EXPOSED
The bread-and-butter deliverables — survey instrument drafting, job analysis write-ups, competency models, engagement report summaries, literature reviews, and standard psychometric runs — are exactly the text-and-statistics work current AI does at usable quality. What holds is the part where a named expert stands behind a selection system in an EEOC adverse-impact challenge, facilitates a leadership team through a restructuring they don't want, and decides which construct actually predicts performance in this specific job. Licensure rarely binds I-O practice (unlike clinical psychology), so the shield is thin; the moat is client trust and expert-witness-grade accountability, not regulation.
Mixed — a routine tier and a judgment tier. An 8 reflects that item-writing, factor analyses, validation tables, and 360-feedback narrative reports — the volume work billed by the hour — now come out of a model in minutes, while the residual defensible core is narrower: running a live focus group with unionized supervisors, deciding whether a criterion measure is contaminated, and defending a cut score under Uniform Guidelines scrutiny.
Fully desk- and screen-based. A 4 is for the occasional on-site visit — observing assembly line tasks for a job analysis, sitting in on an assessment center as an assessor — inside client offices and plants where the physical demand is walking and watching, not handling anything.
Certification preferred, not legally required. At 5, the psychologist title is protected in many states and SIOP/ABPP credentials carry weight with sophisticated buyers, but I-O consulting to organizations is not a licensed act of practice, so a firm can staff a validation study with unlicensed master's-level analysts and nothing stops it.
Some relationship component. A 12 rather than 16 recognizes that CHROs bring back the specific consultant who survived the last layoff round with them and can name their VPs, but much of the market is procurement-driven — engagement survey vendors, assessment platforms, RFP-awarded validation studies — where the contract, not the person, is the relationship.
Meaningful discretion. 12 is earned by calls like whether an integrity test's adverse impact is justified by business necessity, whether to tell the CEO their pet competency model is unvalidated, and how to sequence a restructuring communication — high-consequence but usually advisory, with the client executive owning the decision and the legal exposure.
School Psychologists SAFE
Has AI actually changed your work?