Can AI Help Humans Thrive? | Dreamforce 2026

Hello, welcome everybody. So glad to have all of you here today. So, you know, AI is increasingly doing much more than just answering questions for us. It's actually making decisions and taking actions on our behalf, and that creates tremendous opportunities for us. But while it can build new capabilities and reduce burdens, the benefits are not automatic. We cannot automatically assume that everything that AI brings to us is beneficial. In fact, in a recent American Psychological Association survey, more than 3 quarters of psychologists said that their patients had brought up the use of AI for mental health, and emotional support. Many reported benefits from those interactions. However, many also brought up concerns from dependency and distorted thinking. So the question isn't simply whether AI can help make us more productive. It's whether we can design and deploy AI that genuinely helps people flourish? And this is something that's, that's been a big question and passion of mine for a while. And how would we even know if we're succeeding? And so I am so, so happy today to have with me Dr. Arthur Evans, CEO and EVP of the American Psychological Association, today to help discuss all of these questions. Dr. Evans has spent his career applying psychological and behavioral science to some of the most complex societal challenges. And today, he's really going to help us explore these questions more. So thank you very much and welcome, Dr. Evans.

Well, thank you. I'm so glad to be here. I'm honored to be here.

So let's start with the big question in the title of this very session. Can AI actually help humans thrive? And if it can, what should thriving really mean beyond just productivity, engagement, and efficiency?

Yeah, well, I think that is a very important question. And I think right now it's a really important question for the conversations that we're having right now about AI. If you think about it, a lot of the conversations kind of fall out into 2 camps. So you hear a conversation and it will sound as if AGI or superintelligence is the end. So how do we make the most capable model of AI? The better conversation, I believe, is AI should be a means to an end, that the end is human flourishing, people thriving. And that we're building AI so that people, humanity, is better off. And if we're not doing that, then really you have to question, why are we building AI? And if you look at the mission statements of some of the frontier companies, they actually say that. They say that they're building AI to essentially make humanity better. So if you look at it from that standpoint, then the question is, well, what is thriving? What does it mean to flourish? And there are a lot of different definitions, but they kind of coalesce around a few ideas. One of them is that we're feeling good, that we feel good as, as, as people. But it goes beyond just feeling good. It goes to, are we functioning well? And not only are we functioning well, but are we— do we have good connectedness to other people? And in fact, the issue around connectedness is probably one of the strongest predictors of how we will do in life. And so then the question is, if this is about how do we feel as people, are we functioning, are we in strong, positive relationships, then the question is, well, does AI help us with that or does AI detract from that? And if we really want to have as an end goal, we ought to be asking the question, what does AI do then for how we feel about ourselves? Does it make us feel more competent or less competent? What does it do for how we are functioning? Is it helping us in terms of our critical thinking skills? Is it helping us in terms of our judgment and our decision-making? Is it helping us in those areas that really reflect how we're functioning? And then is AI helping in terms of the relationships that we want to have and we need to have with other people? Is it detracting? Is it creating more trust or less trust? Those are the kinds of things that we ought to be looking at if we're— if we really want to see AI help us to thrive.

I absolutely love that. You know, we spent years optimizing technology for things like engagement, convenience, efficiency. And, you know, what we've learned from the last generation of consumer technology is there's a difference between designing something that people want to use and designing something that's actually good for them. Have you observed this? And are there some examples that you can think of where there's this real divide, like what I want versus what's good for me?

