Practitioner practice & AI-assisted work7

AI literacy — the knowledge and skills required to understand, use, and critically evaluate AI systems — has been proposed as a core competency for social work, relevant even to practitioners who never directly use AI tools.

Sources: ahn2025

Appears on: /pan-lab, /practice/ai-literacy

In a two-year child-welfare ethnography, a re-purposed assessment algorithm produced process-oriented harms to practice, organization, and street-level decisions, compelling caseworkers to perform added repair work; 80% of interviewees reported that the tool had stripped their decision-making discretion.

Sources: saxena2024, ammitzbollflugge2021

Appears on: /pan-lab

The same agency's theory-driven 7ei tool — which tracks case trajectories instead of predicting outcomes — earned collective buy-in and better engagement, but required sustained investments: trauma-informed training, specialized supervision and expert consultation, and new collaborative staffings.

Sources: saxena2024

Appears on: /pan-lab

In a participatory-design study (CHI Late-Breaking Work) with 51 social-service practitioners across two stages (27 in co-design workshops, 24 in contextual inquiry), AI value concentrated in documentation relief, assessment brainstorming, guidance for junior workers, and supervision support — with deskilling and privacy concerns voiced inside the same sessions.

Sources: tan2025

Appears on: /pan-lab, /what-ai-can-do

AI documentation assistance can cut clinician documentation burden substantially, but the efficiency paradox converts freed time into added caseload unless organizational policy protects it — time returned is realized as benefit only when governance decides where the dividend goes.

Sources: vanhara2026

Appears on: /pan-lab, /what-ai-can-do

A human-services AI framework argues organizations should start from their own practice challenges and ask which AI capabilities might help, rather than adopting vendor tools first, and pair that with digital stewardship — discernment, accompaniment, and attunement — noting that most organizational AI investments have shown no meaningful return.

Sources: goldkind2025

Appears on: /pan-lab, /what-ai-can-do

The health-care chapter renders the NASEM Assistance category as reach as much as throughput: post-discharge texting that checks whether a patient obtained their medication and alerts a worker when something is wrong, and round-the-clock operation treated as timely aid, extend assistance beyond the clinic visit. The chapter states this as a practitioner expectation it argues for, and attaches no measured coverage, uptake or outcome figure to it.

Sources: ji2026

Appears on: /pan-lab