Name: AI Hype to AI Habit: Making Experiences That Survive the Pilot
Uploaded: 2026-05-13T18:16:12.538Z
Duration: 51 min 4 s
Description: AI Hype to AI Habit: Making Experiences That Survive the Pilot

Transcript for "AI Hype to AI Habit: Making Experiences That Survive the Pilot": Alright. Hello, everyone. Good morning, good afternoon, good evening, wherever you're joining us from. Thank you all for joining. If you've joined this session, you've likely seen two things at once, AI works and AI plateaus. The value is real, but often the scale is not. More than most, I think you'll understand this contradiction. AI is transforming specific tasks but maybe not transforming entire workflows. Some teams are thriving with it while others are quietly opting out. So today we'll be sharing why AI peaks and how to push past that ceiling. We'll be getting started in just a minute but before we hop in a few bits of housekeeping. Because of the size of the webinar we will be having everyone muted so that there's no kind of crosstalk. To enable closed captioning please hover over the stage and click the CC button at the bottom of your webinar window. We'd also love to hear from you throughout the session, so please leave comments in the q and a widget. We have a team working behind the scenes to help answer your questions, and we'll also try to leave plenty of time at the end to do a live Q and A. Luke and I would be happy to answer any questions that you might have and don't worry about taking notes we'll be following up with a recording of the event afterwards you won't miss a thing. Alright so I think it's time that we introduce ourselves I'm Colin Burke, I'm a staff quantitative UX researcher at Superhuman. Been at Superhuman for almost three years now, joining you live from sunny San Diego, California. I lead many of our research projects here at Superhuman, so trying to help our business understand how people use AI tools, what drives value, what drives pain points and frustration. And I'm really excited to be joined by my awesome partner, Luke. And, Luke, would you like to introduce yourself? Of course. Thank you, Colin. Thanks for the intro. Hey, everybody. I'm Luke Behnke. I'm on our product team here at Superhuman. I'm coming from you coming at you from Superhuman HQ in San Francisco, California. It's a beautiful day out there. And, yeah, I help lead our product team here at Superhuman, building a ton of the capabilities that our enterprise customers have been using, and we have been on a journey, to, take our, multi product suite, into kind of the future of AI productivity, and I'm excited to talk to you about what we've learned, on that journey, particularly in, you know, building products, for our customers today. So on that note, I should take a moment to reintroduce you to Superhuman. Superhuman might be a new name. Excuse me. It might be a new name for some of you. You may know us, as Grammarly, which has been a trusted AI assistant used by 40,000,000 people every day. 50,000 businesses use Grammarly every day to help with their AI writing assistance. It goes well beyond proofreading, to really AI writing style guides customizations. We've also united with, Coda, which is an AI, docs and productivity, team collaboration platform, and with Superhuman, now Superhuman Mail, the most productive email and calendar application ever made. And that these, products have come together along with our newest product that we call Superhuman Go, which is an AI assistant that works like Grammarly. It shows up everywhere where you're typing and helps you with a lot more than just writing. It helps you manage your calendar. It helps you manage your inbox. It helps you, work with your, coworkers. It helps you, with tons of tasks each day. And, all those products together are, now, in the Superhuman suite. So, you know, as you'll hear in, our session, we believe the future value of AI in the workspace will be built on, people. That's why our name is Superhuman. We really believe in the power of humans in this, kind of crazy AI era that we live in, and that belief is really at the foundation of Grammarly. As I've said, Grammarly has, existed for seventeen years, serves, 50,000 organizations, 40,000,000 people each day. And, you know, it's we've had because of this experience of, taking this, AI assistant, you know, to our customers over the last seventeen years, we've really had a front row seat into what actually drives adoption and value with AI. I think when I talk to customers today, I hear so much about, yeah, it's working for some people, but not everyone. How do we really drive value? And I think we've had an opportunity to work with so many customers to really learn. And that's what we're gonna be talking about today is just sharing those learnings with all of you so that we can all scale, our use of AI across our organization and that we can really expand, not just, you know, in spite of some of the limitations of the AI solutions we have today, but we can expand because of these amazing AI solutions. And I think, you know, we're excited to kinda talk you through what we've learned. Awesome. So let's start with what we're all hearing. Every study tells a slightly different story about whether and how AI is translating into productivity or business impact, but if you really read between the lines, most research points are a common takeaway. Many of you are probably familiar with this MIT stat that ninety five percent of AI pilots fail. It's a really shocking number, but once that shock wears off, it reveals something important about AI that's actually driving impact. Luke, you talk to our customers every day. What's the story here? Yeah. I think, you know, what we hear when we, you know, go on-site with customers or we see how they're using various tools is that, you know, it it feels like most, AI in kind of, you know, the post chat g p t era kinda comes in a couple different forms. Like, one is it's been sort of bolted into an application that they're using. Maybe this application existed long before these new AI capabilities, but now it has a little AI widget or an AI assistant built in. And it's usually, kinda constrained to what that tool, knows. Or it