Well, there is a divide. And what's interesting about the question you just asked is that a lot of the things that we design these technologies for actually can work against the thriving that we were just talking about. And it's kind of counterintuitive, right? So you would think that a system that is optimized for speed, for example, would be good for us because that seems like a very reasonable goal. But it turns out that a lot of those things actually work against us. And that's why it's so important to understand what I call the human layer in AI. Have you ever walked up to a door and pulled it only to figure out that you have to push it to get through? And if you step to the side and you watch the next 10 people, they would do the same thing that you did, right? So if you think about that example, the door worked exactly the way it was supposed to work. Yeah. It opened, probably had good hinges, probably had a good handle. So all of the design features of the door actually worked. What they didn't check out was how would people experience the door? How were people actually gonna use the door? Now, if you translate that to AI, other technology, the consequences are much more significant. The consequences are, does this help me be a better manager? Does this help— does this technology help me? to, in terms of my cognitive skills? And the question is, when we design technology and we don't ask those questions, we leave it to chance, that's when we get some of the negative outcomes. And if you look particularly at consumer technology and what happened to youth over the last decade, Yes. it hasn't been good in a lot of instances. I mean, it hasn't been good because we were optimizing for speed, we were optimizing for for engagement. We were optimizing for things that on the surface looked like good things, but then we weren't asking the other questions. What is the impact on youth in particular? But I think, again, this issue, I think, goes beyond youth. It is a much broader issue. And so one of the lessons that we learned, and I'm going to pick on my field for a moment, Many psychologists could see some of the impacts that were happening to youth. And I'm going to use youth as an example for this point. We could see what was happening, but as behavioral scientists, you never want to go— and any good scientist is not going to go beyond what the data say. And these new technologies actually have emerging issues for which we don't have a lot of research. And so people were only willing to go so far in terms of what they were willing to say. And what ended up happening is that a lot of these issues got away from us and we're in this era where we have to undo some of that. So the issue for us is how do we in an era where, in an area where there is a range of data, there's a range of evidence from having decades of well-established studies, having meta-analyses, having all of this robust research to the other end of the continuum where we don't have a lot of research, we're still going to have to answer questions at this end of the continuum. And I think for our field, we wanna make sure that we are leaning into that, that we are developing new strategies and ways of developing the evidence, and that we are being assertive and not passive as these new technologies emerge.

I think one of the things that we often think about at Salesforce is we're not B2C, we are B2B, and we do productivity software. And as employers, employees, you don't always have a choice of what tools you are asked to use or told to use. And I think one of the challenges with AI is that, uh, we're told this is going to make your job better, more productive, what have you. But if you don't feel that sense of empowerment, if you don't feel a sense of control, or If the tools take away your sense of purpose, your sense of expertise that you bring, that has a very different impact in how you're engaging with technology. I'm just curious if you're seeing things, if you're seeing something similar to that.

Absolutely. And I think your point's a good one. And I think it points to what I was talking about a little earlier, which is, When a lot of technologies are evaluated, I think we kind of treat them as a technological implementation when what we're really doing is a human systems implementation.

Yes.

We are taking a technology and we're introducing it into a human system. And that system is gonna function differently than it had previous to that. And as leaders, as managers, we have to understand and anticipate that that's gonna happen. It's going to change relationships. It's going to change how people feel about their own autonomy and agency. It's going to affect the customers on the other end. And so I think any good company is not going to— is going to look at these technologies beyond the technological and really try to understand also what is the impact on the human systems? What is the impact on the humans?

Absolutely. That segues perfectly into my next question. You've described behavioral science as a missing human layer in AI development. What does it actually mean to design that layer intentionally rather than waiting to see how people behave once the product is already out in the world?

Well, it means understanding that these technologies are gonna have an impact on us and we have to anticipate what those are. We know a lot about the AI stack. And when I say we, I'm not a technologist, so I'm talking more metaphorically here. People know a lot about the AI stack. They know what the model and how the model is going to behave. They know how, you know, they understand the data. They understand the layer around the interface. What they What we typically don't know is what is the impact on people and what it means to design for people. Number one is to ask that question. And if you look at what has happened with a lot of companies that have implemented technology, what they've done is they've assumed that a task that a person does is equivalent to the job that they do.

Yeah.

And so they take the person out, they, Assuming that because the technology can do the task when the jobs are much more complex. And so what that means is that if we're going to implement these kind of technologies, we have to understand the full breadth of what people do and then to design the deployment in such a way that we account for all of the full tasks of what people are doing.

A couple of weeks ago, APA had the Flourish Summit, which was just such an amazing experience. And one of the things that was discussed was measuring what matters, and how challenging it can be. Like, it can be easy to be able to measure how many lines of code you've written, which is a terrible measure of productivity. But there are a lot of things that we can measure over and over again, but we don't really know, to your point, about that human layer and what the impacts are. And I think that's one of the big challenges that we have, particularly when we're implementing AI in workplaces, is measuring what matters, being able to know, is the AI really helping employees flourish? If we're just able to do more of the same thing faster, Technically, I guess that's good, but not always, you know. I'm just curious what your thoughts are about, you know, connecting again the conversations that we had a couple of weeks ago with, with here as we're talking about the use of AI.

