Michelle ThompsonBack pain is a major cause of disability globally, yet an AI-driven app is being used in the UK to assist patients in managing their symptoms. But how effective is it?
When I injured my back in October 2024, the exact cause wasn’t clear. It might have happened when I lifted one of my nephews into the air, or more likely, while moving a bed so my wife could place a rug underneath.
Whatever the cause, my overconfidence in my strength resulted in a lower back injury. A few days later, the pain began radiating down my left leg along the sciatic nerve. Standing was fine, but sitting was painful. To sleep, I had to place a pillow under my legs to prevent the shockwaves of pain.
By mid-December, I saw a pain specialist in Washington, DC. He conducted some movement tests, ordered X-rays, and referred me to physical therapy with a diagnosis of “lumbar radiculopathy.” In the US, this process is fairly quick – I went from pain to starting treatment in about three months. However, this still felt 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 have worked wonders. Each week, I go in, discuss my pain, get a soft-tissue massage, and do supervised exercises and stretches. My pain isn’t gone, but it’s finally under control.
Alongside this, I’ve also been using a new UK-based physio, which is partly human. This “physio” consists of pre-recorded videos from a real physiotherapist, powered by AI and accessible via an app on my phone.
This service, from Flok Health, is the first AI-powered clinic to be trialed by the UK’s National Health Service (NHS). It’s been approved by the Care Quality Commission as a registered healthcare provider and began treating patients like me at the end of 2024.
The reason for this innovation is that many people suffer from lower back pain and have difficulty accessing the care they need. By September 2024, nearly 350,000 people in England were waiting for treatment for musculoskeletal issues, the longest waitlist for any condition. The UK government reported that 23.4 million workdays were lost in 2022 due to these untreated issues, creating both a significant economic and human cost. Worldwide, lower back pain affects around 223 million people and is a leading cause of disability.
Flok’s goal is to provide immediate care, reduce the burden on the NHS, and prevent untreated back pain from worsening. It’s a noble mission, but my main question was: Is AI truly the solution?
The app is designed to assist patients in managing their pain by guiding them through a series of stretches and movements.
Finn Stevenson, a trained doctor and former professional rower in Great Britain’s Olympic development programme, understands how physically demanding rowing can be – especially on the back if form isn’t perfect. While competing, Stevenson had immediate access to top doctors and physiotherapists to treat his injuries. However, after retiring from rowing and facing flare-ups of his back pain, he experienced firsthand the challenges of receiving 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. “In theory, I should be well-equipped to manage this. I had an academic background and three years of professional physiotherapy. If I was struggling, others probably were too.”
Stevenson and Ric da Silva, Flok’s co-founder and CTO, initially met as early employees at CMR Surgical, a British startup focused on developing robotics for soft-tissue surgery. However, with Flok, their goal is to provide treatment for those who don’t require surgery. They aim to assist people with manageable cases.
“There are many conditions where 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.” Their approach is to relieve clinicians from treating simpler cases, allowing them to focus on patients with more complex issues.
“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 welcomed by Kirsty, my physiotherapist, who appears in a minimalist setting with all the essentials of a yoga instructor: a black workout outfit, a bamboo-colored mat, a contemporary chair, and a Monstera plant. She asks me about my pain and background, and I answer using multiple-choice options. Each response leads to a follow-up question or instruction, with the AI adjusting based on my answers.
Rather than a live video call, this interaction is a series of pre-recorded videos stitched together by AI in real-time based on my responses.
While recent advancements in AI have largely focused on generative AI driven by large language models, Flok uses AI in a unique way. “We essentially developed a domain-specific language for describing clinical reasoning,” says Stevenson. Unlike a chatbot like ChatGPT that predicts the next word, Flok’s AI avoids the common issue of “hallucinations” where algorithms generate inaccurate information. Instead, this AI functions more like a choose-your-own-adventure book, with over one billion possible “intervention combinations.” The AI’s power ensures this process is seamless.
Stevenson calls it a “perfect software problem.” “Software is an excellent delivery model if you know what you’re delivering and need to scale it effectively,” he says.
