Dr Andrew Ting Explains Why Many Healthcare AI Solutions Fail in Practice

Walk into any hospital today, and you will hear a lot of chatter about the future of medicine. Most of that talk centers on Artificial Intelligence. We are told these tools will solve burnout, catch every hidden tumor, and streamline the messy paperwork that keeps doctors up at night. Experts like Dr Andrew Ting have noted that while the tech is impressive, the execution often misses the mark. There is a massive gap between what a demo looks like in a boardroom and what actually happens on a busy twelve-hour shift. If we do not close that gap, we are just piling more expensive gadgets onto a system that is already breaking.

The Reality of the Digital Disconnect

Software developers are brilliant people. They can build neural networks that identify patterns in data faster than any human brain ever could. But a developer sitting in a quiet office in Silicon Valley or Seattle has a very different lived experience than a nurse in a chaotic emergency room. When a developer builds an algorithm, they are looking for accuracy and efficiency in a vacuum. They want the highest possible area under the curve.

In the real world, a clinician is not looking for a perfect score on a data set. They are looking for a tool that does not add three extra clicks to their already bloated workflow. Many AI tools fail because they are “additive” rather than “integrative.” They require a doctor to log into a separate portal, upload a file, and wait for a result. In a setting where every second counts, that extra step is the kiss of death for any new technology. If it is not part of the natural flow of care, it will be ignored, no matter how smart it is.

Why Technical Accuracy Is Not Enough

We often see headlines about AI beating radiologists at spotting breast cancer or skin lesions. These studies are great for PR, but they often ignore the nuance of clinical judgment. An AI might flag a tiny shadow on an X-ray as a potential problem with 99 percent certainty. On paper, that is a win for the algorithm. In practice, if that shadow is a known benign artifact that the doctor has seen a thousand times, the AI is just creating “alert fatigue.”

When clinicians are bombarded with irrelevant notifications, they start to tune out the ones that actually matter. This is a classic example of a solution being technically “right” but operationally “wrong.” Developers tend to optimize for sensitivity, meaning they do not want the AI to miss anything. Clinicians, however, need specificity. They need the tool to shut up unless there is something truly worth their limited attention. Without a bridge between these two mindsets, the AI becomes a nuisance rather than an assistant.

The Cultural Barrier Between Tech and Medicine

There is also a significant language barrier at play here. Developers talk about “sprints,” “iterations,” and “failing fast.” In medicine, failing fast means someone gets hurt. Doctors are trained to be risk-averse and evidence-based. They want to see five years of peer-reviewed data before they trust a new diagnostic tool. Tech companies, on the other hand, want to ship a “minimum viable product” and fix the bugs later.

This fundamental philosophical difference creates a lot of friction. Dr Andrew Ting has highlighted that moving from a “tech-first” approach to a “clinician-first” approach is the only way to build trust. When a doctor feels like a piece of software is being forced on them by an administration that does not understand their daily struggle, they will resist it. Collaboration cannot be an afterthought. It has to start before the first line of code is even written.

Building the Bridge Early and Often

So, how do we fix this? The answer is deceptively simple but incredibly hard to execute: put developers in the clinic and put doctors in the engineering room. True collaboration means more than just a monthly feedback survey. It means developers shadowing a surgical team to see how they handle instruments. It means engineers watching a primary care doctor navigate a legacy Electronic Health Record system that looks like it was built in 1995.

When developers see the physical constraints of a hospital, their priorities change. They realize that a voice-activated tool might not work in a noisy ward. They see that a tablet-based app is useless if the doctor’s hands are covered in sterile gloves. These are practical, “boots on the ground” realities that you cannot simulate in a lab. On the flip side, clinicians need time and resources to participate in the development process. You cannot expect a doctor to provide meaningful input on AI architecture if they are also expected to see forty patients a day.

Designing for the Human Element

At the end of the day, healthcare is a human-to-human interaction. AI should be the “invisible” layer that supports that interaction, not a barrier that sits between the patient and the provider. If an AI can transcribe a conversation so the doctor can actually look the patient in the eye instead of staring at a screen, that is a massive victory. If it can pre-sort lab results so the most critical ones are at the top of the pile, that saves lives.

The most successful AI tools are the ones that do not feel like AI at all. They feel like a helpful colleague who is always one step ahead. Achieving that level of seamlessness requires a deep mutual respect between the people who build the tools and those who use them. We have to stop treating AI as a magic wand that can be waved over the healthcare system to fix all its problems. It is a tool, like a stethoscope or an MRI machine, and its value depends entirely on how well it fits the hands of the person using it.

Final Word

We are at a crossroads in medical technology. We have the computational power to do amazing things, but we lack a shared vocabulary to make them useful. By prioritizing clinician-developer collaboration, we can stop building “cool” tech that collects digital dust and start building “useful” tech that actually improves the quality of care. It is time to get out of the silos and get into the trenches together. Andrew Ting and other leaders in this space continue to emphasize that the human element is the most important variable in the equation. If we keep the focus on the person at the bedside, the technology will finally live up to its promise.