Yeah, well, I think that, you know, this is an area where I think behavioral science is so important. A lot of these ideas, ideas of flourishing and thriving, we've actually developed measures for that. And if we're going to really promote these ideas, we should be using measures that have psychometric validity, things that we've tested out, and using those kinds of measures to actually measure how we're doing, but also design the interventions that we need to get the outcomes that we want. Yep.

We're moving into a new era where, you know, AI isn't just, again, giving us information. It's actually acting on our behalf. So what changes psychologically when we move from simply asking AI to give us information to actually doing that work on our behalf?

You know, one of the things that we're very concerned about for youth and bringing these technologies into the educational environment is what does it do for the skill set that young people have? And I'm going to use young people in this case. The reality is that people are— young people are developing a lot of the important cognitive skills that they have. And when they are interfacing with a a technology that gives them the answer, there's a real good chance that you're going to have atrophy in very important skills. I liken it to weightlessness for an astronaut. So an astronaut that is in space, they have to intentionally exercise in order for their muscles not to atrophy. We need to be thinking about— take education, for example. If we're introducing these technologies, are we creating atrophy in those skill sets that people, that young people need? And how do then, if that is the case, how then do we introduce the friction that we need? How is it that we introduce the resistance that we need? How do we create strategies that actually force some of the learning that is actually needed for people?

At Salesforce, we often talk about specific tasks that are very consequential and we don't want to hand to AI. And we want to make sure that that human is making that final decision. And I think that is, that's something that's really critical for every enterprise or every workplace to make that very mindful choice of this is where we draw the lines. Are there any things that you have in mind that you think we should draw the line and these are just some of the things that we should not be handing over to AI to do for us?

Well, you know, in healthcare, I think that there are a lot of important decisions. Decisions about diagnosing people is very important. Decisions about ultimately what the treatment is that someone is going to get is very important. And those decisions that really have consequences for people are the kinds of things that we should be aided by technology and not turning those kinds of consequential decisions over.

So we've talked a lot about risks. So let's talk about opportunities. If we get this right, What could AI genuinely make better for us as humans? What's one possibility for human flourishing that you think perhaps we might be currently underestimating?

You know, I'm gonna talk about one of the things that I've been working on and many people in our field have been working on, which is I'm gonna take our psychological health for example. We have had in this country a model of achieving psychological health that is really focused on treating symptoms. So we wait until people are in some kind of crisis and we treat the symptoms of whatever the condition is that people have. The reality is that all of us have psychological health and all of us should be thinking about our psychological health and trying to get to optimal health. In order to do that, we have to have a different approach, and the approach has to be one that looks at the entire population. So for people, if you're in the audience and you're doing well, the question is, well, how do we keep you well from a psychological standpoint? If you are someone who is struggling, you may not have a diagnosis, but you're struggling, the question is, well, how can we intervene so that we reduce the likelihood that you're going to escalate and have deeper problems. If you're someone who is really struggling, how do we bring the best science to ensure that you're going to have the best possible chance of recovery? Well, it turns out that AI can be a wonderful strategy or wonderful tool to help us with an approach like that. AI can help us in terms of our literacy, our brain literacy. You know, if we ask people, well, what are the kinds of things that we can do? Most of us know the kinds of things that we can do for our physical health. We know that if we exercise, if we eat the right diet, that we're probably going to do better physically. But if you ask the same question about our brain health, people typically don't know those kinds of things. And we don't have enough clinicians or other people to do one-on-one interventions to help people. Well, that is a perfect example of the kind of thing that AI is really good at. It can personalize it, personalize the information. It can do it at scale. It can make sure that it can ensure that the information is the latest information. It can go a very long way in helping the population have a level of literacy that increases their psychological health. And also for people who are struggling, They may not need a clinician, but they might need some information about things that they can do to manage their own health better. Again, AI is perfect for that. So my hope is that we're not only looking at how do we mitigate potential harms in the technology, but what are the opportunities that we need, what are the opportunities that we can use to really help elevate the overall psychological health, brain health of everyone in the population.

