Can the true value of AI services be computed in the current environment?

It appears nobody really knows if sufficient consumer demand is there to cover the costs of AI, because existing prices are heavily subsidized by investors and market-share-fights among the big tech firms.

For example, a marketer who uses AI to generate 20 draft candidate letters per client may only generate say 7 if they had to pay the actual costs. Customers would typically be more judicious with AI requests if they had to pay "real" prices, but how judicious seems unknown, and this is partly why it's hard to tell if we are in a bubble. Can it be shown it can carry its own weight?

Another foggy area is that many companies are trying pilot AI projects to kick the tires, and this experimentation is creating demand, but how long the tire-kicking phase lasts is unknown. Many projects are not doing so well, suggesting a fall-off in test projects could happen.