might kind of, you know, be one of these new, chat based kind of, like, meta assistance like ChatGPT, where, you know, you're, going to a new command line interface, and you're asking questions. You know, increasingly, these two kind of options, what we hear from some customers is that they they kind of feel a little bit like bolt ons. Like, you know, they're asking it's a new chatbot. It's a it's a new feature inside of existing tool. Maybe it only works if you stay in that tool, and then as soon as you leave that tool, there's a totally different experience, a totally different assistant that has different context, that, you know, isn't really fluent to the kind of things, that, you know, employees are trying to do across their tools. And so what we have seen is, and we have a lot of, research, on this, that, you know, we have this, big disparity between, AI power users, those that have really leaned in to changing the way they work and to learning these new tools. But it's a small pocket. Five, ten, 15% of the organization will self declare as AI fluent, and the rest, not so much. Right? And so that doesn't really feel like, to us, like, organizational wide transformation. It feels like sort of uneven augmentation across just parts of the workforce. But the pilots here that do succeed, and this is true of superhuman pilots and true of what we've heard from customers about other AI tools as well, is that they they very specifically target high value workflows. These are things that employees are doing often. They have real value, you know, maybe some critical workflow in the sales process that's actually touching revenue and impacting revenue. And then these AI pilots really go deep on that workflow. They really, integrate not just the technical side of things, the data and the context, but into the experience that employees are already using. And we have certainly seen that, you know, with Grammarly, where you don't have to find another tool, you install it, and it just sort of shows up where you're working to make sure that you don't send an embarrassing, you know, a typo or that you can write with more clarity, in a given communication. And, you know, when AI sort of comes to you or when it's deeply integrated into the workflow, beyond just APIs or MCPs, but it's really about helping the people who are doing the work, right now to get more value from something that is a really high value, you know, a part of their day. A 100%. And as Lou said, right now we're in the chat era. Most people when they think of AI, they're thinking of that chat interface. And chat's amazing, but chat really has a hidden tax. The value that you should get from a chat experience depends really on how well you actually prompt it and if you actually know what it can do. In a large scale experiment, researchers from MIT looked at over 18,000 prompts from almost 2,000 users, and nearly half of the performance gains that they saw in those results came from how well the user actually adapted their prompt. Two people using the same model can have much different outcomes. And this really tells us that in a chat based approach, AI value today is really skill dependent. And it's a major lift to actually reskill the entire organization, every employee, to become a prompt engineer. And by the way, we're affording a large assumption here that people actually find the tools in the first place and they recognize that the task within their workflows could potentially benefit from AI and then go to the right tool and then prompt it. So there's a huge kind of barrier to entry that happens before they even get to the process of prompting. This isn't an indictment of employees, it's really a reflection of where AI is today and its maturity. When it's truly an assistant and a digital partner, people won't actually have to have that burden of remembering how to use it, how to build context, how to prompt like they have to now And so there's a huge burden to understand and also find those tools and know how to leverage them. So let's talk a little bit about AI fluency. Realistically, AI capabilities are advancing faster than the average person can adequately understand or effectively use it. Every knowledge worker is aware of AI and that awareness keeps rising every day. And that's not surprising. Nearly every time you read the news, you watch TV, you hear a story about AI and how organizations are issuing these AI mandates, you know, requiring employees to leverage AI within their workflows. Ninety five percent of US adults have said that they at least know a little bit about AI. But fluency, which is the ability to actually use AI within their workflows and for their tasks and their job, is growing much more slowly. And this is starting from a very low baseline already. Just 13% of professionals identify as AI fluent and 22% are actively AI avoidant, meaning they're avoiding AI whenever possible. And that gap creates a lot of pressure. People know they're supposed to use AI, they hear that message loud and clear from their managers, from their leaders, and their organizations, but they don't really feel equipped to do it well. Recently, an HBR article came out that confirmed this feeling with actual data. It showed that right now, AI isn't actually reducing work, it's intensifying it. The pace has increased, and we're in a world of unknowns, and people are really struggling to keep up while also maintaining their day jobs. This research really showed that people are taking on more tasks, they're experimenting more, stretching their job descriptions and their scopes, and they're working longer hours at a pace that just feels really unsustainable. And then there's the growing pains of what they call work slop, where people are unintentionally or maybe intentionally over relying upon AI, producing work that kinda looks like productivity, but really just creates more confusion in work for themselves and those around them. And we can't solve that gap with endless training. Training doesn't scale fast enough to keep up with AI advancements. As you all know, AI tools, they're advancing every single day. There's new features, new models. It's a huge burden to place on already overwhelmed employees