Are you waiting to implement AI in your practice or health system until it gets ‘good enough’?
I know a lot of physicians and practices who say that healthcare AI just isn’t good enough yet to bother with:
It’s not reliable for clinical diagnosis
It still requires a human to oversee it
It’s just hype; most of the companies claiming to use AI are just recycling what they were already using anyway
It will get better
There’s no rush.
Are these people being smart to wait, or are they going to regret not getting on board the rocketship of AI?
The more interesting question is: How will they know when the right time is to incorporate AI?
The Wait Calculation
The concept of the “Wait Calculation” was showcased in a recent article by Ethan Mollick at Wharton. The Wait Calculation is based on a paper about interstellar travel, of all things. The Wait Calculation is an actual equation to predict when civilization should start leaving Earth for other planets based on how quickly technology is improving. For example, if you leave now for another planet, you may actually arrive later than if you left in 10 years because the technology will improve during that time. So should you wait or go for it?
For those who are now worried about when they should join an interstellar journey: “if the time to wait for the development of faster means of travel is far greater than the length of the journey, then the voyagers should go ahead and make the journey. But if the likely future travel time plus the length of wait is equal or less than the current journey time then they should definitely wait.”
Even better than the Ethan Mollick article is the paper he references, which very seriously discusses the prospect of sending early “welcome parties” to these distant planets. It does have some encouraging facts about human progress, though, such as “80–90% of all scientists who have ever lived are alive now” and “recessions are now measured in months instead of years”. Technological progress has rapidly increased over time despite setbacks like the Black Plague, and we can reasonably expect the same of AI.
The Trajectory of AI Progress
We can also look at the history of AI progress for context, like this chart from Our World in Data that shows gradual increases in some areas like handwriting and speech recognition, and very rapid increases in comprehension and understanding. This mirrors the “jagged edge” of AI that we’ve discussed the last several weeks, in which it’s great at some things and terrible at others.
The Wait Calculation for Non-Clinical Healthcare AI
The reality is that AI is already being adopted where it’s obviously helpful, often without physicians and hospitals really being aware this is happening. Do you know if your billing company is using AI? I bet they are. Do you consider AI to encompass all techniques of rules-based machine learning, which your billing company has likely been using for years anyway and just didn’t have to advertise previously? Almost all non-clinical healthcare companies are trying to incorporate AI for a real competitive advantage, not just a talking point. As a recent Bain report notes, “companies that take a wait-and-see approach in terms of AI are at risk of being left behind.”
Non-Clinical Healthcare AI has already left for its interstellar journey. The wait calculation is so clear - that it makes sense to use the technology now rather than waiting for it to improve - that it’s not up for debate. Billing companies and other medical industries that don’t adopt some AI technology soon won’t be able to compete. In other words, they’ll be left on Earth while civilization flees for Barnard’s star.
However, there are many areas of healthcare AI in which the wait calculation is not nearly as clear.
The Wait Calculation for Clinical Healthcare AI
Clinical applications are the best example of complexity in knowing when the AI models are ready for launch.
I’ve seen others suggest that we need to wait to implement healthcare AI until all these conditions are met:
The legal liability issues are completely resolved
The AI model is always better than a human at every diagnosis
The staff and physicians are thoroughly trained and understand how AI works and all its limitations
There is assurance that no jobs will be lost related to AI implementation
Others say that we should implement clinical healthcare AI now because:
There have to be use cases for the courts to resolve liability issues
The plan is to keep a human in the loop regardless so who cares if the AI is wrong sometimes?
The staff and physicians will learn by using the technology
Jobs will surely be changed but not eliminated
Who is right? The answer is complex and likely depends on your practice setting, staff willingness to learn new technology, and the practical barriers to healthcare AI implementation familiar to all physicians. Many of you may be ready to board the healthcare AI rocket ship but are hampered by the hospital’s long sales cycle or the hospital’s unwillingness to pay for the new technology.
This long delay adds another layer to the wait calculation: if I start the process now, knowing that the technology I want isn’t quite good enough yet, will it be the right time when the funding actually comes through? That 12-18 month window for software acquisition and integration can be a time of incredible growth and maturation in AI technology. I can guarantee that savvy CMIOs and Chief AI Officers are doing informal wait calculations when they put in requests for new software.
They’re betting on the following happening in the next year-ish:
Clinical evidence to support AI use cases comes out
A bonus would be if the evidence is strong enough for insurers to pay for its use
Especially if it changes the standard of care such that insurers have to pay for it
The competition increases and the price decreases
Evidence of cost savings overlaps with evidence of clinical effectiveness
They’re also assuming it will take additional months to customize the software to their environment and workflow, as this Harvard Business Review article points out.
Is There A Better Way to Predict the Wait Calculation?
One way to sense-check some predictions is to use a public prediction calculator like Infer, which helps the US Government by aggregating experts’ best guesses or Metacalculus, which has more of a community approach. Their AI section, for example, has questions ranging from “When will the first weakly AGI system be announced?” to “When will AI wholly create a critically acclaimed film?” (predictions: October 2026 and March 2030, respectively). Unfortunately, none of these have a healthcare section other than biosecurity, so physician AI leaders can’t turn to crowdsourced opinions about when a specific feature will be available or hit a specific milestone.
From Metacalculus
What Should We Do While We Wait?
So how will we know when we should take off on an interstellar journey, or buy that AI software you’ve been considering? I’d suggest the doing the following first:
Learn how it works: That way even if you decide to incorporate it later, you will have some skills and won’t be starting as far behind
Try some of it so you’ll know when it really gets good: Be your own expert prediction calculator. If you try enough products and keep a general eye on the market for an AI approach you need, you’ll be a much better judge of when to take the leap
Think proactively about what it would take for you to actually buy AI software: Set up some requirements for when you’d want to incorporate AI into your practice. Does it need to have a clinical accuracy of above 95% and at least two good studies supporting its use? If you pre-define what it would take for you to hop on the AI rocket ship, you’ll be less likely to miss the critical window.
Conclusion
The integration of AI isn't just an option—it's an imperative journey we're all part of. The analogy of the "Wait Calculation," inspired by the vastness of interstellar travel, isn't just poetic; it's profoundly relevant to our decision-making in healthcare technology. This isn't about waiting for AI to reach a mythical state of perfection. It's about recognizing the moment we're in—a moment with the potential to improve patient care, optimize our practices, and make healthcare is more accessible, accurate, and efficient. The decision to integrate AI into healthcare goes beyond mere adoption; it's about actively shaping the trajectory of our profession and ensuring that we remain at the forefront of innovation.