AI, empathy, and health

Thinking about how multi-modal AI can create value for people and payers in health

Jan 28, 2024

We’re back for issue number 2 of the year.

We had a great response to first predictions issues this year, so thank you to everyone, and to all the new subscribers. If you are reading this and haven’t subscribed, please do. It’s a great way to keep up with how emerging technologies are impacting consumer behavior and creating new opportunities for brands and marketers. 🚀

You can also subscribe to my Podcast on Spotify, Apple Podcasts, or YouTube to hear interviews with the builders, innovators, and marketing pioneers shaping what’s new and what’s Brand Next.

If you look at the content and news that finds it’s way into my newsfeed, you’d think that everything in the world revolves around emerging tech. It seems like every day there are new announcements and new capabilities being launched by companies big and small.

I am reminded of the saying that “The pace of change has never been this fast, yet it will never be this slow again.” and it’s hard to imagine things moving faster and yet logically I know they will.

Talking to brands and marketers about the change and the pace of it has left me with a few thoughts and observations:

1. AI Product releases

Every product release brings a new wave of amazement. Watching the product demo’s or playing with the latest product launches can feel like magic and everyone wants the magic for their business

HeyGen let’s you speak almost any language - and many people demonstrated that by trying it and posting the results on LinkedIn and other social channels.

MidJourney can create pictures of people that are nearly indistinguishable from real people. The results are now so good that the New York Times tested people’s ability to recognize the difference as part of an interactive article and based on my small straw poll, it would appear that a lot people guess over 50% incorrectly. The results I got were worse than random.

Click on the picture to go to the article and test yourself

And then there was Google’s demo of Multi-modal AI which felt almost magic - and every one wants magic. Every business knows they need AI, but the problem is that few feel prepared to deploy it, and even fewer feel confident of what they should prioritise to deliver the most value.

So while we are all running, it’s important to remember - The value of Technology is not in what it can do, but in what you do with it. So let’s be deliberate and responsible in what we do with it.

2. IDEAS

I have heard so many ideas of what people want to do with AI. Imaginations are ablaze, because let’s be honest, armed with magic the barriers to launch your start-up or make your marketing ideas realities are much smaller.

That said, I have also spoken to so many people who say that they have an amazing idea but won’t talk about it because they are scared someone might take it. Coming from the agency business, this is a normal thought. When your currency is creative ideas, you guard them.

That said, what is the value of an idea if it is not being discussed, if it is not influencing other people and their ideas, if you are not getting feedback to continue to build and refine your own ideas.

3. RISK

I’m a bit stuck on this one, but I am thinking a lot about what AI means for risk and how best to approach it in business and in marketing.

A lot of my clients have traditionally enjoyed taking the position as fast followers. Why should they be first to adopt a new technology or develop a new service. Let other people go first and then they will be fast followers who catch up after the best practices have been established.

But in an age of AI, when it’s possible for a small business to launch and move quickly with digital services and products, when the underlying technology is moving so fast that by the time you think you have established a best practice there is already a next practice, does this approach still make sense?

Or is it more logical to be the leader and the innovator who establishes the practice, and continues to innovate and disrupt yourself in order to maintain a leadership position. A leadership postion through which you capture the majority of the data allowing you to keep your leading edge and potentially build a moat.

Honestly - the evolving appraoch to risk is something I am thinking a lot about right now and while I am not prepared to give a strong opinion on this - expect to see more from me on the subject and the pro’s and con’s of both positions.

THE POINT

But why does any of this matter. Quite simply because we need to talk more about how brands leverage AI and emerging technologies to create real world impact that can be good for them, the people they serve, and the communities around them.

Technology should be an additive force and by that what I mean to say it that it does not have to yield a net zero result - but only if we are deliberate and responsible and to make that happen we need more dialogue about the ideas that add value.

Welcome to the idea series. This is IDEA number #1 in the series and it was inspired by Google’s Gemini demo which had moments in it that felt like true magic, and yet when thinking about how it could be used to benefit specific industries or peopel, well it left a lot ot the imagination.

So while the potential for multi-modal AI is broad, here is one example / idea to demonstrate just what a significant and positive impact it could have in the area of health.

Multi-Modal AI for health

The following is a hypothetical use case that shows how GLP-1’s, working in partnership with personalized healing coaching powered by multi-modal AI, could provide effective weight loss and long-term management in a cost effective way. Note that this is an exploratory use case only and has not been researched. This is not medical advice. But now onto the Use Case.

The biggest trend outside of AI is GLP-1 drugs, otherwise known as Semaglutides. Originally developed as a treatment for diabetes, this new class of drug appears to have a broad range of other potential uses.

Most notable is it’s impact on appetite and as a result weight management - and with 70% of American’s being either overweight or obese the potential positive impact is huge. So much so that JP Morgan estimates that by 2030, 9% of the US will have been prescribed a form of GLP-1 for weight management. That is roughly 30 million people.

But there are two problem:

Source information about the patient problem is here: GoodRx, Jama Network

If people need to continue taking medication in order to manage their appetite or else risk regaining the weight, medical payers are put into a complicated position.

It’s unlikely that they could afford to carry the high cost of the medication over the full lifetime of the patient, but halting treatment could result in a yo-yo system where payers pay for medication to help you lose the weight, then they stop paying and you regain the weight, at which point the payers start again and help you lose the weight all over again.

This does not sound like an ideal solution for the patient or for the system as a whole.

What we need is a way to help people to both reach and maintain a healthy weight, and to do so in a cost effective way that can be maintained over a long-period of time. And this is where multi-modal AI comes in.

