Artificial intelligence as an augmented partner in school consultation: Applications across the problem-solving process

Synthesis: Wu, Dillon and Lockwood argue that school consultation — an indirect service model in which a consultant such as a school psychologist collaborates with a consultee such as a teacher to address a student's academic, behavioural or social-emotional needs — is the next site of generative AI adoption in school psychology practice. Their contribution is the augmented triad: AI joins the traditional consultant-consultee-client relationship as a partner for organisation, reflection, skill development and data visualisation, without absorbing the interpretive and ethical work. Mapped onto the four cyclical problem-solving consultation phases — problem identification, problem analysis, intervention implementation and evaluation — AI drafts note-taking and planning templates, builds data collection tools such as time-sampling sheets and Goal Attainment Scales, graphs baseline data in seconds, coaches consultees between meetings through custom GPTs and Gems, and synthesises progress monitoring data into summaries and visualisations. Two US surveys (199 school psychologists in 2024; 129 Ohio practitioners about a year later) show adoption rising from roughly half to 79%, mostly for recommendations, summaries and report writing. The authors insist that human consultants remain accountable for accuracy and decisions, that AI outputs are hypotheses rather than directives, and that privacy law, hallucination and bias constrain every phase. It is conceptual, not an outcome study; its evidence comes from adjacent fields and self-report surveys.

Key Findings

1. Consultation is an indirect service with a well-documented time problem. The consultant (for example a school psychologist) works collaboratively with a consultee (for example a teacher) to meet the needs of a client (for example a student), so success depends on relationship quality and the consultee's capacity to translate planning into practice (Erchul & Martens, 2010). Consultation is highly effective (Garbacz et al., 2026), but school psychologists lack time for it: psychoeducational report writing alone consumes an average of 7.5 hours per week (Filter et al., 2013), and most time goes to special education eligibility administration (Farmer et al., 2021).
2. The augmented triad is the paper's organising frame. AI joins the traditional consultant-consultee-client triad as a partner assisting with organisation, reflection, skill development, follow-up intervention support and data visualisation, built on the premise that AI augments human performance best when it supports rather than supplants human reasoning in cognitively complex domains (Joksimovic et al., 2023).
3. Adoption is already substantial in US practice. In a spring and summer 2024 survey of 199 school psychologists, roughly half were using AI at work — most commonly for generating individualized recommendations (53%), assisting report writing (37%) and summarising complex information (28%), with under a fifth doing so at least weekly (Farmer et al., 2025). A survey of 129 Ohio school psychologists about a year later found 79% using AI at work, 43% at least weekly, the same use patterns (recommendations 62%, summaries 55%, report writing 40%) and a median two hours saved per week (Lockwood, Shergill & Choi, 2025).
4. Problem identification: preparation, tools and rehearsal. AI can generate meeting and note-taking templates tailored to the consultant's style and context, build data collection instruments such as time-sampling spreadsheets that automatically calculate interval percentages, Goal Attainment Scales and Check-in/Check-out cards, and support rehearsal of communication skills (paraphrasing, clarifying questions, summarising, reflective listening), with evidence that AI tools can give feedback on those skills (Dillon et al., 2026).
5. Transcript-based reflection is available at every phase. With informed consent and an appropriately authorised platform, consultants can use AI transcription and analysis to receive feedback on fidelity to consultation steps or on the use of a particular approach such as strengths-based framing, and to ask what areas the meeting did not explore (Owens & Dillon, 2026).
6. Problem analysis: planning, knowledge and graphing. AI produces implementation planning templates covering the who, what, when, where and how of an intervention plan, generates psychoeducation handouts and skills-practice materials (for example target behaviours, modelling task steps, behaviour-specific praise), and creates accurate baseline graphs in seconds (a task that would otherwise take the consultant extra time), so the dyad can spend less time on what the data says and more on selecting a manageable goal (Yan et al., 2025).
7. Intervention implementation: on-demand coaching between meetings. Customisable assistants such as GPTs and Gems can act as stopgap coaching tools that explain intervention steps, run reflective questioning and scenario-based simulations, and promote self-regulated learning. The counterweight is explicit: LLMs hallucinate plausible but non-factual content (Huang et al., 2025), and responses are prompt-, context- and model-dependent, so a consultant verifying one response does not guarantee the consultee will receive equally accurate information.
8. Intervention evaluation: synthesis, debriefing and institutional memory. AI can integrate academic, behavioural and socio-emotional data streams into analytic summaries and visualisations and compare baseline with intervention performance (Mandinach, 2025), support guided debriefing dialogues that connect quantitative outcomes with qualitative experience, and — configured as a knowledge repository retaining intervention histories and outcome data — function as institutional memory for continuity across practitioners and settings.
9. Accountability does not transfer to the tool. Considering NASP Standards II.2.1 and II.2.3, school psychologists remain responsible for the accuracy of documents they produce and the appropriateness of recommendations and decisions they endorse; AI-generated insights are to be treated as hypotheses explored collaboratively rather than directives to follow uncritically.

