Just last week, Anthropic’s CEO made waves by suggesting we are not adequately prepared for the magnitude of job losses for white collar workers that are coming in the next few years. In fact, he suggested that:
Over half of white collar jobs could be wiped out in the next 1-5 years
The extraordinary progress AI has made in a short period of time, and the looming narrative of job loss, has me thinking deeply about how AI will impact physical therapy and healthcare more broadly.
I don’t proclaim to be an AI expert (full disclosure: I use ChatGPT to edit my Substacks, but I don’t use it to generate the content itself).
However, I’m already seeing AI’s fingerprints (or maybe 1’s and 0’s?) all over our field, and I believe this is just the beginning.
Looking ahead, I’m anticipating 3 phases of how it will evolve in PT. The phases won’t be distinct or mutually exclusive, but they will bring distinct capabilities and workforce implications.
As I see it, each phase will also have a distinct benefit and a big question.
Here’s how I see it unfolding- and I would love to hear your thoughts. Where do you agree? Where am I off base? Let’s keep the conversation going.
Phase I - Right Now: Administrative Efficiency & Floodgates Opening
I’ll be candid: if you’re not using AI already in your practice, you’re falling behind.
AI tools are transforming the administrative burden in healthcare, streamlining tasks that have long drained our time and energy.
Take documentation, for example. After implementing AI documentation at MovementX, we have seen the time our providers spend doing their notes drop from hours to minutes.
The feedback from our team? It’s shockingly accurate and good.
Companies like ScribePT, Eluve, and Prediction Health, are leading the charge here, using AI to automate documentation, enhance compliance, and optimize reimbursement workflows. Others, like Banjo Health, are automating the Prior Authorization process, a historically painful and tedious task.
Beyond direct patient care and claim cycles, AI is opening floodgates in marketing and outreach. With content generation easier than ever, we’ll see an explosion of voices in the PT space. That’s exciting, but noisy. The question here will be: who can distinguish themselves, how they do it, and what algorithms they will be relying on (be it a platform or a tool).
The benefit of all this? More efficient organizations and a lighter burden on providers.
The big question: As Larry Benz points out, what will employers ask (or demand) providers to do with the time AI frees up?
Phase II - Arriving Soon: Augmented Diagnosis & Treatment
If AI is already streamlining documentation and workflows, the next frontier is clear: augmenting diagnosis and treatment.
Take companies like CareSpace, for instance, who claims to digitize musculoskeletal care via a mobile phone or webcam and recommend personalized plans of care (disclosure again, our team at MovementX has worked with the CareSpace team). Humans then monitor the results after CareSpaces technology does its job.
Others, like Exer.ai, measure motion from a phone camera, and claim to enhance clinical decision making, personalize treatment, and detect falls risk.
Imagine pairing our AI documentation (listening to our patients) with a movement analysis (technology that already exists) to automatically produce a differential diagnosis list, scan for red flags, and augment our clinical diagnosis. Based on the diagnosis, suggested treatments will be available.
If AI does not yet fully and accurately suggest diagnoses and treatments to the point where it’s highly accurate and hard to ignore and accelerate a provider’s evaluation, I’d expect it will arrive very soon.
The phrase I’ve heard most, that I think is closest to reality in this phase is:
AI won’t replace people.
But people who use AI will replace people who don’t.
Or in the physical therapy industry, at the very least, providers who use augmented AI tools will be way more efficient than those who don’t.
The benefit of all this? More accurate diagnosis and evidence based treatment, alongside even lighter administrative burden.
The big question: Just how good will AI get? Could it replace us entirely?
That brings me to the last phase…
Phase III - The Great Unknown: Can AI Bots replace Human PTs?
Once AI is sophisticated enough to effectively diagnose and treat, could it fully replace a human PT or PTA?
In the spirit of improv, a topic for another day, I think the answer will be ‘yes and.’
I do believe AI will be able to handle lesser complex, high-volume cases and potentially fully replace humans.
Think Remote Therapeutic Monitoring (RTMs) on steroids, with no direct human interaction
For example, if someone has simple acute back pain, is looking for a set of exercises to reduce pain and regain function, perhaps they can find an AI bot that will deliver necessary exercises and progressions.
For most, I’d argue that’s a good thing! After all:
We have Silver Tsunami of older adults headed our way
Our workforce might not be large enough to handle them
Access disparities persist, especially in rural settings
And we spend more on low back pain and all musculoskeletal conditions than almost any other condition!
This phase is coming, whether our industry likes it or not. The market will demand it.
The benefit of all this? Information and health empowerment can be democratized in new and meaningful ways to help drive down costs and deliver more (and hopefully better) care.
The question that arises? Will humans actually be needed?
Which brings us to the big, existential question…
Where that Leaves Us (The Humans)
So where does that leave us, the humans, in this evolving landscape?
Should we be preparing for massive job losses in our industry?
And how should organizations engage with AI?
I’ll save that for Part 2 next week.
In the meantime, I’d love to hear your thoughts and opinions. Where do you agree? Where did I miss the mark? Because one thing is for sure: how we engage with this rapidly evolving technology will determine our future.
And the more we can be aligned on how we thoughtfully and intentionally engage with it, the better.
Lots of food for thought!
Eagerly awaiting Part II!
Great article Josh-we have to classify AI in some way and yours makes great sense!