AI Is Great But Stupid [Part 4]: Practical Tips That Actually Help
At this point, I’m sure that you’ve all heard the standard advice: “work on your prompting,” “don’t trust AI blindly,” “just be more specific.”
Short on time? You can listen to this audio summary in under 3 minutes.
Length: 2 min 39 sec
Read Full Transcript
Hi, I'm Allyson Edwards from AdviseUp Consulting, and this is The Bottom Line.
Over the last three episodes, we've talked about what AI actually is, where it fails, and how to govern it. To wrap up our series, we're shifting from high-level oversight to the daily habits that make your team more effective users, because getting value from AI has less to do with finding a "perfect prompt" and much more to do with building realistic habits.
The most useful adjustments are small, practical changes to how you interact with the technology.
First, stop treating AI like Google. According to McKinsey, half of consumers use AI as a search tool. But remember, search engines retrieve information, while AI generates responses. When you give it a generic, search-style prompt like "write an audit plan," you get a generic, high-level output. In practice, AI works best when you treat it like an intern or a junior collaborator, not a database. Give it context, explain what you’re trying to achieve, define your audience, and lay out the limitations. It turns out that many of the same communication habits that make a good manager also make a good AI user.
Second, ask AI to critique itself. Don't just take the first output and run with it. After a response is generated, ask the model: What assumptions are present here? What risks are missing? What argument could be made against this conclusion? These follow-up prompts often provide more insight than the original request, and forcing the AI to walk you through its reasoning makes it much easier to catch inconsistent or overly general logic.
Third, remember that sometimes the best prompt is a constraint. We are used to thinking that more detail automatically means a better answer. With AI, sometimes the opposite is true. If you want clarity, limit your prompts. Ask for just three examples, tell it to avoid technical jargon, or specify that it should identify gaps rather than filling them in. This keeps the model from overproducing or overgeneralizing.
Ultimately, the most useful AI skill isn't technical at all. It’s good judgment. It's knowing when an output feels incomplete, when nuance is missing, or when something is technically right but still just off.
Many of the exact same skills that define strong audit and compliance professionals also define strong AI users. The technology isn't a magic bullet, but it isn't useless either. The key is simply learning where it genuinely helps, where it needs oversight, and how to ask better questions.
Thank you so much for tuning into our series on The Bottom Line. For more insights on risk management, governance, and emerging technology, visit us at AdviseUp Consulting. Have a great day.
None of these suggestions are wrong, but they also don’t get to the heart of what makes AI useful in day-to-day work. That has less to do with finding the perfect prompting format and more to do with understanding how modern AI behaves in practice.
The most useful AI habits for most users are small adjustments that improve consistency, reduce errors, and make the AI’s output more useful for your actual use cases.
TIP #1
Stop Treating AI Like Google
According to McKinsey, half of consumers intentionally seek out AI-powered search tools, both through AI applications themselves and through things like Google’s AI Overview. While this might be tempting, it’s important to realize that AI is not a search engine. Search engines retrieve information; AI generates responses. This distinction matters because many weak AI responses start from search-style prompts: “summarize HIPAA,” “write an audit plan,” “what are the risks.”
The outputs may look polished, but they’re often high-level and generic because the request itself is generic, and better suited to traditional searches which will lead you to more detailed articles. In practice, AI tends to work better when you treat it more like an intern or a junior collaborator than a database:
- Give it context.
- Explain what you’re actually trying to achieve.
- Define who your audience is.
- Mention any constraints or limitations.
- And, if possible, describe what a “good” output actually looks like.
It turns out that many of the same communication habits that make a good manager also make a good AI user.
TIP #2
Ask AI to Critique Itself
Many people use AI to generate an initial output, whether that means to draft a policy, to summarize legislation, or to rewrite an email to get the tone you want. However, the key to getting the most out of those initial responses is to ask AI to critique its output. After you generate your initial response, try asking:
- What assumptions are present in this response?
- What important risks might be missing?
- What argument could you make against this?
These follow-up prompts will often provide just as much, if not more, insight than the original request.
In a similar vein, another useful technique is asking AI to explain why it reached a conclusion. For example, asking it to walk you through its reasoning and what information influenced its recommendations makes it easier to notice where the logic is inconsistent, unsupported, or overly general.
TIP #3
Ask Follow Up Questions
The first AI response doesn’t have to be your final answer, and your first prompt doesn’t have to take into account every single bit of nuance. AI tends to work better as an ongoing “conversation” rather than a single request.
Use the first response to establish direction. From there, you can refine the outputs gradually by narrowing the scope, adjusting the tone, clarifying priorities, and correcting any misunderstandings that the AI has made along the way.
This helps minimize some of the quiet failures of AI: overly vague recommendations, incomplete context, and plausible but actually unhelpful outputs.
TIP #4
Sometimes the Best Prompt Is a Constraint
Most of us are used to thinking that more detail automatically means a better answer. With AI, sometimes the opposite works better. If you’re trying to get to the true heart of something, looking for more will just lead to a lack of clarity; instead, try limiting your prompts: ask for just three examples, ask it to avoid technical jargon, or specify that it should identify gaps rather than filling them in. These constraints reduce the tendency of AI to overgeneralize or overproduce.
TIP #5
The Most Useful AI Skill Is Good Judgment
A lot of effective AI usage really just comes down to judgment: knowing when an output feels incomplete, when nuance is missing, or when something is technically right but still just feels off. None of this requires that you’re especially technical. You just need to ask thoughtful questions, communicate clearly, recognize poor outputs, and validate information carefully.
In that respect, many of the same skills that define strong audit and compliance professionals also define strong AI users.
Final Thoughts
AI is not a magic bullet (yet), but it’s also not useless. Sometimes it’s incredibly useful. Sometimes it’s confidently incorrect in completely baffling ways. And sometimes it’s strangely good at things in a way that makes it sound almost human.
The key is learning where AI is genuinely helpful, where it needs oversight, and how to build realistic habits around using it.
Ultimately, getting value from AI has less to do with finding the “perfect prompt” and more to do with learning how, and when, to ask better questions.
Build Better AI Habits
Predictable risks require thoughtful oversight. Contact AdviseUp to discuss strategies for building realistic daily habits, refining team communication protocols, and implementing practical validation workflows across your organization.
"AI is Great But Stupid" Series
Understanding what AI is lays the foundation for everything that follows: how it fails, how it should be governed, and how it can be used This series is designed to move professionals from AI-hype to AI-competence.
Part 1: Understanding What AI Actually Is
Part 2:
Where AI Goes Wrong
Part 3:
Governing AI
Part 4:
Practical Tips That Actually Help
