35% of 10,000=3,500 (mean1). That's Sample 1 where n1=10000. 65% (6,500) are still skeptical.
950 (mean2) out of 3000 is 950/3000=19/60 (31.67%) are skeptical (do not accept the new software). That's Sample 2 where n2=3000. In this sample, 41/60 (68.33%) accept the new software.
Both of these are large samples so the Normal Distribution can be assumed.
This is a binary situation and variance=np(1-p) where p is the proportion (%ge).
variance1=σ12=10000×0.35×0.65=2275; SD1=σ1=47.70;
variance2=σ22=3000×41/60×19/60=649.17; SD2=σ2=25.48.
We need to compare like with like, so let's consider the difference in mean proportions who are willing to accept the new software until an upgrade becomes available.
0.6833-0.35=0.3333=⅓.
Now we need a SD σ to generate a test statistic. σ=√(σ12/n1+σ22/n2)=√(2275/10000+649.17/3000)=1.58 approx.
The test statistic Z=0.33/1.58=0.21 approx. This is the number of SDs deviation from the mean difference of the means of the two samples.
At 5% significance Z=1.96 so 0.21 is considerably less than 1.96 so lies within the range of what is probably (with 95% degree of confidence) not significant.
The company can conclude that the skeptical population has not decreased significantly.