AI in HR for generalists and HRBPs
Ultramarine Work drafts FMLA notices, handbook addenda, warning letters and termination packets from the files in your HR folder. It prices jobs against your salary surveys, builds people reports and flags anything it couldn’t verify.
Final Written Warning and 30-Day Performance Improvement Plan
Performance concern
Your gross renewal rate was 81%, 78% and 84% in July, August and September, against the 90% target in your role expectations. We discussed the gap in coaching sessions on July 29, August 26 and September 23, and you received a written warning on October 6 under Section 7.2 of the Employee Handbook.
Goals for the next 30 days
Reach a 90% gross renewal rate on the 14 accounts renewing by December 10, and log a renewal plan in the CRM for every account due before January 31. Meet with your manager each Monday to review progress. If these goals are not met by December 10, further action up to and including termination of employment may follow.
CheckedDates match the coaching notes. Confirm the October 6 warning is in the personnel file before this issues.
How it works
- 1
Fill the HR folder
Add your handbook and policy files, job description templates, ER case notes, leave requests, I-9 copies, payroll and HRIS exports, and the salary surveys you participate in. Keep FMLA certifications in their own subfolder.
- 2
Name the piece of work
Ask for the FMLA case file, the warning and PIP or the job pricing. Each file comes back finished: the WH-382 with the five-business-day clock dated, the warning citing the handbook section, a source on every tab of the workbook.
- 3
Review and decide
It names the file behind each number and lists what it couldn’t reconcile, so review means checking the trail instead of redoing it. Nothing posts to the HRIS and nothing goes out. You send the WH-381 and make the call.
Ask it for
“Draft the job description and interview questions for a facilities technician in Renton, WA at $52,000 to $64,000, with the benefits wording the state requires.”
Facilities-technician-JD.docx
“Build the FMLA case file for the intermittent-leave request we got March 3: eligibility math, the WH-381 eligibility notice and the WH-382 designation notice.”
Intermittent-FMLA-case-file.docx
“Write a final written warning and a 30-day PIP for a customer success manager who missed renewal targets for three months, with goals from the coaching notes.”
Final-warning-and-PIP.docx
“Test the Denver selection criteria for the five coordinator roles against the four-fifths rule and tell me whether the layoffs trigger WARN.”
RIF-impact-and-WARN-check.xlsx
“Price the payroll operations specialist role against the three salary surveys in the folder, age the data to April 1, and return ranges with compa-ratios.”
Payroll-specialist-pricing.xlsx
“Put together the Q3 people report from the HRIS export: headcount by team, annualized voluntary turnover, time to fill, with any cell under five suppressed.”
Q3-people-report.xlsx
Inside the Human Resources mode
Hiring and selection
Turns a requisition into a job description, lawful interview questions and a scorecard, and checks the posting against pay range and salary history bans.
HR policy and handbook
Drafts the state addenda your handbook is missing, calls out where two states conflict, and reviews policies against NLRA Section 7 before they go out.
Performance discipline and termination
Puts together the warning, the PIP with dated goals and the termination packet, and sets out the final-pay deadline for the state the employee works in.
Investigation plan and findings report
Lays out the plan, the evidence log and the interview outline, marks where an Upjohn advisory belongs in the record, and drafts a finding on each allegation.
Leave and accommodation case management
Carries the leave case file through eligibility math, WH-381 and WH-382 notices, interactive process notes, intermittent use and return-to-work letters.
Rif and restructuring plan
Builds the selection grid on objective criteria, applies the four-fifths test, works out whether the layoff trips WARN and assembles the OWBPA release terms.
HR compliance and records
Audits I-9s for missing dates and signatures, tracks harassment-training completions, sorts what belongs in the medical file and prepares EEO-1 workpapers.
Small business HR setup
Lists what a first hire needs: EIN, state withholding and unemployment accounts, workers’ comp, new-hire reporting, the I-9 and the required posters.
