Technology selection inside large organisations is decided by whoever is most enthusiastic, and enthusiasm correlates with novelty rather than value. The counter to that is not scepticism, which is just enthusiasm pointed the other way. It is a scoring framework agreed before anyone looks at the candidates, weighted according to what the organisation actually cares about, and then applied without adjusting the weights when the answer is inconvenient.
The supplied case was a mid-sized health network: three hospitals, fifteen community health centres, two specialist facilities, around 8,000 staff and 850,000 patients a year, with electronic medical records in place, basic telemedicine, limited cloud adoption, and low capability in AI, IoT and AR/VR. Five candidate technologies were on the table: robotic process automation, digital platforms, augmented and virtual reality, machine learning and AI, and IoT with big data analytics.
The failure mode was obvious in advance. Compare those five on capability alone and AI wins, because AI is the most capable technology in the list. It is also the one this organisation is least equipped to run, the one carrying the most unresolved ethical exposure in a clinical setting, and the one with the longest payback. A comparison that cannot express those facts will produce a confident recommendation to do the hardest possible thing first.
Applied properly, the framework ranked robotic process automation and IoT above machine learning. AI scored top marks on the heaviest-weighted dimension in the model and still came fourth, because it scored joint-lowest on organisational capability and on ethics. That is not a verdict against AI. It is a statement that this organisation cannot currently run it well or govern it responsibly, and that those are the constraints to fix before the capability becomes usable, which is a different recommendation and an actionable one.
The framework is the deliverable here. The healthcare verdict is the worked example, and it is the kind of verdict the framework exists to produce: unglamorous, sequenced, and arrived at without anyone having to argue against the most interesting technology in the room.
In brief
- Six weighted dimensions defined before scoring, with patient impact heaviest at 25% and capability and ethics deliberately included at 10% each as the only dimensions that can penalise a powerful technology
- Five technologies scored 1 to 5 per dimension with written justification per cell
- Final weighted scores: RPA 3.8, IoT and big data 3.8, digital platforms 3.7, ML and AI 3.4, AR/VR 2.8
- ML and AI scored joint-highest on patient impact (5) and joint-lowest on both capability (2) and ethics (2), demoting it to the second tier despite the highest ceiling
- Top three separated by 0.1, reported as a tier rather than a ranking to avoid false precision
- Three-phase roadmap over 60 months with each phase building the capability the next requires; Phase 1 targets a 30% reduction in manual data entry and a 500-patient IoT pilot
- $8 million across three years against a projected $15 to $20 million five-year return
- Framework designed to be reapplied as new candidates appear, which is the part that outlives the specific verdict
The report as submitted
Executive summary
This report evaluates five digital technologies for a regional healthcare network using a six-dimension weighted framework: robotic process automation (RPA), digital platforms, augmented and virtual reality (AR/VR), machine learning and AI, and IoT with big data analytics.
RPA and IoT with big data offer the highest immediate value, both at 3.8 out of 5, followed by digital platforms at 3.7 and machine learning and AI at 3.4. AR/VR shows the lowest near-term viability at 2.8, on implementation cost and technology maturity. The recommended three-phase implementation takes the deliverable technologies first, then patient-facing platforms, then advanced AI capability.
The investment case is $8 million across three years against an expected five-year return of $15 to $20 million, with change management and staff development treated as first-order costs rather than as overheads on the technology spend.
1. Introduction
Building on earlier assessments of chatbot platforms in the same course (Jordan 2024a, 2024b), which found 52% of patients using health information chatbots and practices achieving 30 to 40% reductions in administrative workload through automation, this report widens the evaluation to five further technologies.
The network faces an ageing population, rising costs, workforce shortages and rising patient expectations. The objective is to align technology investment with strategic priority while managing implementation risk. The assessment considers technical capability alongside financial implications, patient impact, operational benefit, organisational readiness and ethical exposure, on the principle that in a clinical setting patient safety and regulatory compliance are constraints rather than criteria.
2. Enterprise context
The organisation operates in a highly regulated environment under demand pressure, budget constraint and workforce shortage, with strategic priorities in patient-centred care, operational excellence, digital capability, workforce development and community health.