to go through endless training to keep up with that. We have work to do to upscale people and find their best working rhythms with AI, but it's really on us to build more intuitive AI systems that don't have people feeling like they have to take a week off just to learn the tools. Because we build better, more intuitive AI products, adoption will naturally follow. So let's talk about how we combine product experiences and people approaches to go from hype to habit. As builders of AI, we do a lot of market research to understand what people need and then build tools for that. In our research, we've uncovered some insights about how people actually engage with AI tools that I think you all find helpful. It's very clear first and foremost that the workforce is not a monolith. As Luke said there's people in different sort of camps, some are more comfortable with AI, some are less comfortable, and there's very different experiences happening at the same time. Our research team has identified four distinct mindsets across the workplace depending on a person's AI readiness and the capabilities that AI tools offer them. Interestingly these groupings are fairly evenly spread so each kind of quadrant here represents about 25 or a quarter of the workforce And so let's quickly walk through each of these mindsets because each one kinda reveals something about what AI needs to do to outlive the pilot. So the first group is the engaged and enabled. These are the power users. I bet many of the people in this room likely fall into this category. For this group, the human AI fit is already strong. They don't need much training, and in many organizations, they're the ones actually showing people the ropes, teaching them how to leverage AI tools in their work. Ideally, we'd love every employee to be at this level, but that's not how workforces actually work. AI works well for this group because they're willing to make it work. They experiment, they adapt, they push the tools to their limits. We'll always have power users, but if we want AI outcomes that scale across the organization, we have to design experiences for everyone else. So the second group is the interested but inexperienced. This again represents roughly another quarter of the workforce. This group is really interesting because they're curious about AI, they're motivated to improve their work, but they're still kind of building that familiarity and confidence with AI tools. They're not resistant to AI, they're just not yet fluent in how to actually apply it to their day to day work and responsibilities. This is part of the change management that I met I bet many of you are going through today, and this is where addressing the product problems can really shine. What they don't need is more pressure to adopt AI. They need better, more intuitive product experiences that make it easy to get started and easy to see that value very quickly. The next group is the capable of the cautious. These are employees that understand AI and they're confident they could use it, but they're not yet convinced it consistently improves their work. Many of them have experimented with AI, but they've also seen kind of inconsistent outputs or situations where maybe it created more work than it saved. I'm sure many of us have been through this where you're using the AI tool. and maybe it was more work to actually prompt it and get it working the way you want it to, and it didn't quite work out the way you expected. So their hesitation here isn't so much about capability, it's about trust and fit. With this group, it's more of a people problem. Building trust and showing them where exactly AI actually helps in their workflows is critical. And then finally, we have the disengaged and doubtful. These employees have very low comfort with AI and very little interest in actually using it in their work today. Some may have tried it in the past but had very frustrating experiences. Others simply don't see how it fits into what they do every day. As a result, they're the least likely to engage without significant training or a major shift in their experience. To reach this group, AI has felt less like a tool they have to learn all over again and more like a natural extension of the tools they already use. This group could really benefit from people and product support, but better products are likely the highest leverage first step so that they can actually see the improvements in their day to day work with their own eyes to help them get started without having to go through new training to learn it all over again. That's great. Thanks, Colin. Like, I I think the big question that we have here is, like, what does that research tell us? I think that, research how does sort of each group breaks down, you know, 25%, 25%, 25%, 25%, like, is really interesting, and I think we all maybe feel that. We we certainly I think I even feel that here at Superhuman, and, you know, we're an AI focused company. We we definitely have people across that whole spectrum. And so when you step back, you know, I think these four groups are really reacting, to the same underlying experience, which is that this is all changing very quickly, and they're being asked to kind of continuously change the way that they're working, and learn new interfaces. You know, there's kind of the cold start problem as well where you sit down with this chat experience, and you sort of say, what what what can I do? You know, and you, you know, increasingly, those tools have gotten better at helping you get ideas, but you have to be willing to take that jump. And, you know, even we find some people just forgetting that they have various tools at their disposal and they end up, you know, sort of defaulting back to the workflows and the way of working that they knew before. And so when AI depends on, that kind of level of, engagement, involvement from employees, I think that's when we can really hit a ceiling. And, that's maybe where some of these initial pilots or these things that get exciting upfront tend to peak. And, you know, power users, of course, will continue to get, a value from AI, but it won't necessarily scale to the other 75% of the workforce. But I think if we can make, one