AI HEALTH COACHING

An AI health coach providing truly personalized and in the moment support and engagement can help people manage and maintain weight over the long-term.

Don’t get me wrong, just giving people a digital coach isn’t going to do the job. There are already so many digital fitness and diet apps out there, so how could this be different? That’s the right question.

First of all, it does need to be partnered with GLP-1s to start. The medication is proven effective and so that will drive early results. The results are what’s important if we want to drive lasting behavior change.

Let’s be honest - thousands of people set New Years resolutions every year to go to the gym, but for most it doesn’t last. The lack of visible results are discouraging and so behaviors fade out. What’s needed is to see the results of lifestyle working in combination with the results of medication. In this way we can help people create new habits where their self-esteem is at a high point, and we can then support and reinforce these behaviors over time. In this way we reinforce motivation.

At the same time, it’s also likely that people will regain some weight when the stop taking GLP-1’s. Even if they have changed their lifestyle. And it’s also likely that there will be days when they fail to meet a fitness or diet goal. Those failures can’t be allowed to build up in their mind as barriers, but rather as natural elements that happen - and just require a bit more support.

For that they need a coach who can be there for them in the moment, and that can demonstrate empathy.

Being there in the moment is easy for AI, but does AI have empathy. The answer is clearly NO, and yet people who engage with personalized AI created content and bots have been seen to trust it with things they wouldn’t typically share with a human and in many cases to prefer and feel more supported by the AI. So while it’s not real empathy, it can be trained to be supportive and so people will perceive it as having empathy.

But this is about experience. The coach can’t just be a text window like ChatGPT, while people need to know that it is AI, appearing human is likely to be valuable for most people in building regular interactions and trust. What this means:

Appear as a realistic human avatar that people can talk to through their phones or other devices
Be able to interpret data that people share with the coach regardless of format. Patients should be able to share photo’s of what they eat, or images of where they walked or ran to as part of a program of exercise, or just text the coach or even talk to the coach to tell them when they are feeling hungry or having cravings. The AI coach needs to be able to communicate and interpret content in the same ways a human would.
But the AI coach should do more. It should also see data from IOT devices like smart watches, or even smart ovens, and data from Apps like Uber Eats to see what you are ordering. It’s not about policing the person, but about supporting them by increasing the frequency of engagement.
The AI coach must remember everything you have done in the past. Empathy and trust aren’t created in a single conversation but over time. The AI coach would need to develop a relationship with the patient over time - referencing the past in the interactions. Supporting them with real examples of what they have done before. Unlike engagements with LLM’s like ChatGPT where each engagement is a new thread, this will require long-term context to be maintained.

The objective is to not just to tell people what they should do, but to help them do it by supporting them emotionally and connecting in ways that enables honesty, safety, and support through the process.

EFFECTIVENESS

It needs to be said that while there is research and evidence to support different aspects of this approach, it has not been tested and so actual effectiveness can’t be stated. That said, there are a lot of potential benefits to all participants from patients to providers and payers.

However, without testing, nothing in this article should be thought of as medical advice, but rather as an example of how Multi-modal AI is going to have an impact on peoples lives.

SUMMARY

Fuel by AI powered innovations, the world of marketing and digital engagement are evolving faster than ever. The true power of this technology isn't in the 'wow' factor that is so well displayed in product demo’s, but rather in how we harness it to create real-world impact. This is where Multi-modal AI shines, and while the example focused in this newsletter is focused on health, it’s impact will transcend industries.

So whether you're in health, retail, or any other industry, the message is clear: embrace AI not just as a tool, but as a partner in crafting more ‘human’ experiences. And while it may be tempting to see this as an entirely new world, success still requires knowing people, understanding the problems they have that you can solve, and doing it in a manner that shows empathy and authenticity. That’s what’s needed if we want emerging technologies to improve people lives and unlock the possibilities that were only previously possible in our imagination! 🚀

EXAMPLES OF TECHNOLOGIES NEEDED TO MAKE AI COACHING WORK

To make personalized AI powered health coaching a reality will require a multitude of technologies to work together. Examples of these technologies along with links to some of the businesses building in these areas are listed below. There are many other brands building in these areas and I would recommend doing your own research if you are interested in seeing more.

Multi-modal AI - Allows the ”AI coach” to understand and communicate with someone using any any medium (text, voice, pictures, video)
- OpenAI - GPT4
- Google Gemini demo
Extended context windows - Creates the perception of memory by allowing the “AI Coach” to maintain awareness of a large number of data and interactions.
- Claude 2.1 - Has a context window of 200,000 tokens which is the equivalent of about 150,000 words
- Pi says it has no context window, but just remembers past conversations
  .
Life-like avatars - Allows fo human-like interactivity and communication from the “AI Coach” and so enables the semblance of the emotional signals required to build trust over time.
- Synthesia are creating lifelike Avatars. They primary use case today is training, but it’s easy to see how this gets cross-applied. You can create a free avatar video here.
Real-time content creation - Creation of content with such short latency times that it feels as responsive as everyday human conversations.
IOT devices and data capture - Biometric and activity data captured via wearable and which reflects a person's behavior, activities and health
- Think about everything from the Apple watch, to Withings smart scale, and even smart ovens like this one from Anova.
AI translation - Enables inclusivity by letting the “AI Coach” communicate with people in the language of their choice while maintaining a sense of emotional honesty.
- Rask is providing AI translation to many businesses today
- Seamless Express from Meta is also worth looking at

Brand Next