What school consultation is, and where AI enters

School consultation is the indirect service model that anchors much of school psychology practice, and the paper builds on the problem-solving consultation (PSC) model, also called behavioural consultation (Feldman & Kratochwill, 2003; Kratochwill & Bergan, 1990). PSC organises the work into four cyclical phases: problem identification, problem analysis, intervention implementation and evaluation, with a strong consultant-consultee relationship established early and sustained across all phases (Kratochwill et al., 2014). Before formal problem identification, consultants carry out contracting activities that clarify confidentiality, roles, expectations, informed consent and procedures (Fallon & Bender, 2023) — in this paper those contracting conversations become the ethical foundation for data handling once AI tools enter the process.

The paper deliberately separates two things that are often conflated: AI that supports consultation practice (direct work with consultees inside the problem-solving process) and AI that supports consultation training and professional learning (skill rehearsal, feedback and supervision). Its central claim is that AI belongs inside the PSC cycle as an additional decision-support partner rather than as an administrative bolt-on, which is why the authors frame the resulting arrangement as an augmented triad — the consultant, the consultee, the client, and AI alongside them. Anthropomorphising the tool is the authors' own shorthand for a human-AI collaboration in which machine contributions stay procedural and analytical.

Applications across the four phases

The phase-by-phase proposals are concrete. In problem identification, AI prepares documents before and after the meeting and builds data collection tools matched to the concern, such as time-sampling matrices or Goal Attainment Scales. In problem analysis it addresses two known barriers to implementation — consultee knowledge and consultee skill (Coles et al., 2015; Owens et al., 2017) — by producing summaries, explanations, handouts and practice materials, and by graphing baseline data so that the problem-solving conversation is about goal selection rather than about reading a chart.

In intervention implementation, AI extends the consultant's instructional influence beyond scheduled meetings. Custom GPTs and Gems require no coding and are free to set up, and they can answer questions, supply practice prompts and give immediate feedback, which the authors liken to established performance-feedback mechanisms in behavioural consultation and treat as a route to self-regulated learning. In evaluation, AI supports the summative and formative objectives Sanetti et al. (2014) name — measuring goal attainment, assessing plan effectiveness, guiding consultation decisions and planning for maintenance — by unifying data streams, prompting debriefing dialogue and holding the case record. The authors also note that the consultation documentation a custom assistant retains may matter later, if the student is referred for special education evaluation and the team needs the intervention history. A supplement of sample prompts accompanies the article, developed and tested with ChatGPT and Gemini.

Where the consultant ends and the AI begins

The boundary the paper draws is consistent: AI supplies organisation, drafting, visualisation and structured reflection; the consultant supplies interpretation, relational sensitivity and ethical judgement. The authors cite Joksimovic et al. (2023) for the general finding that AI augments human performance when it supports rather than supplants reasoning, and Kohli and colleagues' work in applied behaviour analysis, where AI systems generate treatment recommendations and identify subgroups that respond differentially (Kohli et al., 2022), as evidence that decision-support roles in professional practice are already approximated elsewhere.

The sharpest limit is hallucination combined with context dependence. Because AI output varies with prompt, context and model, the authors advise the consultant to probe a topic with the model first — for instance a behaviour intervention — before inviting a consultee to use it independently. The human-only province is stated plainly: the interpretive and relational work of consultation, and the accountability for what is written and recommended, stay with the professional. AI is a cognitive and procedural partner in the teacher-facing work of building consultee capacity, not the agent of change.

Risks, ethics and professional constraints

Privacy is treated as a precondition rather than a footnote. Student data are part of an educational record covered by FERPA (1974), and consultee data are protected by confidentiality measures in NASP Standard I.2.1 (NASP, 2020), so only AI platforms operating under appropriate data-sharing agreements — Business Associate Agreements in the United States, or equivalent agreements elsewhere — are suitable for transcripts, progress monitoring data or case records. The authors warn that those agreements alone do not eliminate risk: platforms differ in how they store, retain, process and potentially reuse submitted data, so consultants and organisations must also evaluate vendor practices, security protections and transparency against institutional policy.

Bias and equity are the second constraint. AI models may reproduce biases in training data, especially where diverse populations are underrepresented (Ferrara, 2023), which is particularly consequential because consultation decisions must be culturally responsive and contextually appropriate (Newman & Rosenfield, 2024). The authors extend the equity argument to access, arguing that Open Source or district-level tools trained on diverse educational contexts could democratise data-informed consultation rather than concentrating it in well-funded districts, and calling for professional learning in digital ethics and data interpretation alongside systems-level policies.