The readiness assessment matters more than the profile, because it sets the ceiling on what can be delivered. Capability is high in basic IT infrastructure and electronic medical records, medium in data analytics and mobile technology, and low in advanced AI and machine learning, IoT integration and AR/VR applications. Change readiness is supported by strong clinical leadership, established project management and a record of successful system implementations. Budget constraint and regulatory compliance are the standing limits.
That profile is the reason the scoring framework needs an organisational capability dimension. Three of the five candidates sit in the low-capability band, and a comparison that ignores this scores a technology on what it could do somewhere else.
3. Evaluation framework
Six dimensions, weighted before scoring, with the weights argued from what a healthcare provider is for rather than from what is easy to measure.
- Patient impact and experience, 25%. Care quality, satisfaction, access, engagement, health outcomes. The heaviest weight, because a health network that optimises anything above patient outcomes has misunderstood itself.
- Technical feasibility and integration, 20%. Implementation complexity, system compatibility, scalability, reliability, security.
- Financial considerations, 20%. Initial investment, operating cost, return, cost-benefit ratio, financial risk.
- Operational excellence, 15%. Process efficiency, resource use, quality assurance, decision support, automation potential.
- Organisational capability and change management, 10%. Staff readiness, change capacity, cultural fit, leadership support, training load.
- Ethical and social considerations, 10%. Privacy, equity of access, employment impact, transparency, social responsibility.
The last two are the ones that make the framework work. They carry small weights, but they are the only dimensions on which a powerful technology can be penalised for being powerful, and without them the exercise collapses back into a capability ranking.
Each technology scores 1 to 5 per dimension with the score justified in a sentence rather than asserted. Weighted totals map to priority bands: high priority 4.0 to 5.0, medium 3.0 to 3.9, low 2.0 to 2.9, not recommended 1.0 to 1.9.
4. Technology evaluations
4.1 Robotic process automation
RPA has matured in healthcare beyond simple administrative tasks into complex clinical workflows. Proven applications are patient registration and scheduling, claims processing, clinical documentation, supply chain management and compliance reporting. The limits are integration complexity with legacy electronic medical record systems, regulatory compliance requirements, the process standardisation work that has to happen before automation is possible, change management resistance, and ongoing maintenance overhead.
| Dimension | Score and justification |
|---|---|
| Technical feasibility | 4/5, high compatibility with existing systems |
| Financial | 4/5, strong return through cost reduction |
| Patient impact | 3/5, indirect benefits through efficiency |
| Operational excellence | 5/5, significant administrative efficiency gains |
| Organisational capability | 3/5, requires training but builds on existing IT capability |
| Ethical considerations | 3/5, job impact concerns, redeployment opportunities |
| Weighted score | 3.8/5, medium-high priority |
4.2 Digital platforms
Digital platforms have moved from patient portals to systems connecting patients, providers, payers and other parties. Telemedicine adoption accelerated after 2020, and patient engagement platforms increasingly target preventive care and chronic disease management. Applications are telemedicine, patient engagement and health tracking, provider collaboration, health information exchange, and digital therapeutics. The limits are interoperability across diverse systems, the digital divide affecting elderly and disadvantaged patients, privacy and security requirements, varying regulatory compliance, and user adoption.
| Dimension | Score and justification |
|---|---|
| Technical feasibility | 3/5, moderate complexity due to integration |
| Financial | 3/5, significant investment, long-term savings |
| Patient impact | 5/5, direct improvement in access and engagement |
| Operational excellence | 4/5, enhanced care coordination |
| Organisational capability | 3/5, substantial training and change management |
| Ethical considerations | 3/5, digital divide concerns against positive access impact |
| Weighted score | 3.7/5, medium-high priority |
4.3 Augmented and virtual reality
AR/VR has moved from experimental to practical in medical education, surgical planning and patient therapy, helped by falling hardware cost and better software. Applications are immersive anatomy training, 3D surgical planning and navigation, VR-based pain management and distraction, exposure therapy in mental health, and interactive patient education. The limits are high hardware and development cost, technology that changes fast enough to require frequent replacement, limited long-term clinical evidence, user acceptance problems including motion sickness, and integration complexity with existing workflow.