Worker classification and FLSA exemption
Tests a role against the duties and salary-basis requirements, runs the contractor analysis on the actual working arrangement, and estimates back pay exposure.
Benefits administration and compliance
Reconciles carrier invoices against deductions, writes open enrollment notices and drafts ACA, COBRA and Form 5500 workpapers for review.
Compensation benchmarking
Matches jobs to surveys on work content rather than title, ages each source to one effective date, blends them and returns ranges with compa-ratios.
Performance and talent reviews
Manager notes become review records that cite observed behavior, match ratings to history and avoid protected traits, leave status and medical detail.
People metrics and analytics
Produces headcount, turnover, retention and time-to-fill reports with standard definitions and suppressed cells, so the numbers match across teams.
Engagement survey analysis
Reports survey results at the minimum group count you promised employees, scrubs comments of identifying detail and suppresses cuts that expose small groups.
Inclusion program and legal review
Reviews ERG, mentoring and fellowship programs for open eligibility and lawful use of demographic data, and flags preference or quota language.
Headcount and workforce plan
Models the headcount plan at fully loaded cost, with employer taxes and their wage-base caps, benefits, mid-year proration and the backfill effect of attrition.
Close variance and forecasting
Explains the people-budget variance by line and driver (attrition versus plan, late backfills and open roles) and rolls the forecast to year end.
Pay equity analysis
Runs the regression and cohort studies on your HRIS export, sizes the remediation budget and packages the findings for review with legal.
- Office skills
- Excel, Word, PowerPoint, PDF
- Tools
- Read, edit and write files, Search files, Run code, Search and read the web, View images
- Connectors
- SharePoint, Google Drive, Gmail, and additional custom connectors
- House style
- Your fonts, layouts and rules, on every file
Who it’s for
HR generalists, coordinators and assistants, HR business partners, employee experience and inclusion staff, HR managers and heads of people who run the employee lifecycle, advise managers and plan the people function.
Questions
Does it screen candidates or decide who gets promoted?
No, it never screens candidates out and never makes the promotion call. It drafts the criteria, the interview questions and the scorecard, and runs the adverse impact math you ask for, but a person scores the candidates and decides. The EEOC treats an AI tool that informs a hiring decision as a selection procedure under Title VII, and the employer owns any disparate impact even when a vendor built the tool. If Local Law 144’s bias audit applies, the call is for your lawyers.
Can it keep FMLA and ADA medical files separate from personnel files?
Yes, if you set the folder up that way. The app changes files only in the project’s folder, so keep medical certifications and accommodation notes in a separate subfolder and say which documents may draw on it. What it writes cites the files it used, so medical detail appears only where the request points, and you review every file it touched. Supervisors may be told restrictions and accommodations only, and 29 C.F.R. 1630.14 requires exactly that split.
If we get an EEOC charge, are our AI prompts and outputs records?
They can be, and the safe answer is to treat them as records. The records rule, 29 C.F.R. 1602.14, requires employers to preserve personnel and employment records for a year, and every record relevant to a pending charge until final disposition. Real disputes have turned up a prompt typed into a personal AI account and the draft it produced. Ultramarine Work writes its drafts into the project folder, so the same records rules apply. What to keep is your counsel’s decision.
How does it handle sensitive employee data?
Ultramarine Work changes files only in the project’s folder. When a request needs it, it can read files elsewhere on your computer, never passwords, keys or app data, and you can turn that off in settings. The model runs on Paper Instruments’ servers, so the content it works on is sent there to be processed. Paper Instruments is SOC 2 Type II, and our security and compliance details are on our compliance page.
Our handbook covers nine states. Is the state law right in each one?
No, not without your own state-law check. Give it the handbook and the states where people actually work, and it drafts the missing state addenda, flags state conflicts, reviews the language against NLRA Section 7 and says what it could not confirm. The traps lawyers find in AI-drafted handbooks are real: invented final-pay rules, deadlines that no longer hold and thresholds that ignore employer size. Counsel should read anything material before the handbook ships.