AI agents for workforce management: what we are building

Workforce management is where software meets the most complicated rules in a business. Pay rules stack on top of each other — base rates, overtime thresholds, night and weekend premiums, holiday rules, exceptions agreed years ago. Rosters have to respect contracts, rest periods and skills. And every error lands in someone’s pay.

We have been building workforce management software since 2019. We are now building AI agents to sit inside it.

Three agents, three jobs

• The rules agent explains and tests pay-rule configurations. Ask it why an employee was paid what they were paid, and it walks through the rules that applied. Change a rule, and it tests the change against real scenarios before it goes live.
• The scheduling agent checks and builds rosters against contracts, rest periods, skills and demand, and explains any conflict in plain language.
• The anomaly agent reads time and attendance data before payroll runs and flags what looks wrong: missing punches, unusual overtime, patterns that do not fit.

The same guardrails we use ourselves

These agents work on data that affects people’s pay, so they follow the same architecture we run our own company on: each agent has one job, a reviewer checks the output, and a human approves anything that changes a schedule or a payment. The agents recommend; people decide.

Why this matters

The hardest part of workforce management is not calculating pay — software has done that for decades. It is explaining it, changing it safely, and catching mistakes before they reach a payslip. That is work agents are good at, provided someone designed them to be checked.

Where we are

The agents are in build now. If you run workforce management software, or a workforce big enough that payroll errors cost you, come and talk to us at stand B1350.