Two more constraints follow from accountability. Consultants should be trained not only to operate these tools but to interpret their outputs critically, since holding professional judgement as the foundation means AI insights cannot be adopted as directives. And the paper recommends that schools establish AI consultation protocols stating how data are shared, how results are interpreted and how final decisions are made, so that transparency and accountability survive the automation of drafting and graphing. The regulatory detail throughout is US-specific (NASP standards, FERPA, HIPAA, BAAs), with the authors noting equivalent arrangements in other jurisdictions.

Limitations and what would need to be tested

This is a conceptual article, not an outcome study. Its applications across the four PSC phases are proposals, and the paper states directly that empirical evaluation of AI use in school consultation is still required; the supporting evidence it cites is drawn from adjacent fields (scoping reviews, applied behaviour analysis) and from survey data about general AI use by school psychologists rather than from trials of AI-assisted consultation. Its sample prompts were developed and tested with two specific commercial models, so their performance is not a general claim about AI tools.

The research agenda the authors set out doubles as the list of open questions: whether AI differentially affects problem identification, intervention planning, consultee learning, implementation fidelity or student outcomes across behavioural, mental health and organisational consultation models; whether AI-assisted consultation improves process variables (consultee engagement, efficiency of feedback, accuracy of data interpretation) or outcome variables (student behavioural and academic gains); and whether any effects persist longitudinally, since AI-assisted consultation could plausibly promote sustained fidelity and faster problem resolution but has not been shown to. The measurement gap is acknowledged too: constructs such as "AI-assisted fidelity monitoring" and "AI-mediated reflection" need operational definitions and validated instruments before the field can move past speculation. Finally, the framing, legal detail and professional standards are United States school psychology, and the authors themselves flag a declared conflict of interest — one author is a paid consultant to a test publisher on AI-assisted report writing — which readers should weigh alongside the low evidentiary base of what remains a practitioner-facing proposal.

Connected Concepts

• Human AI Collaboration — the augmented triad as a specific arrangement of human and machine contribution
• Teaching — the consultee's capacity as the mechanism through which student outcomes change
• Teacher AI Competency — consultee knowledge and skill as named barriers AI is proposed to relieve
• Generative AI — the model class generating templates, handouts, summaries and recommendations
• Large Language Models (LLMs) — large language models and the cited evidence on IEP goals and psychological reports
• Conversational AI — custom GPTs and Gems as on-demand coaching between consultation meetings
• Problem Solving — the four-phase PSC model this paper maps AI onto
• Special Education — eligibility administration, IEP goals and FERPA-protected records
• Privacy — BAAs, FERPA, NASP confidentiality and consent before recording consultation
• Hallucination Risk — plausible but non-factual output and prompt- and model-dependence
• Feedback — AI feedback on consultation skills, transcripts and consultee implementation
• Bias Mitigation — training-data bias in culturally responsive consultation decisions

Connected Articles

• Can AI deliver appropriate support for diverse student profiles? A large-scale evaluation — Evaluating whether LLM support fits diverse student profiles
• The Integrity of Psychology Assessments in the AI Age: A Critical Examination — Integrity and validity questions for psychology assessment in the AI age
• What do you mean by human-AI collaboration: Prerequisite functions and the affordances needed to achieve it — What human-AI collaboration requires before it can be claimed
• Architecting an AI-Driven Decision Support System for Enhanced Online Learning and Assessment — AI decision support systems for assessment decisions
• Cognitive Offloading in the Age of Generative AI: What Does It Mean for Students With Learning Disabilities? — Cognitive offloading and students with learning disabilities
• Addressing the Void of AI Policies in Education for Students With Specific Learning Disabilities — The policy void for AI and students with specific learning disabilities
• Special-R1: Reinforcement Learning for Special Education — Aligning LLM Tutors to Diverse Learners through Disability-Adaptive Training — Aligning LLM tutors to learners with disabilities through adaptive training
• Beyond Compliance: A Proposed Framework for Ethical Governance of Student Data in Learning Analytics — Ethical governance of student data beyond compliance
• Unpacking Interaction Profiles and Strategies in Human-AI Collaborative Problem Solving: A Cognitive Distribution and Regulation Perspective — Interaction profiles in human-AI collaborative problem solving

Citation

Wu, S., Dillon, C., & Lockwood, A. B. (2026). Artificial intelligence as an augmented partner in school consultation: Applications across the problem-solving process. Journal of Educational and Psychological Consultation. Advance online publication.