Back pain is one of the leading causes of disability worldwide, and now an AI-powered app is being used in the UK to help patients manage their symptoms. But how effective is it?
When I injured my back in October 2024, the exact cause wasn’t immediately clear. Maybe I strained something while lifting one of my young nephews into the air. Or, more likely, I hurt myself while hoisting a bed so my wife could slide a rug underneath.
Whatever the trigger, my overconfidence in my own strength led to a lower back injury. Within days, the pain spread down my left leg along the sciatic nerve. Standing was manageable, but sitting became unbearable. To sleep, I had to place a pillow under my legs just to dull the relentless, downward-pulsing shockwaves.
#1From Human Hands to AI Guidance: Exploring a New Kind of Physiotherapy
By mid-December, I visited a pain medicine specialist in Washington, DC, where I live. He conducted movement tests, ordered X-rays, and diagnosed me with lumbar radiculopathy, referring me to physical therapy. In the U.S., this is a relatively quick process—I went from experiencing pain to starting treatment in about three months. But even that feels like a long wait for something as common and debilitating as lower back pain.
Since early January, I’ve been seeing human physical therapists who work wonders. Once a week, I check in, discuss my pain, receive a soft-tissue massage, and go through well-supervised exercises and stretches. The pain isn’t completely gone, but it’s finally under control.
I emphasize human physical therapists because, well, I’ve also been seeing a new physio in the UK. But this one is only partly human. Instead of an in-person specialist, she exists as a series of pre-recorded videos created by a real physiotherapist, powered by artificial intelligence, and accessible through an app on my phone.
#2Flok Health: Can AI-Powered Physiotherapy Solve the Back Pain Crisis?
Flok Health, a pioneering company, has launched the first AI-powered clinic to be trialed by the UK’s National Health Service (NHS). Approved by the Care Quality Commission as a registered healthcare provider, the service began treating patients—myself included—at the end of 2024.
The motivation behind this initiative is clear: millions struggle with lower back pain, yet access to timely care remains a challenge. As of September 2024, nearly 350,000 people in England were stuck on waiting lists for musculoskeletal treatment—the longest waitlist for any medical condition. In 2022 alone, the UK government reported that untreated musculoskeletal issues led to the loss of 23.4 million workdays, highlighting both the economic and human toll. Globally, lower back pain is a leading cause of disability, affecting an estimated 223 million people.
Flok Health aims to provide immediate access to physiotherapy, reduce the strain on the NHS, and prevent back pain from worsening due to delayed treatment. It’s an ambitious mission, but one key question remains: Is AI really the solution?
#3From Olympic Training to AI Healthcare: The Vision Behind Flok Health
In addition to being a trained doctor, Finn Stevenson was once a professional rower in Great Britain’s Olympic development program. Anyone familiar with rowing—whether on the water or just using a rowing machine at the gym—knows that while it’s an incredible workout, poor form can wreak havoc on the back.
As an elite athlete, Stevenson had immediate access to top doctors and physiotherapists. But when he left rowing and his back pain returned, he experienced firsthand the struggles of everyday people trying to access care.
"It was quite a big shock in terms of what access to care looked like for 99.5% of people," says Stevenson, now CEO of Flok Health. "In theory, I should be pretty well placed to manage this. I had an academic background, I’d had three years of professional physio. If I was struggling, other people probably were too."
Stevenson co-founded Flok with Ric da Silva, now the company’s chief technology officer. The two met while working at CMR Surgical, a British startup focused on robotics for soft-tissue surgery. But with Flok, their mission is different: to provide non-surgical treatment for people with manageable conditions.
"There are quite a few conditions where that’s true—you don’t need a scan, hands-on manipulation, or medication," explains da Silva. "What you need is 10 minutes of stretches a few times a week, and that will solve the problem."
#4A Virtual Physio Experience: How Flok Uses AI Differently
The goal is to free up clinicians to focus on patients with more complex conditions—those who truly require hands-on care and specialized medical attention.
"This AI is more like a choose-your-own-adventure book where there are more than one billion intervention combinations
When I first open Flok, I’m greeted by Kirsty, my AI-powered physiotherapist. She appears in a minimalist setting, embodying the aesthetic of a yoga instructor—dressed in a black workout outfit, sitting on a bamboo-colored exercise mat, with a contemporary office chair and a Monstera plant in the background.
Kirsty starts by asking me questions about my pain, and I respond through multiple-choice options. Each answer triggers a new follow-up question or instruction as the AI adapts in real time. But unlike a live video call, this interaction isn’t happening in real-time—it's a series of pre-recorded videos seamlessly stitched together based on my responses.
While much of today’s AI revolution is driven by generative models like ChatGPT, Flok takes a different approach. "We essentially developed a domain-specific language for describing clinical reasoning," explains Stevenson. Because Flok’s AI isn’t a chatbot predicting the next word in a sentence, it avoids the risk of hallucinations—where AI fabricates information—an issue that makes generative AI particularly concerning in the medical field.
Instead of generating unpredictable responses, Flok’s AI functions more like a choose-your-own-adventure book—only with over a billion possible intervention combinations, according to Stevenson. The power of AI enables this to be delivered seamlessly, adapting to each patient's unique needs.