| Dimension | Score and justification |
|---|---|
| Technical feasibility | 2/5, high complexity and infrastructure needs |
| Financial | 2/5, high costs, uncertain return timeline |
| Patient impact | 4/5, significant potential in specific applications |
| Operational excellence | 2/5, limited current operational impact |
| Organisational capability | 2/5, requires significant new skills |
| Ethical considerations | 4/5, generally positive, few ethical concerns |
| Weighted score | 2.8/5, low-medium priority |
4.4 Machine learning and AI
Machine learning is advancing quickly in diagnostic imaging, predictive analytics and clinical decision support, with increasing regulatory approval for AI-based medical devices, though there is a real gap between claimed and demonstrated capability that has to be discounted for (Naudé 2019). Applications are imaging analysis with reported accuracy above 95% in radiology, early warning systems reducing patient deterioration by 20 to 30%, decision support integrated into electronic medical record workflow, natural language processing cutting documentation time by two to three hours per physician per day, and treatment personalisation from genomic and clinical data.
The limits are the reason this scores where it does. Data quality and bias, with documented disproportionate effect on minority populations (Gall 2019). Regulatory approval requiring extensive clinical validation. Clinical integration difficulty from workflow disruption and clinician resistance. Explainability problems that create trust issues in high-stakes decisions. And a talent shortage in healthcare AI where demand exceeds supply by a factor of three (Allen 2018).
| Dimension | Score and justification |
|---|---|
| Technical feasibility | 3/5, moderate complexity, dependent on data infrastructure |
| Financial | 3/5, significant investment, potential returns |
| Patient impact | 5/5, high potential for improved outcomes |
| Operational excellence | 4/5, enhanced decision-making capability |
| Organisational capability | 2/5, requires significant new expertise |
| Ethical considerations | 2/5, significant bias, privacy and accountability concerns |
| Weighted score | 3.4/5, medium priority |
4.5 IoT and big data analytics
IoT in healthcare is expanding through better sensors, better wireless connectivity and better analytics. Combined with big data analytics it supports real-time monitoring, predictive insight and population health management. Applications are remote patient monitoring through wearables, medical equipment asset tracking, environmental monitoring for infection control, building automation, and population health analytics for disease surveillance. The limits are the privacy exposure created by collecting at that volume, interoperability across diverse devices and protocols, data overload without adequate analytics, reliability for clinical-critical applications, and regulatory compliance.
| Dimension | Score and justification |
|---|---|
| Technical feasibility | 3/5, moderate complexity, network requirements |
| Financial | 4/5, reasonable costs, strong operational savings |
| Patient impact | 4/5, enhanced monitoring and preventive care |
| Operational excellence | 5/5, significant insight and automation |
| Organisational capability | 3/5, builds on IT capability, new skills needed |
| Ethical considerations | 3/5, privacy concerns balanced by health benefits |
| Weighted score | 3.8/5, medium-high priority |
5. Reading the results
AI scores top marks on the thing everyone cares about and still comes fourth. Machine learning ties with digital platforms at 5 out of 5 for patient impact, the heaviest-weighted dimension in the model. It then scores 2 for organisational capability and 2 for ethical considerations, the joint-lowest in the table on both. Weighted out it lands at 3.4, behind two technologies that are less impressive in every respect except deliverability. That is the framework doing its job.
The top of the table is a genuine tie, and pretending otherwise would be false precision. RPA and IoT both weight out to 3.8 and digital platforms sits at 3.7. Three technologies inside 0.1 of each other is a band, not a ranking, and treating a 0.1 gap as a decision is exactly the kind of spurious authority scoring frameworks are prone to. Reported as a tier, sequenced by what enables what.
