Beyond the Hype: 5 Questions to Find the AI Use Cases That Matter

The push to adopt artificial intelligence in HR is everywhere. Executives are asking how it can be implemented, and HR teams are eager to find tools that give them leverage and save time. Scroll through any tech vendor’s website and you'll see "AI-powered" in every demo and message.

This simultaneous pressure to adopt AI and the flood of potential ways to do that makes it hard to be thoughtful about when and how you use AI. Most organizations have arrived at the conclusion that you should try to use AI, but now the hard question is how do you use it to create real value?

To help you cut through the noise and find the features and use cases that will actually have an impact, here are five questions you can use to evaluate any AI feature or application:

1. Is it useful — or just impressive?

It's easy to confuse a technically impressive demo with a genuinely useful feature.

Imagine a demo in which an AI chatbot window pops up and the presenter types, “Show me all the employees who are below their salary range midpoint.” The list appears instantly. It looks like magic.

But how would you do that without AI? Well, your spreadsheet already had a “Compa-Ratio” column. You’d sort by that column and get the same data in two clicks.

The AI chat feature is exciting, but is it actually easier or faster than the existing user interface? In many cases, these AI features are just a new, conversational layer on top of a simple user experience that already exists. The key test is: Is this actually faster or easier compared to doing it myself in a well-designed tool? Don't let the novelty of the interaction distract you from the underlying value.

Just because you now can use your software by typing out an essay doesn’t make that easier than clicking a button.

2. Why wouldn't I just use ChatGPT?

Or why wouldn’t I just use Gemini, Claude, etc.?

In the push to be seen as adopting AI rather than falling behind, many vendors have added very simple integrations of AI capabilities: You have a chat window in the app that is a thin wrapper over a foundation model that has been provided with some basic context (the vendor product documentation, for example).

This might save you a bit of time, but this advantage is eroding fast. For example, don’t be impressed by an HR tech demo that uses AI to generate an example job architecture or a compensation planning merit matrix: The major AI providers can do these kinds of tasks out of the box. They're also gaining more context about your specific company (think Microsoft Copilot in your full Office suite or Gemini in your Google Workspace).

Be especially critical if you have to pay extra for a feature like this. Are you paying a premium for marginal functionality that you could get for free? A truly valuable AI capability should be deeply integrated into your tools and data, such that you can’t easily get the same result by opening your favorite AI chat and prompting a bit.

3. Is it secure and compliant?

If your company hasn't already created an AI-specific security and compliance evaluation, it's coming. We have been seeing these AI questions for multiple years and are starting to see very in-depth AI-specific questionnaires from the enterprises we work with.

Startup vendors that are moving fast to integrate AI may not always possess the resources to have even thought through all the compliance questions your company will have.

This raises the bar even further for the actual usefulness of the AI capability you're being sold on. If it doesn't pass your security review, it doesn't matter how useful or impressive it is.

4. Does it remove work or just create new, different work?

We assume AI will save us time. But what if it just changes the kind of work we do, without reducing the amount?

A recent study tested engineers on tasks in a project they were familiar with to see if they were faster using AI or not. It turned out AI made them around 20% slower. The key finding? They spent a lot more time than expected correcting a bunch of things the AI got close, but not quite right. That was with coding tasks, where billions have been invested in making AI great.

Ever asked AI to draft you an email and then spent much longer than you expected cleaning up places where it didn't sound quite right? Yes, you saved initial draft time, but what happens if you add the edit time?

Here's my rule of thumb to find use cases that save time: Use AI for execution leverage, don't outsource your strategy.

For most people, AI will give the best return on investment (ROI) as a brainstorming partner or for tasks where you know what you want and can describe it clearly, but it’s much faster to describe than to do.

Some examples:

• As a brainstorming partner: Instead of “Write me a draft total rewards philosophy for our company,” try "Give me five questions that a total rewards philosophy for a company should answer."
• For clearly defined tasks: Instead of “Give me a strategy for improving employee retention,” try "Make a table showing which of these 50 exit interviews indicated compensation was a contributing factor."

If the job doesn't fit those categories, be extra critical of the value you’re getting. With all the leadership and social pressure to use AI, you have to be careful not to lose sight of the goal. Using AI is a means, not an end in itself.

5. Does it show its work, or does it just give me results?

For now and in the near future, using AI means leveraging it to perform a task and then checking its work. The check part is only possible if you either already know what a good result looks like or if the AI shows the steps it took and you can verify those.

If the AI’s effort is a black box, then what happens when it makes a mistake? Can you identify what went wrong? Can you undo it? Will you even catch the mistake?

Consider these scenarios:

• An AI tool "intelligently" adjusts your salary ranges based on "market trends." Can you see why it changed them and what specific data sources it used? If you don’t already know what adjustments should be made, can you trust this?
• An AI generates performance review language. Can you trace back the inputs to specific notes that you took?
• An AI recommends merit increases. Can you audit the logic to understand if you agree with it?

The key principle is this: AI should augment your decision-making, not replace your ability to understand and reverse those decisions.

The best AI tools act as an assistant. They show their work, and they cite sources you can look up. If you don’t already know what "good" looks like and you can’t reverse-engineer the AI’s reasoning, then the output has little value.

Conclusion

The directive to adopt AI is coming from the top for a good reason: There is real value to be found. There is also a tremendous amount of noise. Not every "AI-powered" feature is a step forward, and some are a step back in a fancy new package.

I've used these five questions to think critically about AI features, both when evaluating vendor capabilities and when deciding what problems AI can actually solve for our team. They've helped me separate genuine opportunities from expensive distractions.

The key is to be disciplined. Don't let impressive demos or executive pressure push you toward AI adoption that doesn't pass these tests. The companies that will get real ROI from AI in HR are the ones asking where it actually helps, not just racing to check the "we use AI" box.

Author

Peter McKee is founder and CEO of Aeqium, a compensation management platform that enables leaders to make better compensation decisions faster by instantly adapting planning workflows to your unique compensation philosophy and automatically surfacing the insights that matter to your business.