Artificial intelligence is being used across healthcare to support clinical work, streamline administration and help organisations make better use of data.
AI is a broad term covering different technologies and purposes. Two of the most commonly discussed types are generative AI and predictive AI.
Understanding the difference between them can help healthcare organisations assess where each approach may add value, what risks need to be considered and how the technology should be governed.
Generative AI creates new content based on patterns found within existing data. This can include text, images, audio, code and other forms of information.
In healthcare, generative AI may be used to summarise lengthy documents, produce draft correspondence, support clinical documentation, create synthetic data for research or assist with training and education.
One increasingly visible use case is ambient clinical documentation. These systems can listen to a consultation and produce a draft clinical note or summary for review by the healthcare professional.
This can reduce time spent on administrative tasks, but generated content still needs appropriate human oversight. Generative AI can produce inaccurate, incomplete or misleading outputs, so clinicians must be able to check and correct the information before it is relied upon.
Predictive AI has a different purpose. It analyses existing and historical data to estimate what may happen in the future.
Rather than creating new content, it identifies patterns and calculates the likelihood of particular events or outcomes.
Potential applications include identifying patients who may be at increased risk of deterioration, forecasting demand for beds or clinical services, predicting hospital admissions, supporting preventative care and helping organisations plan staffing and resources.
For example, a predictive system may analyse observations, medical history and other available information to highlight patients whose condition could be worsening.
These systems can support earlier attention and decision-making, but their accuracy depends heavily on the quality and relevance of the data used to develop them.
The main difference is the purpose of the technology.
Generative AI creates content. It may produce a summary, draft document, image or suggested response based on the information it receives.
Predictive AI estimates outcomes. It uses patterns within existing data to indicate what may happen or which patients, services or events may require attention.
In practice, healthcare technology may combine both approaches. A predictive model might identify a potential risk, while a generative tool presents the information in a clear summary for a member of staff.
| Generative AI | Predictive AI |
|---|---|
| Creates new content | Forecasts possible outcomes |
| Commonly works with text, images or audio | Commonly works with structured and historical data |
| Can support documentation, summaries and research | Can support risk identification, planning and forecasting |
| Outputs need checking for accuracy | Predictions need monitoring for reliability and bias |
Both approaches have the potential to reduce administrative pressure, improve access to information and support more informed decisions. They also introduce important risks.
Generative AI may produce convincing information that is inaccurate or unsupported. Predictive systems may generate unreliable results if they are trained on poor-quality, incomplete or unrepresentative data.
Healthcare organisations must also consider patient confidentiality, information governance, cybersecurity, bias, clinical oversight and whether the technology falls within medical device regulation.
AI should support professional judgement rather than replace accountability for clinical decisions.
The right type of AI depends on the problem an organisation is trying to solve.
Generative AI may be suitable where staff need help creating, organising or summarising information. Predictive AI may be more appropriate where the objective is to identify risk, anticipate demand or support planning.
Before adopting either, healthcare organisations should define the intended purpose, understand the quality of the underlying data and establish clear governance around how outputs will be reviewed and used.
The value of AI in healthcare will depend on more than the sophistication of the technology. Successful adoption requires safe implementation, meaningful clinical involvement and a clear focus on the needs of patients and staff.
