AI Copilot in the Doctor’s Office: myth, miracle or menace?

Introduction

The promise of AI in healthcare is often framed in terms of speed, scale, and diagnostic power. But some of its most transformative potential lies in something far more human: the conversation between patient and healthcare professional (HCP).

In simple terms, a copilot is an intelligent assistant, usually text or voice-based, that helps people perform a task more effectively. In the healthcare setting, that means:

AI copilots are emerging not to replace clinical judgement, but to reinforce it. These tools are designed to support structured, effective communication - helping both patients and HCPs prepare for, navigate, and follow up on consultations with greater clarity and confidence.

Crucially, AI copilots do not diagnose. Instead, they help ensure that key information is captured, expressed, and understood, which is especially important in rare diseases, where symptoms are often vague, multi-systemic, or easily dismissed.

So why are AI copilots gaining traction now? The timing is right: the technology is ready, healthcare systems are under pressure, and patients are more digitally engaged than ever. Copilots offer a practical way to make every conversation count: helping patients feel more clearly heard and supporting HCPs to communicate with greater clarity and empathy. For rare diseases in particular, where communication gaps can delay diagnosis or disrupt continuity of care, this mutual understanding can lead to more timely referrals, more confident decision-making, and ultimately, better outcomes.

Why rare disease communication needs an upgrade

Rare diseases often sit at the intersection of complexity and fragmentation. Patients may live for years without a diagnosis, navigating multiple specialists and repeating their story countless times – often without feeling truly understood or any closer to answers.

Guided conversations can help change that.

This matters, because in rare disease, symptoms are often non-specific, episodic, or span multiple systems - and when presented piecemeal, they can be difficult to interpret.

From the HCP’s side, structured inputs reduce reliance on memory and vague description. A well-designed copilot can prompt for relevant information, flag inconsistencies or patterns, and support a higher quality referral, which is particularly valuable when GPs, who are not typically trained in rare conditions, need to escalate to secondary or tertiary care.

These tools can also bridge silos. Many rare disease patients are under the care of multiple specialists, often across different institutions. Copilot-generated summaries that are consistently structured, shareable, and regularly updated, can support continuity and coordination by making key information accessible to the right people at the right time. Importantly, these summaries can also be tailored to their audience, whether that’s a patient needing clear, actionable next steps, a GP seeking a high-level overview, or a specialist requiring detail aligned to their clinical focus.

Importantly, structure doesn’t have to mean rigidity. The most effective copilots combine guided frameworks with space for natural expression, emotion, and nuance. In doing so, they can help both patient and HCP feel the conversation was not just efficient, but meaningful.

Empowering patients before, during and after consultation 

A key strength of AI copilots lies in their ability to support the patient journey at every stage - not just within the four walls of the consultation. For individuals living with rare diseases, this continuity can be critical.

Ultimately, the value here is not just in the data collected, but in how it supports the patient to become a more informed, confident advocate for their own health. In the complex world of rare disease, that advocacy can make all the difference.

Reinforcing clinical confidence in complex cases

AI copilots also offer valuable assistance to HCPs by helping them focus on what matters most, especially when time is limited and complexity is high.

Rare disease consultations are particularly challenging for healthcare professionals. Symptoms are often non-specific or overlap with more common conditions, and are compounded by lengthy, complex patient histories. For GPs in particular, this can make accurate assessment and referral even more difficult. AI copilots can help by generating structured summaries, offering adaptive prompts, and intelligently highlighting symptom patterns. In doing so, they support clearer documentation, more consistent consultations, and higher quality referrals – reducing the risk of delays, repetition, or miscommunication. 

For specialists, copilots can assist with continuity of care which is an increasingly critical challenge in rare disease management. Patients are often seen by multiple teams across different settings, with relevant information scattered across notes, systems, and siloed records.

This consolidated view is especially valuable in multidisciplinary environments, where coordination is key but often difficult to achieve in practice. In these settings, tools that enable HCPs to quickly access and interpret the full patient story aren’t just helpful - they’re essential to delivering joined-up, high-quality care.

And importantly, AI copilots are not just information organisers. When thoughtfully designed, they can actively support clinical reasoning – highlighting critical issues, identifying red flags, and prompting HCPs to explore inconsistencies or follow up on unclear details. This kind of “intelligent prompting” not only safeguards against cognitive overload, but also but also supports greater consistency in how symptoms are explored, and decisions are made, particularly for GPs managing rare or unfamiliar conditions. By guiding structured, high-quality conversations, copilots help ensure no important detail is missed, even when expertise may vary. The result? HCPs spend less time deciphering fragmented information and more time making confident, informed decisions. 

