will recover half of the total cost incurred for the launch from foreign satellites mounted on the PSLV.