The future of AI for doctors: what could the next decade bring?

If it feels like artificial intelligence has suddenly arrived in healthcare, Professor Enrico Coiera, Director of the Centre for Health Informatics at Macquarie University's Australian Institute of Health Innovation, would disagree.

“I think even the framing that this is the beginning is wrong,” says Coiera, who has worked in health and AI for more than 35 years.

AI has been used in healthcare since the 1990s, he says, including in ECG interpretation, medical imaging and labs.

What has changed is how and where doctors encounter it.

“Until now, it's all been one step back from the workflow,” says Coiera. “We're in the acceleration now. The great acceleration.”

How AI could change everyday clinical practice 

For many doctors, it is the ambient scribe that has pushed the AI conversation into the mainstream, with more than 40% of Australian GPs now using one in their practice. 

“I think of the scribe as a gateway drug to other kinds of AI,” Coiera continues. “But they’re the least exciting part. We already have apps where you can have a look at a mole and diagnose it. There'll be more of that sort of technology.”

Coiera points to AI-enabled ultrasound as one example, where AI could guide someone without specialist ultrasound training to capture the right image.

“You can have the AI basically saying, ‘Left a bit. Right a bit. Up a bit. Down a bit. That's the liver,’” he says.

He also sees potential for AI and point-of-care testing to make clinical trials, testing, and treatment more accessible remotely, reducing the need for regional and remote Australians to travel long distances.

CSIRO's 2026 AI Trends for Healthcare report points to similar possibilities, such as the development of an AI capable of generating radiology reports from chest X-rays.

But Dr David Hansen, CEO and Research Director of the Australian e-Health Research Centre at CSIRO, says some of AI's most useful applications could also be much simpler.

For a time-poor GP, that could mean using AI to search years of medical history and surface information relevant to the patient sitting in front of them. It could also help doctors navigate lengthy clinical guidelines.

“These clinical guidelines are huge,” he says. “Being able to tailor a summary to the patient in front of the doctor is potentially really powerful.”

Hansen also hopes AI could take some of the load off doctors, allowing them to work at the top of their practice and spend more time with patients. But he cautions that removing the mundane parts of a job isn't necessarily an unqualified good.

“I've got software engineers who use them a lot, and they go, ‘I'm exhausted because I'm not doing any of the mundane stuff anymore. I'm just thinking all the time,’” he says. “I think it's the same with doctors. We need to make sure doctors don't lose the skills they need.”

The skills doctors will need in an AI-enabled future 

The question of retaining skills is part of what Hansen describes as building “AI maturity”. The ability not only to understand the tools being used, but also developing the judgement to question them: “Is this the right thing? Does this feel right?”

Sometimes, the answer may be that AI isn't needed at all. 

“Imagine a GP practice wanting to identify how many of its patients have diabetes. If the records are properly structured and coded, the answer can be found quickly and cheaply, rather than using AI to search through every patient record,” he explains. “Just because we can do something with AI doesn't mean we should.”

Professor Coiera has a similar warning about “vibe coding”, where people use generative AI to build software without necessarily having the technical expertise to assess what they have created.

“You wouldn't go home and cook up a new drug and say, ‘I've tested it on myself and it seems to work,’” he says. “But equally they might have a fantastic idea that needs to be done well.” 

According to Hansen, some of the earliest GP medical record systems were developed by doctors. What has changed is how accessible technology has become, and he is already seeing doctors seek help to take their ideas further.

“Now those same doctors are turning up saying, ‘I've built this little tool using ChatGPT. How do I actually turn this into something real?’” he says.

Safety first when it comes to AI in healthcare 

However AI is being used, Coiera expects doctors will also need to demonstrate that they can use AI safely.

“I think one of the big drivers will be accreditation,” he says. “You'll have to show safety and quality in what you're doing, and demonstrate that you're a safe and effective user of the technologies you use.”

While doctors can already seek out short courses, he believes colleges will have an important role in ensuring training reflects what clinicians actually need. And being a safe user will require more than technical competence.

Automation bias is one risk. A doctor using an ambient scribe might initially read and correct every document carefully. “After a week they'll say, ‘It seems to be pretty good,’ so they'll just start to skim it,” explains Coiera. “After three months, they're just clicking accept.”

The level of scrutiny, he says, should reflect what is at stake, with clinical decisions around tests or treatments requiring far greater oversight than routine administrative tasks.

But responsibility doesn't sit with doctors alone. With “technology pouring out the funnel faster than we can evaluate”, Coiera argues that AI used in healthcare needs to be medical grade, and backed by evidence of its quality and effectiveness.

Australia’s regulatory frameworks are already beginning to catch up. In February 2026, the TGA published updated guidance clarifying how existing medical device regulations apply to AI-enabled software. The third National Policy Roadmap for AI in Healthcare also points to progress, including new Ahpra guidance for health professionals.

When patients bring their own AI

The next decade will not only change how doctors use AI, but how patients use it too. 

Hansen expects patients to arrive at consultations having already discussed their health with AI. Rather than dismissing those conversations, he sees an opportunity to build on them.

“Hopefully that's a process whereby we are increasing people's health literacy,” he says.

Coiera expects patients' use of AI to go even further, envisaging patients having AI agents that know their health information and can act on their behalf, potentially handling tasks such as requesting repeat prescriptions. 

He cautions against assuming today's technological limits will last. “Whenever people say, ‘Oh, it can't do X,’ I always reply, ‘For now,’” he says.

The future of AI for doctors: key takeaways

• AI isn't new to healthcare, but its use in everyday clinical practice is accelerating
• AI could reduce doctors' administrative and information burden, but clinicians will need to judge when it genuinely adds value
• AI literacy, safe use and appropriate human oversight will become increasingly important
• Patients' own use of AI is likely to change how they engage with doctors