For WorldatWork Members
- Beyond the Black Box: Why ‘Explainable AI’ is Non-Negotiable for HR, Workspan Daily Plus+ article
- Q&A: Using an AI Clone to Streamline and Scale HR Messaging, Workspan Daily Plus+ article
- From Hesitation to Action: Navigating AI in Payroll Management, Workspan Daily Plus+ article
- From Audits to Algorithms: Use Proactive AI for Sustained Pay Equity, Workspan Daily Plus+ article
- Rewards Require Architecture: Structuring AI Incentives That Work, Workspan Daily Plus+ article
For Everyone
- HR’s AI ‘Future’ Is Now; WorldatWork Can Help You Make the Most of It, Workspan Daily article
- Buying Beyond the Buzzwords: Critical Questions for Using AI in Comp, Workspan Daily article
- What’s Your Role in Making AI a Responsible HR/TR Tool? Workspan Daily article
- The AI Revolution in Sales Incentives: Rethinking How You Pay to Sell, Workspan Daily article
- Talking Tech: How AI Is Impacting Rewards, Skills and Compensation, Workspan Daily article
Most compensation professionals have had the same experience with artificial intelligence (AI):
- Type in a quick request;
- Get back something generic and a little bland; and,
- Quietly conclude the technology isn’t ready for the nuance of our work.
Pay structures, merit messaging, regulatory edge cases: Surely a chatbot can’t handle that.
It likely can. The problem usually is the brief, not the model.
Shifting the Mindset and the Returns
A mental shift can help change how you use these tools. Most people treat AI like a search engine: Type a few words, get an answer. Cutting-edge HR and compensation professionals treat it like a junior analyst: Capable and fast, but only as good as the brief you hand it. You wouldn’t drop “write something about merit increases” on a new analyst’s desk and walk away expecting a board-ready memo. The same rule applies here.
So, consider writing prompts the way you would brief that sharp new analyst. Over time, you may use that approach almost reflexively. We, at Novo Insights, call it CRAFT, and yes, the acronym is the point: CRAFT a better prompt, get a better result. (See the table below for details).
|
The CRAFT Framework | ||
|
The Step |
What to Do |
Background and Tips |
|
Channel a Role |
Tell AI who it is before asking what to do. Assign a professional role to unlock the right expertise and voice. |
Compare “explain merit budgets” to “act as a senior compensation consultant advising HR business partners and managers.” The second prompt gets you the framing, caveats and tone a seasoned practitioner would actually use. The key question here is simple: “Who is the AI in this conversation?” |
|
Root an Objective |
State what you are solving for in one sentence. Use one prompt for one goal — resist stacking multiple asks. |
This is where the most self-inflicted pain occurs. Comp pros typically stack three asks into one request: Draft the communications, build the frequently asked questions (FAQs) and summarize the policy. Then they’re surprised when all three come back half-baked. Pick the single outcome. “Draft an FAQ document explaining our merit increase guidelines.” That’s one job, clearly defined. The key question: “What single outcome am I solving for?” |
|
Add Helpful Context |
Give it what it can’t assume: company size, audience, timing, relevant systems and any organizational nuance/sensitivity that matters. |
The more of your real situation you hand over, the less generic the answer. The key question: “What does it need to know?” |
|
Frame the Output |
Specify format, length, tone and audience. The more precise you are about the deliverable, the less editing you’ll do. |
“Write an FAQ” gives you something. “One-page FAQ, eight questions max, written in common language” gives you something you almost can use as-is. The more precise you are about the deliverable, the less editing you do on the back end. And for those who spend cycle season editing, that’s the whole game. The key question: “What does the deliverable look like?” |
|
Tighten with Constraints |
Tell it what not to do. Guardrails prevent outputs that are technically correct but organizationally wrong. |
This is the most underrated step in the whole framework, and it’s the one comp pros are uniquely positioned to nail. We live in the land of things you cannot say: No legal claims, HR jargon and/or promises about future pay. An output that is technically correct but organizationally wrong is the kind of mistake that turns a helpful draft into a compliance headache. Spell out the boundaries up front. The key question: “What should it never do?” |
Value In, Value Out
Put it all five steps together and a throwaway request becomes a real brief: “Act as a senior compensation consultant advising HR business partners and managers. Draft an FAQ explaining our merit increase guidelines. We’re a 2,000-person company mid-cycle, our guideline document is attached and pay is a sensitive topic right now. Give me a one-page FAQ document, eight questions max, in common language. No legal claims, no HR jargon, no promises about future pay.”
That prompt takes maybe 90 seconds longer to write than “help me explain the merit guidelines.” The output is the difference between something you rewrite from scratch and something you tweak and send.
Know Before You Go
A couple of honest caveats before you proceed.
First, CRAFT doesn’t make AI accurate, it makes it aimed. You still own the final product. The model will occasionally invent a statistic or soften a message past the point of truth, and a constraint reduces that risk without eliminating it. The judgment stays with you. That’s a feature, not a bug. Your value was never the typing — it’s the expertise.
Second, not every task needs all five steps. A quick brainstorm doesn’t need tight constraints. A high-stakes manager communication during a tense cycle needs every one of them. The framework scales to the stakes.
Third, iteration is expected. Your first prompt often won’t return perfect results, so treat the response like a first draft from that junior analyst and refine the context or constraints until it lands.
At the end of the day, the comp pros getting real leverage from AI often aren’t the ones with secret prompts or fancier tools. They’re the ones who learned to write a good brief. Their expertise leads them to what they want (that’s their primary value) and they CRAFT the right prompt to generate the details most efficiently. You already do this every day with your analysts, vendors and business partners. CRAFT just brings that same discipline to the chat window.
So, the next time you’re about to type “help me with comp comms” and hit enter, pause and take another 90 seconds to channel, root, add, frame and tighten. Then, watch what comes back.
Want to Learn More About AI? Examine These WorldatWork Webinars
- Build Your Own AI Coach for HR & Total Rewards
- Design Your AI Collaboration Agent: A Hands-On Lab
- Expanding the Impact of Your Job Architecture with AI
- Finding AI Use Cases in Your HR & Total Rewards Role
- From Workshop to Workflow: Automating HR Tasks with AI
- How a Comp Guy Learned to Stop Worrying and Love AI
Click here to view the full slate of webinars.
Editor’s Note: Additional Content
For more information and resources related to this article, see the pages below, which offer quick access to all WorldatWork content on these topics:
#1 Total Rewards & Comp Newsletter
Subscribe to Workspan Weekly and always get the latest news on compensation and Total Rewards delivered directly to you. Never miss another update on the newest regulations, court decisions, state laws and trends in the field.
