The main reason, assuming you hooked up everything correctly and all the equipment was working correctly, is the difference in input impedance between the DMM and your A/D circuit.
The input impedance of the DMM should be clearly listed in its manual. It may even be written write on the unit. A common value is 10 MΩ.
The input impedance of the voltage divider is the sum of the two resistances. The input impedance of a typical CMOS A/D is probably much higher than your divider resistances, so you can ignore its contribution.
Unless you have a very unusual A/D, the sum of your two divider resistors is much less than the input impedance of the DMM. Therefore consider what the voltage source you were measuring does with a 10 MΩ load, and a few kΩ load that your A/D circuit presented.
Of course the dividers add some error too. Look up the accuracy of the resistors and compute the overall accuracy of the divider. Most likely, the accuracy of the DMM is higher than even your bare A/D. Even without circuit loading issues, the DMM is likely significantly more accurate than your A/D circuit. Then the A/D has quantization error too. If it's a 10 bit A/D, for example, then you get about ±0.05% error just from quantization. Likely the resistors swamp that error, though.