really simple but significant shift, I think a whole new world opens up with these AI tools. And so, you know, the thing that we've been asking ourself here and that we've talked to a lot of customers about is, well, what if instead of asking people to change everything about how they work, what if AI kind of came to them? It adapted to them. It understood how they work, and it, came to them at the right times, to help them get work done in in an opinionated way. And what if it understood the context, not just the tool that you're in today or in that moment or, you know, not just a handful of maybe integrations that it might be limited in the amount of data it can see through an MCP, but it really understood what you've been doing, and help helps you when you actually need it. And, you know, ultimately, really feels like that super intelligent coworker, who's there to help you get work done. And we think that if AI can work the way people do, we're really just getting started. It's not, that we've peaked, that I think the opportunity to keep growing is very high. So maybe we can break this down, a little bit. To get, you know, AI to work the way people do, I think we have to understand a few of the friction points that Colin referred to earlier. And, you know, there are very solvable challenges here. So first, frankly, we hear all the time people say, you know, my I built this incredible, customized, fine tune chat experience, but nobody goes to it every day. We get, you know, fewer than one query per employee per week on something that we've built. And it's not that that tool isn't great. It's just that people forget to use it. They forget that they have these tools at their disposal. I think, and, you know, those people really need, you know, they they have to change their current flow to get value from these tools. The second is that people have to know how what they can do with these tools. How to write prompts, what they can ask, what kind of things the tool is capable of. There's some new, you know, nouns and terms, that have been invented that people might not be familiar with. And so, you know, as Colin mentioned, the difference in output can really be just knowing what you can and can't do with these various tools, which puts a lot of pressure on employees to learn. And last, tools that are not really technically integrated are not integrated across the employees' tool stack, you know, where something is held in one walled garden, whether that's context or memory, and it's really not how people work. We think you know, we have research that shows that the average enterprise, maintains, over 800 different software tools at any given time. And certainly not every employee is moving between 800 tools, but they're often moving between dozens, maybe even hundreds of tools over the course of a week. And when context differs in every one of the tools you are in, it puts a pretty high cognitive load on employees to know, you know, what they can do in in that tool at that time. So true. And this is where our learnings from our 40,000,000 daily active Grammarly users comes in. Where AI is ubiquitous, proactive, and connected, the employee experience really changes, and the value in adoption kinda breaks out of those pockets of power users. Many of you know Grammarly is kind of that small g that shows up wherever you're writing. It's always there, but it's almost invisible until you actually need it. And that's our client. It works more works across more than a million apps and websites. You never have to leave your your kind of workflow. It shows up wherever you're at. That's what we refer to as Ubiquiti. We also have a standalone editor. It's the same capability. It's a different experience, but you have to go to it. And the result is that millions actually use that Grammarly client every single day, but far fewer actually use that editor. It's the same intelligence, but completely different adoption curves driven in large part by Ubiquiti. Luke, talk to us about productivity and connectivity. Yeah. So once AI is ubiquitous, like Colin said, it shows up in, we we see Grammarly being used across a million websites and apps each day. Then, you know, it just it doesn't just sort of become available, but it can become even more helpful. If you think about Grammarly, as Colin said, it's sort of there every time you're writing. It becomes almost that you forget it's there in many cases, until you don't see it for some reason. And or you're, you know, maybe you're working in, on someone else's machine or something and it's not there. And suddenly, you really miss it. And that's the power of just sort of being there. You don't have to ask it for help. You don't have to know what to do. It's showing up. And, you know, it didn't try to make everybody into an English major, but it really simply made writing clear, error free. It makes people feel more confident before they send that critical email to a customer or to their boss. And that's what proactive AI should feel like. It's there, to help you, to check you before, you know, you do something. And it it really solves the remember to ask problem. And it even can solve the what can I do problem? If it's smart enough to show up at the right time and say, hey. Here's what you can do. Here's how I can help you in this moment. Like a true partner, you know, well beyond, you know, sort of just showing up to give you, you know, writing advice. There's many more capabilities that it could provide. If it anticipates what you're doing, it knows where you're working, and it can step in at the right time. So that's proactivity. And then connectivity, we feel like, is really the final piece. A lot has been written about how these LLMs are only gonna be as powerful as the data that they have access to. Of course, as IT leaders, we have to be careful, also about what data, we are, you know, making available from a security and compliance perspective. But once we find that balance of providing these tools with the right context, it really changes the game. It really takes it from these broad models on one extreme who can talk about anything that they've found on the web, but they don't really understand you. They don't