Only one candidate is clearly separated. AR/VR at 2.8 is the single unambiguous result: high cost, immature technology, limited near-term operational impact. It scores 4 on ethics, the best in the table, and still comes last. Being harmless is not a reason to buy something.
| RPA | Digital platforms | AR/VR | ML and AI | IoT and big data | |
|---|---|---|---|---|---|
| Weighted score | 3.8 | 3.7 | 2.8 | 3.4 | 3.8 |
| Priority | Medium-high | Medium-high | Low-medium | Medium | Medium-high |
| Implementation timeline | 12-18 months | 2-3 years | 3-5 years | 2-3 years | 12-18 months |
| Primary applications | Administrative automation, claims processing | Patient engagement, telemedicine | Medical training, therapy | Diagnostic support, predictive analytics | Remote monitoring, asset tracking |
| Key success factors | Change management, process standardisation | User adoption, interoperability | Technology maturation, cost reduction | Ethics framework, expertise development | Security framework, data management |
| Expected return timeline | 18-24 months | 36-48 months | 60+ months | 48-60 months | 24-36 months |
6. Strategic recommendations
6.1 Prioritisation
Tier 1, immediate: RPA and IoT with big data, both at 3.8, strong operational benefit with manageable implementation complexity.
Tier 2, medium term: digital platforms at 3.7 and machine learning and AI at 3.4, both offering significant patient impact but requiring more organisational development first.
Tier 3, future consideration: AR/VR at 2.8, real potential but high barriers pending technology maturation and cost reduction.
6.2 Implementation roadmap
Phase 1, months 1 to 18. RPA pilots on administrative process, targeting a 30% reduction in manual data entry, with governance frameworks for automation established alongside. Concurrent IoT infrastructure for remote monitoring, starting with a 500-patient chronic disease pilot to validate the technical architecture and the clinical workflow before scale.
Phase 2, months 19 to 36. Digital platform deployment integrating telemedicine with existing electronic medical records, targeting a 25% increase in patient engagement scores. IoT scaled past 2,000 patients with predictive analytics for early intervention.
Phase 3, months 37 to 60. AI and machine learning for clinical decision support, focused on radiology assistance and sepsis prediction, plus predictive analytics for operational optimisation targeting 15% better resource utilisation and 20% fewer preventable readmissions.
The sequencing matters more than the total. AI arrives in Phase 3 into an organisation that by then has the data infrastructure Phase 2 built, the governance Phase 1 established, and the change-management experience of two prior rollouts.
6.3 Investment
$8 million across three years, split $2.5 million, $3.0 million and $2.5 million, against an expected five-year return of $15 to $20 million from operational efficiency, patient outcomes and service delivery. Every phase funds the capability the next phase depends on.
6.4 Risk mitigation
Change management with dedicated clinical champions and structured communication. Phased implementation so that each stage produces evidence for the next. Security frameworks addressing healthcare-specific threats including ransomware and data breach. Regulatory compliance planning with early engagement of the Therapeutic Goods Administration and privacy authorities rather than at submission. And continuing staff development, including digital literacy and specific AI ethics training, since the ethics score is a capability gap and not a fixed property of the technology.
7. Conclusion
Success requires balanced investment in technology, people and process, with change management and staff development weighted accordingly. The framework itself is reusable: as new candidates appear, they can be scored against the same six dimensions and the same weights, which is the part of this exercise that outlives the specific verdict on these five.
References
Allen, R. (2018). Business challenges with machine learning. Medium.
Chui, M., Harrysson, M., Manyika, J., Roberts, R., Chung, R., Nel, P., & van Heteren, A. (2018). Applying artificial intelligence for social good. McKinsey Global Institute.
Gall, R. (2019). Machine learning ethics: What you need to know and what you can do. Learning Hub.
Jordan, K. (2024a). Chatbot evaluation and comparison report: Healthcare applications. INTE2553 Digital Innovation Assessment 1, RMIT University.
Jordan, K. (2024b). Amazon Lex vs standalone chatbot platforms: A comparative analysis of platform-based and independent solutions for healthcare applications. INTE2553 Digital Innovation Assessment 2, RMIT University.
Madhavan, R. (2018). Natural language processing, current applications and future possibilities. Emerj AI Research.
Naudé, W. (2019). AI’s current hype and hysteria could set the technology back by decades. The Conversation.
Ozman, M. (2018). Can digital social innovations tackle big challenges? The Conversation.
Reavie, V. (2018). Do you know the difference between data analytics and AI machine learning? Forbes Agency Council.
Tuffley, D. (2019). Virtual tools, real fires: How holograms and other tech could help outsmart bushfires. The Conversation.