So I know I asked about opportunities, but I also come back to a concern of taking that human out of the loop that we have. Part of the magic of therapy is having that human connection, having that other person to hear you, hear you out and have that, that relationship. And an AI can do an amazing job of simulating that. How is the APA thinking about AI mediating out human relationships?

Yeah, I think, you know, I think that that is an area where the science isn't well developed. We know that that human element that you're talking about is very important for people. We don't believe, especially if you look at what's out, what's available now, that the technology has evolved to the point where It can do such a good job of understanding the nuances, being able to understand and read the nonverbal cues that people are having. The technology isn't there yet for that. It is good, though, for knowledge, more knowledge-based things, psychosocial education and information, but not quite yet for understanding the kind of nuance that we can have in these one-on-one relationships. One of the things that we're very concerned about is that people may experience the technology in that way and may believe and develop a relationship with an entity that is not a person, that may become dependent on that person. And we've seen some pretty bad outcomes when that has happened for people who are quite vulnerable.

Yes. Yeah. I think the survey that I mentioned in the beginning alluded to that very issue.

Yeah. You know, one of the things that's interesting about the survey, so the survey was given to clinicians and they were asked, how many of your patients are using AI? And it turns out a very large, overwhelming majority were. One of the notions, I think false notions that many people have had is that AI is replacing clinicians. It turns out it's not. What's happening is that people are seeking out help.

Yeah.

And as they're seeking out help, they're using AI as a supplement. And I think that's what a lot of people are much more comfortable with, particularly people who have very serious kinds of concerns.

Excellent. So wrapping up, for the leaders and technologists in this room who are building and deploying AI today, what's one thing that you would ask them to do differently if they want this technology to genuinely help people thrive?

I would say that the most important thing is to ask the question about the human layer. I think if you're someone who is developing these technologies, the question is, what do we need to add to our metrics? What do we need to understand about the development of these technologies that help us understand the impact on human beings? If you are someone who is deploying the technology, it is making sure that you're thinking not only about how the technology is going to be used and what— how the technology will be used, but also what will happen to people and to human systems because of the technology. If you are thinking about it from the customer standpoint, you want to understand what is the impact going to be on the relationships between my company and people who are now interfacing with this technology? And so I think the first thing is just to be aware that this is a really important issue. And in the absence of paying attention to these things, companies are not going to be competitive. They're just not going to be because not only is it very expensive when these implementations fail, But you can also really damage the relationships with the customers. I think the second thing is being willing to invest in ensuring that as we understand better that human layer and the impact that technology is going to have on people is to understand what are the interventions that we can do to ensure that we're getting the right kinds of outcomes. And ultimately, I would say it kind of goes back to the point I made in the beginning. Is the technology the end in and of itself, or is the technology one that not only helps us to accomplish our work, but a technology that actually helps us to be better? And I'll just give you one example. If you think about the features of AI, you know, I've made the point that That some of those features actually can work against us, work against our capability, work against our capacity. But we could be intentional about designing into the technology things that actually could help our abilities and our skills. I think that that's a real opportunity that we should be thinking about. The equivalent would be we could put Fluoride in water.

Yes.

What, so what would be the equivalent with these technologies? What kind of things can we embed in technology that will actually enhance our skills and enhance our capabilities? And then being intentional about doing that.

I love that. I'm gonna do a shameless plug for some training that the APA has, and we actually implemented at Salesforce for some of our conversation designers, and it's on designing agents or chatbots for well-being. Would you like to talk a little bit about that? Because I think that could be a tremendous resource for some individuals who are thinking about— they want to have something tangible, a resource that they can bring back with them for, again, how do we create a specific technology that— Or create technology in a way that is specifically for human wellbeing?

Yeah, you know, the idea is emotionally intelligent design and making sure that the kinds of things that we've been talking about today are that people have the skillset, number one, that they have the knowledge around those things, that they understand what we know from behavioral science about how to design those things into technology. And then how to make sure that we can measure, because in the absence of being able to measure, we're not gonna be able to get the outcomes that we want.

I love it. Absolutely love it. I want to ask everybody, please give a big hand to Dr. Evans. I wanna thank you so much. I really enjoyed this conversation, and I'm so grateful for all of the work that the American Psychological Association is doing in this area.

Thank you. Thank you.

Thank you. Bye.