AI fatigue is real. Social media posts (including LinkedIn) are riddled with long, drawn out narratives talking about a new product or course and trying to start discussion. I don’t know what they say, I don’t read them. Maybe my attention span isn’t what it used to be, or maybe deep down I know there’s no substance and I won’t learn anything if I read it.

This feels like a step backwards. AI is supposed to make our lives easier, but sometimes the noise can be overwhelming. How can we prevent AI fatigue? By making it meaningful. If you are intentional with what you post, more people will enjoy it. Focus more on what you want your audience to know, don’t just prompt ChatGPT with your topic for the week and copy/paste without editing to make it meaningful.

Remember, AI is a tool. We need to make sure we use it properly. Stay tuned to the end for practical tips on creating quality posts without increasing AI fatigue.

Here’s what’s happening this week in AI

Plus: Tool of the week

Let’s dive in.

🏥 AI scribes may be increasing healthcare costs: Early data suggests AI documentation tools are actually raising costs in the short term, even as they improve efficiency. Insurers are already signaling they may adjust reimbursement downward as adoption grows.
Why this Matters: Efficiency gains do not automatically translate into savings, especially in fee for service systems.

🧠 AI chatbot prescribing psychiatric meds: Utah just launched a pilot where an AI chatbot can renew prescriptions for common psychiatric medications without direct physician oversight. It is limited to low risk refills, but still, this is a big step into clinical autonomy.
Why this Matters: This forces us to define where AI can safely replace routine clinical decisions.

🫀 AI predicts heart failure years early: A new model out of Oxford can detect heart failure risk up to five years before symptoms using subtle inflammation signals in cardiac CT scans. It was trained on over 70,000 patients and hits about 86 percent accuracy, which is impressive for something we currently miss until late.
Why this Matters: Earlier detection here is not incremental, it fundamentally shifts when we intervene.

💊 AI suggests lower cost prescriptions: A new system called EcoRxAgent generates clinically appropriate but cheaper medication alternatives by combining safety checks with cost effectiveness analysis. It essentially creates multiple “equivalent” prescriptions ranked by affordability.
Key Takeaway: AI is starting to optimize not just clinical outcomes, but financial ones too.

🧪 Call for standardized evaluation of clinical AI: Experts emphasize the lack of consistent benchmarks and validation pathways for AI tools in medicine.
Why this Matters: Without standards, adoption will remain uneven and potentially unsafe.

💸 Financial incentives around AI under scrutiny: Researchers highlight growing financial relationships between AI vendors and clinicians, raising transparency concerns.
Why this Matters: Trust in AI tools will depend on clear conflict of interest policies.

📋 FDA expands oversight of adaptive AI systems: The FDA signaled plans to tighten oversight on continuously learning AI models, with new guidance expected around lifecycle monitoring and post deployment performance tracking. The focus is shifting from one time approval to ongoing surveillance.
Why this Matters: Clinical AI will no longer be a “set it and forget it” tool, continuous validation will become part of routine practice.

LLMs are there to help you improve your writing, not do it for you. We already know that relying too much on LLMs reduces intelligence.

Try these useful tips to get started:

Whatever you’re posting about, if you start with the end in mind, you’re already on your way to being successful.

The biggest shift this week is not better models, it is where they are being used. New information is coming to light that AI is not cost effective if used improperly. Screening, prescribing, and cost optimization are all moving into real workflows, but can they reduce the $5.3 trillion in healthcare spending every year in the US?

If there is one takeaway, AI is not going to make anything easier, unless we are intentional about it’s use. That is where practice will actually evolve.

If this sparked a thought, send this newsletter to a colleague or reply with what you are seeing locally.