4 of 8

Some HR practices translate, some change form, some have no answer yet

With the territory mapped, here is how it works in practice: each process of human resource management, side by side for people and for AI agents, whether that is a person with an agent, an agent on its own, or agent to agent.

  • Translates the practice carries over
  • Changes form it carries over in a different shape
  • Open question no agent equivalent is known yet
Each human resource management process, for people and for AI agents, and whether it translates
ProcessFor peopleFor AI agentsStatus
Workforce planningHeadcount and the mix of skills an organisation needs.Which tasks go to agents and which stay with people; the human–agent mix and the compute budget behind it.Translates
Recruitment and selectionInterviews, tests, references and trial periods.Choosing a model and an agent set-up, evaluating it on the organisation’s own tasks, and probing how it fails before it is trusted.Translates
Job designRoles, discretion and reporting lines.Breaking work into tasks, scoping tools, setting how many steps an agent may take alone and where it must stop for approval.Changes form
Onboarding and socialisationInduction, mentoring and learning “how things are done here”.System instructions, examples of good work, and retrieval over the organisation’s documents, all held in a finite context window.Changes form
Training and developmentCourses, coaching and experience that build capability over a career.Fine-tuning, memory and reusable skills; much of an agent’s “development” arrives as a new model version rather than through experience.Changes form
Motivation and incentivesPay, promotion and recognition that change what people do at work.Reward shapes behaviour during training, not on the job; a deployed agent responds only to instructions, context and constraints. What plays the role of incentives is an open question.Open question
Performance managementAppraisal, feedback and development conversations.Continuous evaluation: tracing each step, automated and model-based judging, and human spot checks; for the people involved, measuring what they add.Translates
Norms and cultureShared values and the social pressure of peers.Policies and written constitutions in an agent’s instructions; conventions that may emerge among agents that interact repeatedly.Open question
Pay and costWages, benefits and the cost of people’s time.Tokens, compute, latency, and the human review time every agent decision consumes.Changes form
Discipline and exitWarnings, dismissal and offboarding.Revoking permissions, rolling back to an earlier version, and retiring a model when it is superseded.Changes form
Employee relations and voiceGrievances, representation and consent.Agents give no consent and have no voice, yet someone must answer for what they do; accountability has to be designed into the chain.Open question