Healthcare exec + AI builder. Startup to Fortune 20. I build at scale, not hype. UCLA Anderson professor. Fluent in payer, provider & tech. Opinions my own.
Los Angeles
Joined June 2008
- Tweets3K
- Following500
- Followers701
- Likes394
For those headed to @sxsw, this tool is pretty great. (Also, the @claudeai character in a cowboy hat is just perfect.)
SXSW is amazing - but there are 4k+ events! That's a lot to sort through and there needed to be a better way -> introducing SouthyWesty.
southywesty.ai
- All the official and unofficial SX events in one place, searchable with natural language.
- It will build a personalized schedule for you.
- Download events direct to your calendar of choice.
- Claude Code inspired (there are / commands)
I hope you find it helpful. #sxsw, #sxsw2026
These are adapted from Rams' originals with two additions for an era where products think, generate, and act.
I wrote the full version with more context for each principle here: linkedin.com/pulse/dieter-ra…
Rams himself said it best: "each generation should review [these principles] and, if necessary, modify or supplement them."
This is my attempt.
12 — Good AI is as little AI as possible.
A demo shows off everything a model can do. A product is disciplined about what it *should* do. Strive for purity. Strive for simplicity.
Less, but better.
11 — Good AI is accountable. (New)
AI that acts on your behalf (renews prescriptions, submits claims, writes clinical notes) must make those actions visible, reversible, and attributable. Autonomy without accountability is recklessness.
10 — Good AI respects human agency. (New)
AI should extend what a person can do, not decide what a person should do. Clinical decision support nudges providers toward evidence-based care, but transparently, with the provider retaining authority. The line is between informed guidance and covert influence.
09 — Good AI is environmentally friendly.
AI carries real computational and energy costs. Does the task require a large model or a small one? Does the query need to be made at all? Efficiency is not just an engineering problem. It is a design obligation.
08 — Good AI is consistent down to the last detail.
The product should respond to similar inputs in reliably similar ways. Uncontrolled variability in a tool is not "creativity." It is a defect. Consistency is the foundation of security and trust.
07 — Good AI is enduring.
The interaction patterns, the trust relationship, the core utility. These should be designed to last, even as the underlying technology evolves rapidly beneath them.
06 — Good AI is honest.
It must not pretend to know what it doesn't know. It must not simulate confidence it hasn't earned. It must not present generated content as verified fact. Honesty is not a feature. It is a foundation.
05 — Good AI is unobtrusive.
The best AI disappears into the workflow. Products that constantly remind you they're "AI-powered" have confused marketing with design.
04 — Good AI makes a product comprehensible.
The user should always understand what the system did and why. A product that produces useful outputs but can't explain itself is a black box dressed up as a tool.
03 — Good AI is aesthetic.
The quality of language, the rhythm of an interaction, the feel of a response at just the right level of detail. These are the aesthetics of intelligence. They matter as much as the interface.
02 — Good AI makes a product useful.
It should be useful first, impressive second. If a feature exists only to show what AI *can* do rather than what the user *needs* done, it fails this test.
01 — Good AI is innovative.
But novelty alone is not innovation. A new capability is only good AI if it solves a real problem in a way that wasn't previously possible.
Dieter Rams wrote ten principles of good design in 1976. They're almost perfect for AI. Almost.
I swapped "Good Design" for "Good AI" in his original principles and changed as little else as possible. Added two new ones for problems he never had to think about.
A thread: 12 Principles of Good AI.
(Me, 12 years ago, trying and failing to keep it extremely cool in front of my design hero.)
My thanks to HIMSS & ScaleLA for hosting last night’s discussion on the future of #virtualhealth. Enjoyed sharing how CareMore Health is using virtual care to enable advanced care delivery models and convenience for the most complex and vulnerable patien…lnkd.in/gqk9P2c
Great start to the day at UCLA facilitating a workshop on #healthcare payment models and their role in #digitalhealth adoption for the clinicians, engineers and MBAs in the inaugural Biodesign Program.
Congrats to Jennifer McCaney and Desert Horse-Gran…lnkd.in/e5ArE9w
A new report from @LEK_Consulting shows that as seniors seek
better management of their care at a lower out-of-pocket cost, Medicare Advantage penetration will continue to rise — to as high as 70% between 2030 and 2040.
lek.com/sites/default/files/…
Hoping we inspired more tech innovators to focus their efforts on serving the chronically ill!
Thank you to The Cedars-Sinai Accelerator and Health 2.0 Los Angeles for hosting us. lnkd.in/dabUFqF
Hoping we inspired more tech innovators to focus their efforts on serving the chronically ill!
Delighted to host #health20LA @dodigitalhealth and panel guests @fiveplusfour of @CareMoreHealth, Adam Kaufman of @CanaryHealth & Dr Mona Patel of @ChildrensLA tonight as they share their insights on successes and challenges in managing chronic disease populations!
