Many AI startups are also focusing on prompt engineering to minimize the number of tokens consumed while interacting with foreign AI models. By designing prompts more efficiently and reusing existing responses wherever possible, companies can contain expenses.