Identifying AI-amenable tasks in the patient care workflow and matching enabling AI technologies

Overview

Contemporary patient care is delivered through complex, multi-step workflows that span triage, assessment, diagnosis, documentation, care coordination, treatment, and follow-up. Many steps in these workflows are repetitive, cognitively burdensome, or prone to delays and errors, and a substantial share of clinicians' time is consumed by tasks that require limited clinical judgement, such as documentation, information retrieval, and administrative coordination. Persistent workforce shortages, rising demand, and clinician burnout have intensified interest in whether some of these tasks can be safely delegated to, or automated by, artificial intelligence.

However, enthusiasm for AI in healthcare has often outpaced clear evidence on which tasks are genuinely suitable for automation and which AI technologies are best suited to each purpose. Not every bottleneck is an AI problem, and not every AI capability maps cleanly onto a real clinical need. This project takes a deliberately structured approach. It first characterises where the major recurring issues in a defined patient care workflow lie and prioritises those that are plausibly amenable to AI support or automation using explicit criteria. It then maps candidate AI technologies and algorithmic models to those prioritised tasks, and assesses the feasibility, risks, and evidence base for each match. The intended contribution is a defensible, criteria-driven framework for deciding what to automate and with what, rather than a broad catalogue of possibilities.

Research activities

The project is organised into four sequential phases, each producing a defined output that feeds the next.

• Phase 1: Workflow characterisation and issue identification.
• Phase 2: Prioritisation of AI-amenable tasks.
• Phase 3: AI technology and model mapping.
• Phase 4: Feasibility, risk, and synthesis.

Outcomes

• Aim 1: Identify and prioritise the current major issues in a defined patient care workflow that could plausibly be outsourced to, or automated by, AI.
• Aim 2: Identify and match candidate AI technologies and algorithmic models capable of servicing the prioritised tasks, and assess their feasibility, risks, and evidence base.

Skills and experience

• Data science and analysis of clinical or administrative data.
• Familiarity with AI and machine learning concepts, including natural language processing and predictive modelling.
• Qualitative and mixed-methods research skills, including interviewing and thematic analysis.
• An interest in health services, clinical workflow, and the safe translation of AI into practice.
• Python programming is desirable.

Scholarships

You may be eligible to apply for a research scholarship.

Explore our research scholarships

Keywords

• Patient care workflow
• Clinical automation
• Healthcare AI

Contact

Contact the supervisor