Artificial intelligence already has changed anaesthesia. Before we called it AI, we called it machine learning, and before that it was simply statistics. These lie on a continuum of methods that allow us to offload thinking. We use their products daily in the form of risk calculators, pharmacokinetic models, and waveform analysis.

However, there are fundamental differences with the artificially intelligent tools of the future. For decades it was thought that if only we could layer enough rules, or heuristics, we could create tools that could cope with the totality of clinical complexity. We could sit down and devise a tool to manage intraoperative hypotension, coming up with a decision tree to cover every possible scenario, but we’d soon end up in the minutiae and we’d still miss edge cases.

Instead, we now have connectionist models, popularised by large language models like ChatGPT, and based on mathematical representations of neural networks. These are masters of association and pattern recognition, so instead of writing heuristics for every possible situation, the model uses its training to react intelligently to the task at hand.

The catch – predictability. Whereas we can see exactly how a heuristic model reaches its conclusion, we cannot see into the heart of a connectionist model to divine its workings. One you can audit, the other you must supervise.

Just as the human anaesthetist relies on their experiences, so too would the artificially intelligent anaesthetist, but with the key difference of scale. A consultant may build a 40-year repertoire of thousands of cases. Your AI co-anaesthetist could absorb millions. Learning and experiencing from every anaesthetic given around the world, only instead of sight, sound, and touch, its senses would be ECG, pressure waveforms, and gas analysis.

You would have a domain expert on your shoulder. One that has read every review article, learned from every anaesthetic, and capable of responding to queries, supplying suggestions, and predicting likely problems before they’ve happened.

The cacophony of alarms will fade, with the AI alerting you intelligently only when necessary intervention is required. So too, would it be able to consume and summarise your patient’s medical history far more efficiently than you or I, recommending individual adaptations specific for that patient.

However, what then happens to us, the human anaesthetist? Once we’ve offloaded so much of our cognitive load, what do we become? Technicians implementing the will of our AI supervisors?

For years, we have assumed that it’s our capacity for reason that would make us irreplaceable, but just as AI models are knocking on the door of near-human thought, we have failed at making anything resembling the complexity and dexterity of the human body.

We are also the greener option. The human brain runs at approximately 12 Watts, whereas current models require orders of magnitude more power to process a fraction of what we can in a split second.

We are efficient and adaptable machines, capable of interacting and testing our surroundings to an extent we nowhere near able to replicate artificially.

Finally, and most importantly, we are human and our patients are human. We bring the human touch; we build trust, we hold the responsibility, and we keep the best interests of our patients to heart.

So where does that leave us? Not as technicians bowing to a black box, but as conductors of a larger orchestra. The AI can watch every waveform, recall every similar patient, and suggest the next best move – yet only we can read the tremor in a patient’s voice, reassure a frightened parent, or decide that today the safest option is to postpone.

AI will amplify our vigilance, but cannot replace our empathy and judgment. Our task is to learn how to use this new instrument responsibly and ethically, while ensuring it remains patient-centred. If we do, the theatres of the future will be safer, calmer, and more humane than ever before. The future of anaesthesia is what it always has been – a partnership between us and technology, with the baton firmly in our hands.

These are my speaking notes for a talk on the role of AI in Anaesthesia that I gave at the Association of Anaesthetists Resident Doctor Conference in June 2025, which hopefully explains some of the rhetorical style. I largely still stand by most of the sentiments expressed. I do think our practice will be changed fundamentally by the increasing pervasiveness of machine learning and artifical intelligence in our lives. Though just as some anaesthetists today look back nostaligcally before the complications of TIVA, BIS, and quatitive ToF, we may someday look back with fondness to a time before our AI assistants.