The reason your AI responses are generic (and how to fix it)

If you've been told to "run it through AI" at work but nobody's shown you what that actually means, this one's for you.

Here's what usually happens. Someone tries it, gets a response that's way too generic to be useful, and quietly decides the tool isn't for them. They're not wrong about the response. They're wrong about why it happened.

The problem isn't the tool. It's that the tool is missing everything you already know. Your audience, your history, the nuance of the situation, the three things that always go wrong in your specific process. That context is sitting in your head, and the AI has none of it.

So here's the fix: before you ask for anything, ask for questions.

Let the AI interview you. Once it understands your situation, what it produces will actually fit.

Here's what that looks like in practice.

You need to explain a new tax rule to someone who wasn't in the room

Picture this: a new revenue recognition policy just dropped. Your controller asks you to brief the sales team on how it affects contract timing. The sales team does not speak accounting. If you get this wrong, you'll either lose them in jargon or oversimplify it to the point of being useless.

This is where most people paste in the rule and ask AI to "make it simple." What comes back reads like a textbook. It doesn't mention deals, or quotas, or the specific question the sales team is actually going to ask on Friday.

Try this instead:

"I need to explain a tax or accounting topic to someone at my company who isn't in finance. Before you write anything, ask me questions about who I'm explaining it to, what they already know, what decision or action they need to take, and what's likely to confuse them. Keep asking until you have enough to write something actually useful."

What you save: an hour of second-guessing your own explanation, plus the awkward follow-up when someone misunderstands and acts on it wrong.

What you gain: a version tailored to your actual audience that you can use in a meeting, an email, or a quick Teams message.

You need to document a process that lives entirely in your head

You've closed the books the same way for three years. You know every exception, every workaround, every thing that breaks in quarter three. Now your manager wants it documented before you go on leave, or an auditor needs to see the steps, or someone new is joining the team.

You sit down to write it and realize you don't know where to start because you've never had to explain it before.

AI cannot write this from scratch. It doesn't know your system, your naming conventions, or that one journal entry that always has to be reversed manually because of how the ERP handles intercompany eliminations.

Try this instead:

"I need to document a process I do regularly at work in finance or accounting. Don't write anything yet. Instead, ask me step-by-step questions to help me describe what I do, what tools I use, where things typically go wrong, and what someone would need to know to do this without me. Once you have the full picture, turn it into a clear written process."

What you save: the two hours you would have spent staring at a blank doc, plus the back-and-forth when the person who reads your draft has six follow-up questions.

What you gain: a document that actually captures how the work gets done, not a sanitized version that looks good but misses everything important.

You have to walk leadership through a number they won't like

Variance is up. A project came in over budget. A tax estimate was off. You know the story behind the numbers, but you also know this room, and you know someone is going to ask the one question you haven't fully prepared for.

Asking AI to "help me prep for a hard meeting" gives you advice so generic it could apply to any meeting about anything.

Try this instead:

"I need to prepare for a meeting where I have to explain a financial result that leadership won't be happy about. Before giving me talking points, ask me questions about what the numbers are, what caused them, what the audience already knows, and what I'm most worried they'll ask. Then help me prepare for both delivering the message and handling the pushback."

What you save: the mental spiral the night before, and the moment in the meeting where someone asks something you hadn't thought through.

What you gain: a clear narrative, answers to the questions you're actually dreading, and a lot more confidence walking into the room. It may not change the outcome, but that may not have been on you to begin with.

These are simple examples, but the same approach works for things that are a lot more complex. The bigger the task, the more the AI needs to understand your situation before it can help. And the more context you give it, the faster you get to something you can actually use.

In every one of these, you're not just getting a response. You're getting your own knowledge back in a form you can work with.

More soon.

One more thing: if you found the prompts in this useful, there's a good chance someone on your team would too. Forward this to them. The more people around you who know how to get useful output from AI, the easier your own work gets.