The case study prompt is a clever workaround, but I've found it just moves the problem. You get specific, invented numbers instead of vague claims, but the numbers are still hallucinations. Writesonic, in my tests, tends to make those numbers overly optimistic to fit its "hype" voice. A fictional startup's costs might be suspiciously low, which is arguably more dangerous than generic advice.
The real issue is that for cost content, any number without a clear, auditable source is a liability. You're now editing fake specifics instead of vague claims, which might actually take longer.
Measure twice, spend once
That "competent but bored MBA student" description is painfully accurate. You've hit on the core quality issue: structure without substance.
You can fix bad structure with a rewrite, but you can't fix a lack of critical analysis. When a tool lists "vendor lock-in" as a trade-off for a startup, it's missing the point entirely. For an early-stage SaaS, the bigger trade-off is often operational simplicity versus potential future migration cost. The tool can't weigh that because it doesn't understand business context.
Your brief was excellent. The fact that neither tool could meet it is the real answer to your budget question.
Keep it constructive.
Yep, exactly that. Neither one offered any real metrics or a framework for calculating them. They'd mention "per-millisecond pricing" and "potential savings," but the guide ended right where the actual work begins.
It's like getting a recipe that just says "cook the ingredients until done" without giving temperature or time.
For your specific AWS example, the output never got to the level of suggesting a sample workload, defining baseline utilization, or showing a spreadsheet structure. You're left with a paragraph about the *importance* of cost comparison. Frustrating when that's the whole assignment, right?
Ship fast. Learn faster.
That's a strong analogy. The "CI/CD pipeline without tests" is exactly what you're buying if the tool only handles structure. It creates the appearance of quality, but you're still on the hook for the real verification.
Your "real budget is time" point is the core of it. If you're already planning to source the substance from your SME, paying a subscription to generate an empty structural shell is a bad trade. You could build a reusable Notion template for that shell once and be done.
The question becomes whether the tool saves more time in editing than it costs. For concrete analysis, it often doesn't.
Keep it constructive.