I've worked with some associates on automated pokerbots for a couple years. It is actually extremely, extremely difficult to make one that can be successful long term.
Number of problems you will run into:
1) Gameplay changes on every variation of the different limits, as well as the table sizes.
2) You need to track, log, and record the past history of the players you are currently sitting at the table with. Then you need to compute this data in a way the bot understands so it doesn't fold to crazy bluffers, or those who only bet with good hands.
3) You need to add variables to your gameplay, ones that win. It's extremely easy to catch onto a "if I do this, this person does this" action, you can't fall victim to that.
4) Sitting at the table too long, and the money dries up, and people change their gameplay as well as further understand your gameplay. Thus your bot needs to know when to leave, and what tables to join.
5) Testing. You need to run about 15,000+ hands before you can accurately analyze the way your bot will play long term, and can only then make the appropriate adjustments. It's both costly and time consuming to do this over and over and over.
Of course there are tons of other problems you will run into, but it is far from an easy task.
