Simulation
Use this simulation to discuss the randomness of p-values and that:
The p-value cannot provide a definitive answer about whether µ1 = µ2
There is always some chance of being misled by an 'unlucky' sample.
Identical populations (H0)
Take several samples. Mention that in about 1/20 of the samples, the p-value will be less than 0.05, etc.
There is a (small) probability that the p-value will misleadingly suggest that the means are different.
Different means (HA)
Again take several random samples. Mention that the p-values are usuall (but not always) fairly close to zero.
There is again a small chance of getting no evidence of a difference between the means.