Emerging evidence reinforces this potential. A recent study found that a large language model could match, and in some cases outperform, physicians in diagnostic reasoning across a range of clinical scenarios.¹ The model performed particularly well in emergency settings and low-information situations, where rapid, high-quality decision-making is vital.¹ This suggests copilots may be especially valuable in high-pressure or uncertain environments - supporting HCPs while potentially helping to reduce the human and financial costs associated with diagnostic error, delayed care, and limited access.¹ As with any clinical tool, careful integration and robust testing through rigorous prospective trials will be essential, but the trajectory is promising.

Earning trust and designing for impact

AI copilots hold promise, but only if they’re designed and deployed responsibly. To gain trust, they must be clinically aligned, ethically built, and clearly positioned as support tools, not diagnostic systems. Their value lies in enhancing communication, not replacing clinical judgement. Copilots can flag patterns or prompt questions, but decision-making must always rest with trained professionals. Misrepresenting this balance risks regulatory breaches and undermines confidence. 

There’s also the question of bias and data quality.

Training data must be scientifically robust, inclusive, and clinically validated – particularly in rare diseases where symptoms vary and edge cases matter. 

Equally important is the risk of over-reliance. While copilots can enhance recall, structure and insight, there’s a danger they could deskill HCPs or over-amplify weak signals, which can lead to unnecessary referrals or anxiety. The solution lies in balance: letting AI do what it does best, while keeping the human in the loop for empathy, judgement, and discretion.

We’ve seen what happens when digital tools enter the consultation room without clear boundaries or clinical relevance. The rise of “Dr Google” has, at times, strained the trust between patients and HCPs - whether it’s the frustration clinicians feel when faced with inaccurate self-diagnoses, or the discomfort patients experience when their HCP appears to rely on generic search results. With AI, we have the opportunity to take a different path: one grounded in clinical purpose, clarity of role, and mutual trust. Copilots must be introduced as thoughtfully designed tools that support, not substitute, the expertise and empathy that define good care.

And finally, the patient experience must never be an afterthought. Systems that feel cold, robotic, or overly scripted can undermine trust. Copilots should be designed with empathy in mind - validated with patient groups, endorsed by HCPs, and grounded in transparency about what they can (and can’t) do. Done well, they can even enhance patient experience. It’s worth noting that patients may interact differently with AI – perhaps more honestly, openly, or without fear of judgement – creating richer, more accurate conversations that ultimately support better care. 

Better conversations. Better Care.

AI copilots strengthen the connection between clinical judgment and patient advocacy. By helping structure conversations, surface critical details, and ensure clarity at every stage of the consultation, copilots can reduce friction, improve understanding, and foster a stronger partnership between patients and HCPs. In rare disease, where journeys are long, symptoms are complex, and emotional fatigue is high, this support is more than useful. It’s necessary. 

But building copilots that work isn’t just about capability. It’s about purposeful design. The best tools reflect real needs, which means defining the right problem, engaging real users early, and designing for clinical accuracy, emotional nuance, and inclusivity. Done right, copilots don’t just capture information - they help build confidence, continuity and connection, making every conversation count.

If you’re exploring how AI copilots could support patient-HCP conversations in your therapy area, we’d love to help you shape something that’s clinically sound, human-centred, and built to make a real difference. Get in touch to discuss how we can help you scope, design, and deliver a copilot solution tailored to your brand, audience, and your clinical objectives. 

If you would like to discuss the content of this article or a potential AI project then please do not hesitate to contact Claire: claire.dobbs@solrishealth.com

Disclaimer 
We use AI to push creative boundaries, deliver impactful work and improve efficiency. We are committed to transparency around AI use and will always use it in line with the terms of client contracts. We use pre-vetted AI tools, audited under the Mission AI Acceptable Use and Guidance Policy, to meet our intellectual property, data protection and client confidentiality standards. If any work is in part generated by AI, we will let you know and advise you of any limitations in fidelity, resolution, adaptability, reproducibility or licensing.

References: 
1. Brodeur, PG, et al. Superhuman performance of a large language model on the reasoning tasks of a physician. (2024).  arXiv preprint arXiv:2412.10849. https://doi.org/10.48550/arXiv.2412.10849