Noom CEO on the need for AI in wellness and health

While generative AI technology is changing healthcare and wellness, there's still a need for a human in the loop.

Noom, a consumer wellness company that helps people lose weight by tracking their food intake and exercise habits, has found that while AI technology is beneficial, AI technology cannot eliminate the need for humans.

On June 27, Noom introduced Welli, a health AI assistant.

Welli helps Noom users get answers to complex questions and frees up human coaching time by handling routine inquiries.

Welli is built on top of the latest models from OpenAI and the Google Vertex AI platform, according to Noom.

The day after introducing Welli, Noom released an AI food-logging assistant. That assistant lets users describe their meals using voice or text.

It also lets users take photos of their meals so the AI technology can recognize ingredients.

While the beginning stages of Welli led to many hallucinations, the current version is configured with retrieval augmented generation, a now widely used technique that helps large language models find and use the most current and specific information.

In this Q&A, Noom CEO Geoff Cook discusses the advantages of AI technology in wellness as well as the need for collaboration between humans and AI.

What has been the response to adding AI into your wellness products?

Geoff Cook

Geoff Cook: It's been well received. One of the things that we wanted to make real clear is when you're chatting with an AI and when you're chatting with a human.

That was the thing we didn't want to have any confusion about, because we felt like that would be an area of concern if you're not sure.

What was interesting about adding the AI into our coaching flow is that we were then able to free up some coaching time to bring those coaches into the premium service of our product and offer one-on-one, [direct messaging] chat -- basically a more intensive, dedicated coach for you on your weight loss journey.

We were able to leverage some of the efficiencies gained from launching Welli and taking that human time that freed up, reallocating those coaches to our premium users and giving them a dedicated coach.

People still demand human interaction and expect human interaction and connection.

Humans are quite good at helping folks with motivation. People feel accountable. They feel accountability to humans in a way they don't, at least today, to an AI.

AI is extremely good at answering things like, "I want to know, is this food orange or green? What's a good recipe right now?"

It's impossible for an AI to have a fully encapsulated view of who the user is and all the data they've shared and, therefore, be able to make in-the-moment suggestions that may be more relevant.

We try to use the humans for the human bits and the AI for the always-on.

How has your used of AI augmented your human coaches?

Cook: The last thing we would ever want to do is replace our human coach with AI technology.

People value their human coach, and no one thinks an AI and a human are equivalent in what value they provide.

They're different. You must be thoughtful with leveraging AI into an existing system. If you just go for efficiency, you may be removing a really valued part of the relationship between a coach and someone on a weight loss journey.

You've got to be careful about how you roll it out.

The way we think of it is, we want to give those coaches superpowers that they wouldn't otherwise have had.

We're able to give the coaches a quick synopsis right next to the chat box so they are quickly able to really understand what this member has been up to for the last two weeks. They don't have to spend five minutes, 10 minutes for each person trying to do that to respond to them.

It helps them get up to speed very quickly.

The future of medicine may well be to do something different or something like that -- not with coaches but with doctors.

You can imagine a system that is able provide the doctor with AI-enabled superpowers, where your every expression is detected, and where a condition might be suggested to the doctor that he might not have noticed because he lacks all the data collection powers and processing of that AI.

Leveraging AI to parse all this data and then help a human make the right decision or the right suggestion at exactly the right moment -- that's what AI done right in a healthcare setting looks like.

How does Noom handle hallucinations from its AI system?

Cook: The way you must deal with hallucinations is to create different anti-hallucination models to try to minimize that as much as possible. You have to make sure that any humans consuming the AI is aware of those limitations.

It's also the case that with the big LLMs -- the large language models from makers like Google, Meta and OpenAI -- they're very much working on this problem too.

How is AI changing the health and wellness industry?

Cook: I was on a panel at a University of California, San Francisco conference a month or two ago. One of the themes from that conference was that there's just not enough doctors to provide the needed care.

There's a lot of other problems in the healthcare system. But even just from a demand and supply perspective, even if you could get everyone a doctor, there's just not enough of them.

People feel accountable. They feel accountability to humans in a way they don't, at least today, to an AI at the same time.

The way population dynamics are working, you're going to have a lot more people retiring, which means a lot of doctors will be retiring.

The actual working workforce will be shrinking. But the number of people they need to support will be increasing, because retirees are living longer.

Therefore, as bad as the doctor shortage might be right now, it will only get worse over the next 20 years.

The way out of that problem is probably not manufacturing more doctors. It takes a great amount of time to do that.

AI could well be one of the more efficient care models and is almost certainly part of the solution.

That was where the excitement was around AI at that recent event. That is where we're excited too.

Esther Ajao is a TechTarget Editorial news writer and podcast host covering artificial intelligence software and systems.