Skip to main content

AI in HR operations, with the controls designed in.

PES helps organisations decide where AI should take on HR work from hire to retire, redesign the processes and support model around it, and lead implementation from the client side.

Three stages of work

  1. Assess
    Where AI could reduce manual work, what it would cost and what has to be controlled.
  2. Design
    The process, support model, decision rights and controls around the new way of working.
  3. Implement
    Client-side leadership of delivery, testing, acceptance and transition into operation.

Start with the work, then choose the technology.

HR, payroll and workforce platforms now include AI features, and vendors are adding agents to service desks, recruitment and workforce management. Switching them on is the easy part.

The harder questions are which work should change, what the AI may decide or only recommend, which data it relies on, and who answers for the outcome. Automating a poorly designed process makes its problems harder to see.

PES looks at the process, data, controls and support model first. The answer may be a feature you already license, a new product, a change to the process, or no AI at all.

Where AI is being applied from hire to retire.

The same questions come up in every process area: what the AI does, what a person still decides, and what evidence is left behind.

Recruitment and selection

Work AI can take on

Screening and ranking applications, scheduling interviews and answering candidate questions.

What to settle first

What the recruiter sees, whether lower-ranked candidates stay visible, and what record is kept of how the ranking was used.

Onboarding and offboarding

Work AI can take on

Coordinating tasks between HR, IT, payroll and managers, and chasing what is outstanding.

What to settle first

Which approvals stay with people, and how system access is granted and removed on time.

Employee and manager service

Work AI can take on

Answering policy and entitlement questions, triaging cases and routing exceptions to the right team.

What to settle first

Which source answers each type of question, which wins when sources disagree, and when the AI should hand over.

Time, attendance and rostering

Work AI can take on

Finding timesheet and roster exceptions and preparing the evidence for a supervisor or payroll officer to review.

What to settle first

Who approves each correction, and how evidence is kept to meet time and wage record obligations.

Payroll

Work AI can take on

Explaining payslips, investigating pay queries and preparing reconciliations.

What to settle first

Which pay decisions stay with authorised people, and what audit trail supports each change.

Workforce data and reporting

Work AI can take on

Answering workforce questions from HR, payroll and rostering data.

What to settle first

Which system is the record for each data item, who may see what, and how answers are checked.

The HR support model changes too.

When AI answers some requests, the work left for people changes in volume and difficulty. The support model has to be redesigned to match.

What PES designs with you

  • Which requests AI answers, which go to HR shared services and which go to specialists or HR business partners
  • Hand-off rules, so an employee is not left repeating a question once it reaches a person
  • Who owns the knowledge the AI answers from, and how it is kept current across policies, agreements and locations
  • Service levels, roles and team capacity once request volumes and work mix change
  • Workforce transition planning, and whether the change triggers employee consultation obligations

Controls to settle before anything goes live.

HR and payroll processes handle private information and decisions with legal consequences. These controls belong in the design and the test plan.

  • What the AI may answer, recommend or execute, and where a person must approve
  • Which data it can see, and the access, retention and data residency terms that apply
  • How it is tested before each release, including policy, award and enterprise agreement edge cases
  • Who owns the process, the technology and incident response once it is live
  • Whether a decision falls under the automated decision-making disclosure requirements that apply from 10 December 2026
  • What records are kept, including time and wage records that must be held for seven years

These are design prompts. They do not replace legal or privacy advice for your organisation.

How PES helps.

Each piece of work uses an existing PES service, applied to AI. You can engage PES for one of them or across all four.

AI opportunity and controls assessment

A review of processes, request volumes, effort, systems and existing licences to find where AI could reduce manual work, what it would take to do safely, and what benefit is realistic.

Typical outcome

A ranked list of opportunities, a baseline to measure against, control requirements and a business case.

Process and support model design

Redesigning the process before automating it. This sets out who does what across employees, managers, HR, payroll, technology platforms and AI tools, with approvals, exceptions and controls.

Typical outcome

Future-state processes, an HR support model and the control requirements for build.

Selecting AI products and assessing vendor claims

Requirements, demonstration scenarios and evidence requests that test what a product's AI does with your data, where a person reviews the outcome and what is recorded.

Typical outcome

AI requirements and scoring built into the selection, and documented vendor answers.

Implementation leadership and assurance

PES leads or supports the client side of delivery: planning, governance, design decisions, testing, acceptance and transition into operation. Build, configuration and integration sit with your vendor, systems integrator or internal technology team.

Typical outcome

Agreed acceptance criteria, tested controls and a service that is ready to run.

Independent of the products.

PES does not sell AI products, build agents or take referral fees from vendors or implementation partners.

That leaves PES free to recommend a feature you already pay for, a new product, a process change with no AI in it, or waiting until the product or your data is ready.

About PES

Considering AI in your HR operations?

Send a short note on the process, product or decision you are looking at, and we'll arrange a focused discussion about where to start.