It is hard to think clearly about AI in a People function while you are inside any one workflow. The detail of should this prompt include the JD or just the requirements list swallows the larger question of where, across the whole function, is AI worth investing in at all.
So this piece is a map. Not a strategy. A picture of the estate. Five domains, each with the patterns that tend to work, the patterns that tend to disappoint, and a sense of where the leverage actually sits.
Use it as a wall to pin your own work against. Anywhere a domain is empty in your function is a place to look. Anywhere it is crowded is a place to consolidate.
Domain 1 — Talent Acquisition
The most-explored domain, with the most mature patterns and the most over-claimed vendors.
Where AI consistently helps: inbound triage with human-in-the-loop decision; JD drafting from real role context (not generic templates); interview scorecard synthesis; structured de-biasing of language in candidate-facing comms; outreach personalisation grounded on actual public information; offer-letter assembly.
Where AI consistently disappoints: end-to-end "AI sourcer" tools that promise to find candidates autonomously; chatbots that try to replace a human touchpoint with a candidate; anything that claims to predict performance from a CV.
The grain: TA is a high-volume, high-judgment function. AI scales the volume work and protects the judgment work — but only if the team holds the line. The teams that get this right use AI to clear the bottom and top of the funnel and put recruiters on the messy middle.
Domain 2 — Onboarding & Lifecycle
Less explored, often higher leverage. The first month of an employee's tenure is where the most can be improved with the least friction.
Where AI consistently helps: first-week sequencing and nudging; manager prompts for new-joiner check-ins; personalised learning path drafting from role + level + tenure; offboarding handover packs assembled from where work lives.
Where AI consistently disappoints: "AI buddy" chatbots that try to replace human connection in the first week; anything that automates a moment where the new joiner needed a person.
The grain: lifecycle work has clear handoffs and templates already. AI is mostly making the templates living, not replacing the humans in the moments that matter. The bar for adoption is low because the alternative is "a coordinator forgot."
Domain 3 — Performance & Development
The hardest domain to get right, because the consequence of a wrong call is the highest and the data is the messiest.
Where AI consistently helps: pre-read assembly for review cycles; calibration session preparation (surfacing inconsistencies in language across managers); 1:1 prompts grounded on prior conversation themes; self-reflection scaffolding for employees writing their own reviews; learning recommendations grounded on real role and project history.
Where AI consistently disappoints: anything that scores or rates an employee directly; "AI coach" tools that try to replace a manager conversation; bias-detection tools that produce a number rather than a conversation.
The grain: performance is a judgment domain. AI prepares the table; humans decide what is on it. Teams that respect this line make their performance cycles better. Teams that cross it produce performance theatre that the company eventually rejects.
Domain 4 — Operations & Compliance
Quietly the highest-ROI domain, and the most under-invested.
Where AI consistently helps: policy Q&A grounded on the actual handbook; document drafting for letters, contracts, references (with human review always); reconciliation workflows across HRIS, payroll, finance; audit trail assembly; regulatory horizon-scanning summaries.
Where AI consistently disappoints: anything that signs, files, or commits without a human review step; legal advice generation; anything that touches data the model was not supposed to see.
The grain: ops work is high-volume, low-creativity, high-correctness. The AI shape that fits is "draft, log, present for approval, never commit." Done well, this is the domain where the team gets back the most hours per week per build.
Domain 5 — Strategy & Insight
The most-talked-about, the least-built. Reasonably so — the data quality bar is highest here.
Where AI consistently helps: survey response synthesis at scale; theme extraction from open-text feedback; meeting note synthesis for People exec syncs; pattern surfacing across exit interviews; turning rich qualitative data into shape that informs quantitative decisions.
Where AI consistently disappoints: "AI dashboards" that try to predict attrition or engagement from thin data; insight generation that the team cannot trace back to source; anything that claims to find a pattern the team had not already half-noticed.
The grain: insight work is where AI most often hallucinates confidently. The discipline that makes it useful is grounding everything in retrievable source — quotes, transcripts, individual responses — and treating model output as a draft for a human to verify, never as truth.
How to use this map
Three things, in order.
First, walk across all five domains and ask: where is the team already losing time we wouldn't get back? That is your priority list.
Second, walk across again and ask: where is the data clean enough to build now? The intersection of those two lists is your first quarter of work.
Third, leave the rest alone. The domains where you are not yet building are not failures — they are the next quarters' work. Trying to move on all five at once produces five half-built things and one exhausted champion. The map exists so you can see the whole picture and still pick.
The function that has built across all five domains in eighteen months looks unrecognisable. The function that tried to build across all five domains at once looks unchanged. Same map. Different sequencing. The shared substrate underneath every domain is the team's AI workspace, and the shared scaffold underneath that is an AI operating system shaped to People.
What this connects to
Auto-recommended next reads in the People Ops cluster, ranked by shared concepts and headings:
- Automation patterns that pay off
- AI workspace setup for People teams (Claude, ChatGPT, Copilot)
- Designing the AI-native People team
- A workflow assessment framework for People Ops
If this resonated, there's more.
Subscribe to receive new Intelligence pieces as they're published. No noise — just the work.
By subscribing you agree to our Privacy Policy. Unsubscribe any time.
