Randomness and confidence intervals

I've created a sampling function that contains within it a feature for outputting the Min and Max of a Confidence Interval:

minCI1 = (ax.mean-3.169*(ax.deviation/(math.sqrt(len(a)))))
maxCI1 = (ax.mean+3.169*(ax.deviation/(math.sqrt(len(a)))))

The variable 'ax' is just a weighted average of the input list.

My first thought to create more randomness into the interval would be to use a lower T-value than 3.169, which corresponds to a 99% confidence interval. The effect of this makes my model less accurate, so I'd like to use some randomness feature that I can use to control the degree of randomness from the mean, as well as the frequency with which it occurs, like so:

def randomizer(tvalue,random_frequency_percent,minCI)

Firstly, would this be any better than using a different T-value, and secondly, does this make statistical sense or is there a better way to go about it?