The Bandwidth: Issue 11

A sharper perspective on this week's top issues that matter at the intersection of health and AI.

Algorithmic pricing on the rise, but maybe not in healthcare

Last week, I focused on the bipartisan alarm over AI chatbots and the risks they pose to vulnerable users.

Another concern is percolating in a different corner of the AI debate, and it has nothing to do with mental health or medical advice. It’s about algorithmic pricing.

This isn’t necessarily a new issue. The FTC dove into it back in 2024. But New York is the first state to regulate AI-driven personalized pricing. The law targets practices that let companies adjust prices based on what they know—or can infer—about you: your browsing behavior, income proxies, location, past purchases. It’s not surge pricing; it’s surveillance-priced goods and services.

On paper, this practice seems aimed at retail consumption. But it could have broader implications as lawmakers are waking up to the idea that AI can quietly reorganize markets long before anyone notices. Algorithmic pricing can operate in the background and can be impossible for a consumer to detect. New York’s move is essentially an attempt to drag that process into the light.

So what does this mean for healthcare?

My instinctive answer is: probably not much. For better or worse, healthcare is not a true market. It violates many core assumptions of market economics. As we all know, healthcare is highly regulated, starting with the fact that we have long outlawed differential pricing based on health status. That principle runs through the ACA’s community rating rules and protections for people with preexisting conditions.

Fee schedules also probably provide some buffer. Algorithmic pricing is built on fluidity, while healthcare pricing is built on fixed rates that are highly scrutinized (most of the time).

Against that backdrop, it’s hard to imagine any political or policy appetite for a world where your ER copay changes based on your “willingness to pay” score. Lawmakers are unwilling to agree on allowing AI in prior authorization (more on that below). They’re not going to casually look past personalized pricing for chemo.

But that doesn’t mean the underlying dynamics can’t seep in. As more healthcare activity shifts to direct-to-consumer channels (think GoodRx, CostPlus, the emerging TrumpRx platform, DTC labs, etc.), AI could look for the same optimization levers it uses in retail.

To sum up: Algorithmic pricing for essential healthcare services isn’t happening. Policy and political reality won’t allow it. Elective and consumer-driven care is a different story. Cash-pay sectors like concierge primary care already operate on retail logic. If algorithmic pricing shows up in healthcare, it’ll be there first.

🛜 Other Field Signals

A quick hit on this week’s key policy shifts and industry trends.

HHS releases a new AI strategy to modernize agency-wide operations
HHS unveiled a sweeping AI strategy focused on accelerating research, improving internal operations, and creating stronger safety and transparency guardrails across its agencies.

FDA expands its internal AI capabilities with new agentic-AI systems
The agency is deploying agentic tools to speed reviews, automate internal workflows, and support scientific analysis. And here’s a second article from STAT News on the topic. (My commentary: I think the general take is, let’s hope this goes better than the first AI rollout at the agency.)

OpenAI’s infrastructure partners rack up $96B in debt to fuel mega–data center buildout
A new $38B loan package brings total leverage to $96B as OpenAI-aligned builders race to meet frontier compute needs. Rising capital pressure could eventually push model pricing upward.

Can AI rescue U.S. healthcare from the gathering storm?
AI Magazine argues the system is heading toward structural failure as workforce shortages and rising costs collide — and AI may be the only scalable relief valve left.

Three regulatory trends shaping healthcare AI
Forbes highlights rising pressure around risk-based oversight, transparency, and data-privacy alignment. The regulatory climate is shifting faster than many health systems expect.

DeepSeek drops two near-frontier models rivaling GPT-5 and Gemini 3 — at bargain prices
China’s DeepSeek released two open-source models that match top-tier performance at dramatically lower cost, intensifying geopolitical tech tensions. Frontier capability has never looked cheaper. Not to be outdone, Amazon announced this week a new family of frontier artificial intelligence models—and a new way for customers to build frontier models of their own.

OpenAI declares “code red” as Google gains ground
Altman is delaying advertising, health agents, and the “Pulse” assistant to focus fully on closing the performance gap with Gemini 3. A rare moment where OpenAI is visibly on its back foot.

Health plans face worsening headwinds — and are turning to AI
A new HealthEdge survey shows insurers squeezed by utilization, regulation, and admin costs, with many betting on AI to rebalance operations despite uneven readiness.

Tampa General rolls out AI voice agents across patient access
The system and startup Hyro launched AI voice agents for scheduling, triage, and call-center traffic, reducing operational strain while introducing new governance challenges.

House bill would prohibit AI models from denying Medicare procedures
New legislation from Reps. Landsman and Watson Coleman would bar Medicare contractors from using AI for coverage denials — a direct response to rising concerns about automated UM. (My Commentary: Per my flag above, this seems to be one area of bipartisan agreement, especially in healthcare - ensuring AI does not run roughshod over clinical decisions).

AI tools show both promise and pitfalls in dementia care
Early models improve detection, monitoring, and caregiver support but remain error-prone and challenging for older adults to use safely.

Insiders say the future of AI agents will be smaller, cheaper, and more specialized
Fortune reports the next generation of agents will move away from massive frontier models and toward lightweight, workflow-native systems that run locally. (My commentary: If this pans out, it could assuage users who have ethical concerns about energy usage and environmental/community impacts.)

Patients are pasting their medical records into chatbots — and doctors are worried
Continuing the coverage regarding concerns over medical chatbots, the New York Times reports patients are feeding full medical histories into public chatbots for second opinions, raising accuracy, privacy, and safety concerns as clinicians see early-warning signs of misdiagnosis and misinformation.

🛠️ Practical Edge

Actionable tips, tools, and thoughts to help leaders strengthen capacity and apply AI in their work.

Accenture x OpenAI Enterprise Partnership
Accenture is rolling out ChatGPT Enterprise to tens of thousands of consultants and launching a program to help clients deploy AI agents at scale. This marks a shift from “pilot mode” to industrialized AI service delivery across major consultancies.

What’s Lost When We Work with AI, According to Neuroscience
In this HBR article, Neuroscientist David Rock warns that over-relying on AI agents and meeting summaries erodes attention, deep thinking, and insight formation — the cognitive processes leaders actually need. Efficiency gains mean nothing if they hollow out judgment.

How to Be “Team Human” in the Digital Future
Author and documentarian, Douglas Rushkoff, argues that the real risk in a tech-driven world isn’t AI taking over, but humans outsourcing so much agency and connection that we forget how to think, collaborate, and build community. His case is simple: technology should amplify human values, not replace them — and leaders need to design systems that pull people in, not push them out.

Using Claude to Help a Family Navigate Assisted Living Decisions
One user of The Rundown shared how they used Claude to build a tailored assessment guide for their 88-year-old mother choosing an assisted-living facility, including criteria and a structured checklist. A good reminder that the most meaningful AI use cases are often deeply human.

🌅 On the Horizon

A quick look at the developments and events expected to shape the weeks ahead.

And finally, if you like what you are reading, please share and encourage your network to subscribe.

Till next time,

BC