HR analytics usually lives in a separate system, or in nobody's system at all. Payroll is in one place, benefits enrollment in another, expense reports in a third, and the questions that cross all three, what does an employee really cost, who is about to leave, where is the money going, get answered once a year in a spreadsheet that someone builds under protest.
A NetSuite account that runs payroll and expense reporting holds more of that data than most people realize. Four of the prompts in my NetSuite AI Prompt Library are written for it, and they cover the questions a CFO and an HR lead both want answered: benefits utilization, compensation benchmarking, employee expenses, and turnover.
If you're new to the library, it's a set of 150 prompts for NetSuite finance work that I released in January. Each one is a structured system prompt that you paste into Claude or ChatGPT along with your NetSuite data, and it turns the model into a specific kind of analyst with a defined method, a required output, and rules about what it's not allowed to make up.
Benefits Utilization Analyzer
This one casts the model as a senior benefits analyst and works through six blocks: health, retirement, insurance, leave, other benefits, and the cost analysis that ties them together. The questions it's built to answer are the ones that get asked at renewal time. What are benefits costing per employee? What's the participation rate in the 401(k)? Which categories are growing faster than headcount?
It ships with five SuiteQL queries: a benefits cost summary, cost per employee, 401(k) participation, cost trend, and cost by category. The thresholds are plain. A cost increase under 5% is fine, 5% to 10% is worth a look, and over 10% is a concern. Cost per employee under $15,000 is fine, and over $25,000 is a concern. You can change those, but the model can't decide on its own that a 12% increase is nothing.
Compensation Benchmarker
The compensation prompt is the one I'd expect to be used most carefully, because it's the one whose output goes into difficult conversations. It analyzes internal compensation distribution, looks for pay disparities among similar roles, positions salaries against ranges, and compares departments and levels.
Its queries pull compensation by department, by job title, by tenure, and a manager-versus-direct-report comparison, plus an outlier detector. The risk table has three rows that anyone who has done compensation work will recognize: pay variance within a role above 25% is high risk, a gap under 10% between a manager and their reports is compression, and three or more cases where a report out-earns their manager is an inversion problem. Every one of those findings is flagged for human review before it goes anywhere, because the data can tell you there's a gap and can't tell you why.
Employee Expense Analyzer
This is the most immediately practical of the four, because expense reports are the HR data most NetSuite accounts have. The model becomes an expense management analyst and works through travel, meals and entertainment, office and supplies, communications, policy compliance, and anomaly detection.
The queries are an expense summary by category, a top-spenders analysis, a policy compliance review against limits, duplicate detection, and a trend analysis. The duplicate detector is the piece I'd run first in any account: same employee, same amount, same category, within a few days. The prompt's thresholds treat a duplicate rate over 3% as high risk, and a policy violation rate over 15% likewise. The output is a category breakdown, a compliance report, and a list of anomalies with the evidence attached.
Employee Turnover Analyzer
Turnover is the prompt that reads the employee record rather than a transaction. From hire dates and release dates it computes monthly turnover rates, breaks them down by department and role, analyzes tenure at departure, and estimates replacement cost. The last block looks for predictive indicators, which is where the prompt is most careful about what it's allowed to claim: a pattern in past departures is evidence, and a prediction about a named person is not something the model is permitted to make.
Thresholds: annual turnover under 12% is fine and over 20% is a concern; early exits, meaning departures inside a year, over 35% are a concern; management turnover over 15% likewise. The deliverable pairs the rates with a cost estimate, because turnover that's expressed as a percentage gets nodded at and turnover expressed in dollars gets a budget line.
What They Have in Common
All four require the same discipline as the finance prompts in the library. The role is defined. The method is fixed. Every rate carries its calculation. Anything about a specific person is flagged for a human, and the model is told to report a missing data source rather than work around it. For HR data that last rule matters more than usual, because the cost of a confident wrong answer about someone's pay is higher than the cost of a gap.
A Note on Sample Runs
For most of the prompts in the library I've been able to run them against one of my NetSuite test accounts and show the output. Not these. The test account has 31 employee records and a handful of expense reports, but no payroll, no benefits enrollment, and no time entries, because a demo account doesn't carry them. Running the prompts would produce a report that says "no data" four times, which is the correct behavior and not worth a screenshot. If your account runs payroll in NetSuite, these prompts have what they need.
Who They're For
Controllers who own the payroll and benefits accounts and want to understand them before the renewal meeting. HR leads who don't have an analytics tool and do have a NetSuite login. And finance consultants who get asked the "what does an employee cost us" question and would like an answer with a method behind it.
All four are in the NetSuite AI Prompt Library, under HR and Workforce Analytics.