"It’s a perfect software problem," Stevenson explains. "Software is an extremely good delivery model if you know what you're trying to deliver and just need to do that in a more scalable way."
#5Flok’s Expansion and the Growing Role of AI in Pain Management
Flok is rapidly expanding its reach. After launching in Scotland at the end of last year, the service has recently secured contracts to operate in England—though Stevenson hasn’t yet disclosed the exact locations. Within the next 12 months, he aims to cover at least half of the UK.
Beyond lower back pain, Flok is also broadening its treatment focus to include hip and knee osteoarthritis as well as women’s pelvic health.
However, Flok isn’t the only AI-driven solution tackling musculoskeletal pain. Another AI-powered app, selfBACK, is undergoing clinical trials in the US. Designed as an add-on to conventional care, it helps patients manage lower back and neck pain—highlighting the growing role of AI in digital physiotherapy.
#6The Promise and Challenges of AI in Healthcare
Much like Flok, selfBACK uses AI to analyze past treatment outcomes for patients with similar conditions and generate personalized exercise plans. Studies suggest it can effectively reduce pain for users. However, a recent trial revealed slow adoption—nearly one-third of participants never used the app, while another third only engaged with it a few times. Both patients and practitioners concluded that such apps work best as a supplement to traditional care, not a replacement.
This aligns with broader research cautioning that self-management apps should complement, rather than replace, professional treatment. Still, mobile health apps remain a promising tool for managing back pain.
However, integrating AI into healthcare raises important questions. Elizabeth A. Stuart, professor and chair of the biostatistics department at Johns Hopkins University Bloomberg School of Public Health, emphasizes that AI-driven healthcare tools must meet the same rigorous standards as traditional medical interventions—including regulatory approvals and peer-reviewed evaluations—to ensure safety and effectiveness.
#7The Need for Rigorous Evaluation in AI Healthcare
"In some ways, these AI tools are no different from other healthcare interventions that we want to study," says Stuart. "We need to assess how well they work, whether they work, and who they work for—just as carefully as we have done for traditional medical treatments."
While AI doesn’t require an entirely new evaluation framework, Stuart points out a crucial difference: many AI systems continue to evolve even after deployment. "Some AI tools are self-adapting," she explains, meaning their performance can shift over time in ways that aren't always transparent.
She compares this to annual flu vaccines, which undergo rigorous reassessment each year to ensure they target the latest viral strains. In contrast, AI models often function like a "black box," continuously adjusting behind the scenes. This lack of transparency raises concerns about how AI-driven healthcare tools change over time—and whether those changes are always beneficial or well understood.
#8The Future of AI in Healthcare: Specialization vs. Generalization
Pranav Rajpurkar, an assistant professor of biomedical informatics at Harvard Medical School, sees great potential in AI for triaging patient care. However, he notes a surprising challenge: "Forced collaboration between AI and clinicians often underperforms compared to a clear division of labor," he explains. When doctors and AI review the same cases side by side, their combined accuracy is sometimes only marginally better than that of doctors alone. A well-defined separation of responsibilities can help avoid these inefficiencies.
In Flok’s case, the system relies on pre-recorded videos featuring professional physiotherapists who guide patients through exercises. However, these videos don’t actively track a patient's movements in real time. Instead, the system depends on patients following instructions correctly and self-reporting any issues. To bridge this gap, Flok employs human physiotherapists who are available to answer patient questions after their AI-assisted sessions.
Looking ahead, Rajpurkar predicts a shift from narrow AI tools—like Flok, which focuses on a specific condition—to broader, generalist AI systems capable of handling multiple medical tasks across different specialties. His own company, a2z Radiology AI, is working toward this vision by developing AI-powered analytics to support radiologists, aiming to create a more integrated and versatile approach to AI-driven healthcare.
#9AI in Healthcare: Augmenting, Not Replacing, Human Expertise
"The systems that succeed won’t be those claiming to 'replace doctors,' but rather those that thoughtfully redistribute clinical work to make healthcare more accessible, efficient, and human-centered," says Rajpurkar.
During my first 20-minute session with Flok, Kirsty—my AI-driven physiotherapist—asks a detailed series of questions about my pain and guides me through various stretches. Based on my responses, she prescribes a set of exercises for the week, emphasizing the importance of not pushing too hard or doing anything that feels uncomfortable.
This experience differs significantly from my in-person therapy sessions in the U.S., where human physical therapists actively correct my posture and ensure I perform movements correctly—something I often get wrong. The key limitation of Flok’s AI approach is that Kirsty can’t see me. Her pre-recorded videos rely on my ability to follow instructions properly and self-report any issues.
However, Flok does incorporate a human element. At the end of each session, patients have the option to leave a voice note for a real physiotherapist, who will review their concerns and provide personalized feedback—a reassuring human touch behind the AI-driven experience.
#10AI vs. Human Care: The Final Verdict
My back felt better after using Flok, but the app likely isn’t for me. I’m clumsy and uncoordinated—I need someone watching my form at all times, or I risk making my injury worse. While AI may one day evolve to track patient movements and provide real-time feedback, we’re not there yet. Until then, I’ll be sticking with human physiotherapists.