Lower back pain is a leading global cause of disability.
Flok is expanding its reach, having initially launched in Scotland at the end of last year. Recently, it secured contracts to expand into England, although Stevenson has yet to disclose the specific locations. Within the next 12 months, he hopes to cover at least half of the UK. Additionally, Flok is broadening its scope to include treatments for hip and knee osteoarthritis and women’s pelvic health.
However, Flok is not the only AI-based solution tackling musculoskeletal pain. Another app, selfBACK, helps manage lower back and neck pain and is undergoing clinical trials in the US as an adjunct to traditional care services. Like Flok, selfBACK personalizes treatment by analyzing successful outcomes from previous patients with similar conditions and recommending tailored exercise plans. Some studies suggest it can help reduce pain, but a recent trial found that nearly one-third of participants never used the app, and another third only used it sparingly. This led both patients and practitioners to conclude that the app should supplement, not replace, traditional care. Additional research has emphasized the need for such apps to complement, rather than replace, medical care.
Despite these challenges, research indicates that mobile health apps like Flok and selfBACK hold promise for helping patients manage back pain. Still, there are numerous questions regarding the broader integration of AI in healthcare. Elizabeth A. Stuart, a professor at Johns Hopkins University, emphasizes that AI healthcare tools must undergo rigorous evaluations like traditional medical interventions, including regulatory approvals and peer reviews.
“AI tools are no different from other healthcare interventions that we want to learn about the effects of,” says Stuart. “We need to assess how well they work, who they work for, and in what contexts, just as we have done for other healthcare interventions.”
Stuart notes that while AI doesn’t need to be reinvented, it’s essential to recognize that these systems continue to evolve even after deployment. “Some AI tools are self-adapting,” she explains, which contrasts with traditional medical tools like annual flu vaccines, which undergo deliberate, predictable assessment processes. AI, however, can change “under the hood” in real-time, making it unclear how adjustments are made.
Pranav Rajpurkar, an assistant professor at Harvard Medical School, sees potential in AI applications for triaging patient care but points out that collaboration between AI and clinicians often underperforms compared to clearly defined roles. “When doctors and AI review the same cases side-by-side, their combined accuracy is sometimes barely better than doctors alone,” he says. “A clean separation of responsibilities sidesteps these issues.”
“Her pre-recorded videos don’t watch my movements and stretches – they rely on me following her instructions correctly and reporting if something is amiss”
In Flok’s case, the company employs professional physiotherapists who appear in the AI-generated videos that patients interact with. These physiotherapists are also available to answer questions after the AI-powered video sessions.
Pranav Rajpurkar predicts that we will eventually transition from specialized AI tools to more generalist medical AI systems capable of handling a range of tasks across different areas. Rajpurkar is working on this vision with his company, a2z Radiology AI, which is developing AI-powered analytics for radiologists.
“The systems that will succeed aren’t those claiming to ‘replace doctors,’ but those that thoughtfully redistribute clinical work to make healthcare more accessible, efficient, and human-centered,” says Rajpurkar.
In my first 20-minute session with Flok, Kirsty asks a series of questions about myself and my pain, guiding me through various stretches and exercises. She tailors the treatment based on my responses and provides clear instructions, advising me not to push too hard or do anything that causes discomfort.
This experience differs from my in-person sessions with my human physiotherapists in the US, who frequently correct my posture and guide me when I’m doing something wrong—something that happens often. The main difference here is that Kirsty can’t see me. Her pre-recorded videos don’t monitor my movements; they rely on me following her instructions and reporting any issues. At the end of the session, there’s an option to leave a voice note for a real physiotherapist, who will address my concerns or questions personally.
While my back felt better after using Flok, I don’t think the app is right for me. I’m clumsy and uncoordinated, and I need someone to watch my form at all times to avoid further injury. The ability for AI to monitor and provide feedback on posture could be a possibility in the future, but for now, I’ll stick with human physiotherapists.
Emily Martinez
