What innovation has taught me about patient safety: A blog by Muhammad Abid

Summary

I have spent almost three years as an NHS stop smoking advisor in Luton.

A client called me five days after removing her nicotine patch due to a skin reaction. She had started smoking again. She was distressed and convinced she had failed. The answer to her question took me thirty seconds. She could have had it five days earlier if there had been anywhere to turn at the moment she needed it.

That moment made me ask whether technology could do what the NHS structurally cannot. Provide trusted, clinically grounded support at any hour, in any language, in the moments when relapse is most likely to happen. So, I developed an AI-powered stop smoking support tool. This blog is about what building this innovative product taught me about patient safety.

Content

The gap that innovation has to fill

Relapse in smoking cessation does not usually happen because someone stops wanting to quit. It happens in unguarded moments between appointments. At 11pm on a Saturday. After a stressful day at work. When something goes wrong with nicotine replacement therapy (NRT) and there is nobody to call. That structural gap is not a failure of the NHS. It is a limitation of what any appointment-based service can provide.

Innovation exists to fill gaps that existing systems cannot reach. This was mine to fill.

The innovation I built alongside my NHS role

While continuing in my NHS role, I built an AI-powered stop smoking support platform delivered through WhatsApp. The choice of WhatsApp was deliberate. No app download is required. It works on any smartphone and is available in six languages. In Luton, where significant communities speak Urdu, Bengali, Arabic, Polish and Romanian as their first language, removing every possible barrier to access was a patient safety decision as much as a design one.

The platform provides real-time nicotine craving support, NRT guidance, behavioural nudges, relapse prevention messaging and proactive check-ins. Every response is grounded in verified NHS clinical guidance using a technique called retrieval augmented generation, meaning the AI draws from a curated clinical knowledge base rather than generating health information from general training data.

The innovation is not the technology itself. The technology exists. The innovation is applying it to a specific, underserved clinical gap with genuine patient safety discipline built in from the beginning.

Why patient safety had to come before innovation

Before I wrote a single line of code, I had to answer an uncomfortable question. What could go wrong if this AI got something wrong?

In a stop smoking context the risks are real and specific. A pregnant client might ask about NRT safety. Someone in mental health crisis might reach out through the tool. A user might receive confident sounding information that is clinically incorrect. These were not hypothetical concerns. They were situations I had encountered as a human advisor.

I completed a full clinical hazard log covering fifteen clinical and technical risks before the platform went live. I built human escalation logic as the first feature not the last. When the AI detects language suggesting crisis, risk or a clinical situation beyond its scope, it immediately directs the user to their advisor, a crisis line or emergency services.

The innovation only works if the safety net is stronger than the gap it is trying to fill.

The innovation lesson I learned from getting it wrong

My first multilingual responses were translations of English text rather than naturally generated responses in each language. They were grammatically correct but culturally flat and in some cases confusing.

For communities in Luton where English is not the first language this was a patient safety issue not just a usability one. A client who misunderstands health information because the language feels unnatural may make the wrong decision at a critical moment.

I rebuilt the language handling so the AI generates responses directly in each language as a native speaker would write them rather than translating from English. Sometimes the most important innovations are not the ones you planned. They are the ones you discover by getting something wrong.

What this innovation does not yet know

I am currently preparing the AI-powered stop smoking support tool for a pilot with NHS stop smoking services in Luton in partnership with University of Bedfordshire and Luton Borough Council Public Health. The evaluation will compare quit rates against NICE benchmarks and traditional support methods.

But I want to be honest about what this innovation does not yet know. Whether AI can fully replicate the human connection that makes stop smoking support effective. How clients with complex needs will interact with the tool in real-world conditions. What risks will emerge in practice that did not appear in design.

Innovation in health is not finished when the technology works. It is finished when the evidence says it is safe, effective and reaching the people it was built for.

We are not there yet. The pilot is where that work begins.

Opinions expressed in blogs and other content are those of the author. Patient Safety Learning welcomes sharing content and opinions that promotes safer patient care and for the reduction of avoidable harm. The views expressed on the hub however do not necessarily represent Patient Safety Learning's views or values. References to a specific product or service does not imply a recommendation or endorsement.

About the Author

Muhammad Abid is the founder and product engineer of WellTechAI CIC, an AI and LLM powered digital health solution for smoking cessation, behavioural health and prevention. With a BSc in Software Engineering, an MSc in Public Health and Data Analysis, four years of full stack development experience, and almost three years as a qualified NCSCT stop smoking practitioner delivering NHS linked services in Luton, he brings clinical, academic and technical expertise to the problem.