“We need AI capability.” I hear this from healthcare executives regularly. But when I ask what that means—what roles they need, what skills they’re looking for, how the team would be structured—the answer is often uncertain.

Building internal AI capability is essential for sustainable AI adoption. Relying entirely on vendors creates dependency. Here’s how I think about building healthcare AI teams.

What Capability Do You Actually Need?

Before defining roles, clarify what you need the team to do. Different objectives require different capabilities:

Implementing vendor AI solutions. If your strategy is buying and deploying commercial AI products, you need:

• Evaluation and procurement expertise
• Integration and implementation capability
• Clinical governance and monitoring
• Change management

You don’t necessarily need deep AI technical skills—the vendor provides those.

Developing custom AI applications. If you want to build AI using your own data, you need:

• Data science and machine learning expertise
• Data engineering capability
• Clinical domain expertise
• Validation and regulatory knowledge

This is a different (and larger) team than implementation alone.

Both. Most organisations need some of both. The mix determines team composition.

Core Roles

Clinical Informatics Lead. This is arguably the most important role. Someone who understands both clinical practice and information technology deeply. They bridge between AI capabilities and clinical needs, ensuring technology serves clinical purposes.

In Australia, health informaticists often come from clinical backgrounds (nursing, medicine, allied health) with additional qualifications in informatics. This dual perspective is invaluable.

Data Scientist / ML Engineer. If you’re developing custom AI, you need technical AI expertise. Data scientists build models. ML engineers deploy and maintain them in production.

For healthcare AI, look for people with health data experience. The nuances of clinical data—missingness, coding inconsistency, temporal patterns—require specific knowledge.

Data Engineer. AI runs on data. Data engineers build the pipelines that get data from clinical systems into formats AI can use. This is often underestimated—data engineering is typically 70% of AI project effort.

Project/Program Manager. AI initiatives are complex projects requiring coordination across clinical, technical, and governance stakeholders. Strong program management keeps everything on track.

Clinical Champion(s). Clinicians from relevant specialties who advocate for AI adoption, provide clinical guidance on development, and lead change management with their peers.

Governance/Quality Lead. Someone responsible for clinical AI governance—monitoring performance, managing incidents, ensuring regulatory compliance. This might be combined with broader clinical quality roles.

Team Structure Options

Centralised model. A single AI team serving the whole organisation. This concentrates expertise and ensures consistency, but can create bottlenecks and distance from clinical areas.

Distributed model. AI capability embedded in clinical divisions or departments. This provides closer clinical alignment but risks inconsistency and duplicated effort.

Hub and spoke. A central team providing core capabilities (data science, governance, infrastructure) with distributed clinical informatics roles in clinical areas. This tries to get benefits of both but requires careful coordination.

Most large health services I work with are moving toward hub and spoke models. Central expertise with local clinical connection.

Hiring Challenges

Healthcare AI roles are competitive. Big tech companies, consulting firms, and AI vendors compete for the same talent. Healthcare organisations often struggle to match salaries.

Strategies that help:

Emphasise mission. Many AI professionals want work that matters. Healthcare offers meaningful impact that financial services or advertising AI can’t match.

Offer flexibility. Remote work options, flexible hours, and work-life balance can compensate for salary gaps.

Create development paths. Clear career progression keeps talent engaged. If AI team members see their future at the organisation, they stay.

Build from clinical staff. Healthcare staff with interest and aptitude can develop AI skills. This creates AI capability that understands clinical context deeply.

Partner effectively. Where you can’t hire, partner with firms that have the expertise. AI consultants Melbourne and similar organisations can provide capability while you build your own. The key is treating partnerships as capability-building, not just project delivery.

Skills to Prioritise

Beyond role-specific skills, some capabilities matter across the team:

Communication across boundaries. AI teams need people who can explain technical concepts to clinicians and clinical concepts to engineers. Translation skills are essential.

Pragmatism. Healthcare AI is messy. Data is imperfect. Systems don’t integrate cleanly. The best AI team members are pragmatic problem-solvers, not perfectionists who need ideal conditions.

Ethics awareness. Healthcare AI involves decisions that affect patient wellbeing. Team members need ethical awareness and willingness to raise concerns.

Regulatory understanding. TGA, privacy law, professional standards—healthcare AI operates in a regulated environment. Understanding regulatory constraints (or knowing when to seek guidance) matters.

Common Mistakes

Hiring data scientists without clinical context. Technical AI skills without healthcare understanding leads to solutions that don’t fit clinical practice. Ensure clinical informatics expertise is part of the team.

Underinvesting in data engineering. Organisations hire data scientists, then discover they spend 80% of time on data preparation because there’s no data engineering support. Build data engineering first.

No clinical governance ownership. AI governance is no one’s job specifically, so it doesn’t happen. Designate someone responsible.

Expecting immediate ROI. Building AI capability takes time. Teams need runway to develop and learn before delivering major outcomes. Short-term ROI pressure undermines long-term success.

Isolating the AI team. AI teams that operate separately from clinical and IT operations struggle to deliver impact. Integration with existing structures matters.

Starting Small

You don’t need a large team to begin. A minimal viable AI capability might include:

• One clinical informatics lead (could be partial FTE)
• One data scientist/engineer (or contracted)
• Clinical governance integrated into existing quality structures
• Access to data science partnership for larger projects

This is enough to implement commercial AI solutions, run small pilots, and build organisational learning.

Expand as AI adoption grows and as you develop clearer views on what capability you need.

The Long View

AI capability isn’t a one-time build. Technology evolves. Applications expand. What you need in five years will differ from what you need today.

Design your team and structure with adaptability in mind. Hire people who learn quickly. Build processes that can scale. Create governance that evolves with technology.

The organisations that will succeed with healthcare AI aren’t necessarily those that move fastest. They’re those that build capability thoughtfully and sustain it over time.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.