ChatGPT itself is skeptical of this and thinks the whole AI agent thing is overhyped. Here is a lengthy response from it when I asked about making an AI agent to handle property management tenant collections: At its core, integrating AI like OpenAI into an app often feels like just another layer of programming. The idea of “prompt engineering” is really just programming, except instead of writing traditional if-else logic, you’re crafting structured text inputs for the AI. This can make it feel like AI is overhyped because the marketing often downplays the amount of work required to integrate it effectively. Here’s why it feels this way and when it is worth the effort: Why AI Feels Overhyped 1. Prompts Require Logic Crafting prompts isn’t magic; it’s logic phrased as natural language. For example: - Traditional programming: if overdue_count > 3 then escalate() - AI prompt: “The tenant has been late 3 times. Should we escalate?” You’re still solving the same problem, but now you’re phrasing it differently. It doesn’t replace programming—it just shifts the complexity. 2. You Still Need Context AI doesn’t “know” your business. You have to explicitly tell it everything relevant in each prompt. This means: Querying your database. Formatting that data. Passing it into the prompt. This step often feels more tedious than just coding a solution directly. 3. Cost and Latency Calling an AI API introduces cost (OpenAI charges per token) and latency (waiting for the response). For simple problems, writing plain logic is faster and cheaper. 4. Limited Autonomy AI doesn’t “do” anything on its own—it just gives suggestions or outputs. Your system still needs to act on those outputs, which means building middleware and execution logic. When AI Is Worth It AI starts to make sense when: 1. Tasks Are Dynamic or Complex: For example, drafting personalized tenant messages where tone and content vary based on late payment history, tenant profile, and overdue amount: - Coding this logic in traditional programming would require dozens of conditions. - An AI can handle it with a well-crafted prompt. 2. High Variability: In cases where rules or outputs change frequently, AI can reduce maintenance. Instead of updating hardcoded rules, you tweak the prompt. 3. Language Tasks: Tasks like generating human-like emails, summarizing conversations, or translating legal documents are areas where AI excels. 4. Unstructured Input: If the input isn’t clean or predictable (e.g., analyzing free-text tenant complaints), AI can make sense of it faster than traditional programming. When AI is Not Worth It 1. Straightforward Logic If the problem can be solved with basic conditionals, loops, or database queries, traditional programming is usually faster and cheaper. 2. High Accuracy Needed AI outputs are probabilistic, not deterministic. If your use case can’t tolerate ambiguity (e.g., financial calculations or compliance rules), AI is riskier than plain logic. 3. Heavy Preprocessing If you spend more time preparing data and crafting prompts than actually using the AI, it’s a sign the task may be overcomplicated for AI. Is It All Hype? In many ways, yes—AI is overhyped. But it’s not that AI is useless; rather, it’s often misapplied. Businesses frequently try to use AI where simpler solutions would work just as well or better. The Programming Reality AI doesn’t eliminate programming; it shifts what you’re programming. Instead of: - Writing direct logic, - You’re writing logic to gather context, build prompts, and handle responses. In many cases, it doesn’t save much time or effort, especially if the problem is simple. Where AI shines is: - Complex, language-based tasks. - Flexibility where requirements change often. - Tasks requiring human-like nuance.