really understand your work. And it also fills that big gap between, on the other side, you know, the, tools that are built into just one product who really only know, information, you know, from that particular product. Connectivity does both of those two things at once. It can broadly answer questions, but it can be very hyper specific to you as a person, the things you know to get your work done. And if the AI knows just as much as you know or even more than you know, that really helps you, act faster, and make the AI even more trustworthy in terms of the kind of answers it's going to provide. So I wanna transition a little bit into, like, out of the theoretical in, to showing you how we have solved this problem here and how we think about the world here at Superhuman. And, take a look at, our newest product that is, Superhuman Go. It's our new AI assistant, which delivers, ubiquitous, proactive, and connected, AI everywhere, someone, an employee is working. And so let's, just bring this down to what this might really look like in real life, and an average person's work experience. So, let's just pretend here, where it's a normal, it's a normal Wednesday, like it is today. And an email lands, from our manager, Carla. And a launch is approaching. This is as a product person, I kind of viscerally feel this stress. We have a lot to do to get ready for this launch, and suddenly, my manager is asking me, what's the status, on these key bugs, these blockers that might really impact the launch? And, she's been helpful. She's pasted a list of things that she's heard about directly into the email. And, you know, that certainly is helpful, but it's also created now a ton of work for me. I'm trying to get work done to get this launch out the door, and suddenly, I need to get answers here. And so, you know, what I would normally do here is click into Jira, click into Asana, make sure I have all the context, go read a bunch of threads, and, you know, we've treated these different apps for years like different destinations. And every single click that we do is a context switch. It brings you from window to window, an opportunity to get, if you're like me, get distracted, completely forget you were doing this task and move on to something else. But when AI is ubiquitous, you can actually see how the pattern changes. So in this case, without leaving email, I can actually just click right over. Thanks to Superhuman Go being connected to these various services, I can just hover over each of these links and get an instant summary of, the latest on these bug reports. No tab switching, no distraction, no digging around. And, you know, Carlo's great, but managers rarely send everything we need. And so the full story is usually never in one place. It's also scattered through Slack. We're huge Slack people here at Superhuman. We spend a lot of time discussing there, and context can easily be lost, across all of these different tools. And, you know, because, again, we have ubiquitous AI that is connected, to all of these places, that context can come to you. And so here you can instantly see the relative Slack conversations tied to those bugs. We are searching behind the scenes for places where, these keywords or where these, bug codes are being searched or being discussed, and we're bringing that context directly to you without having to leave your email. And that's really the magic of Ubiquiti. But let's talk about our second, our our second proposition around proactivity. So here's the thing. There it actually turns out, we you already sent a status report to your manager and said that these bugs were fixed. You're busy. You're stressed. I think we've all been there. And now you're being asked to repeat yourself. And I think, you know, we've all had that moment where we, write a terse email that says, you know, as I have told you before, but here's where Proactive AI can come in, and say, you know what? Here's a recommendation on how you can rephrase your tone a little bit and rephrase your writing, to save you from perhaps a career limiting move. And it gives you a software way, a suggestion. Of course, you can decide to ignore or accept, but in this case, let's accept it and, just move on and, give Carla what she's asking for. So your response isn't now just polite. It's proactive. You're sharing updates your manager didn't even know existed. And, the AI is still working beside you. It's looking for ways to improve your clarity, save you time, give you more information. And, you mentioned the latest status report. The AI recognizes it. You know, Carla, even though you did send it, Carla probably just hasn't seen it. She hasn't gotten to it yet. And so, you know, that's why she emailed you. And so instead of just summarizing it, the AI can find the link directly for you and suggest, immediately adding it in, without you having to hunt and peck for that link. And so what's happening here, if you step back just a little bit, you know, this is, really, you're getting these very smart suggestions in the moment, not just because you asked, but because our system understands the context. It understands, where, you know, has APIs or MCPs that your IT team has connected, to help you bring in the right data at the right time. And, what we really need though is intelligence. It's this awareness of who you are, what you're working on, what they're doing, and why. And that really requires a platform, not just a set of integrations. And so when AI can unify the context, communication, and action, it stops feeling like an add on, and it becomes kind of a new operating layer for how you're getting work done. And this is just one of those many examples of what can be done across, all of these different tools. But let's, take a look a little bit at what happens when, Connected AI starts actually taking actions for you, not just offering you suggestions. So in this case, the system is seeing multiple moving pieces in this email, and it recommends clarifying next steps. You wanna share the report with your team. And you know what? You just need to get a meeting together. Sometimes we just need to get together and schedule a quick review. And so instead of just suggesting this as an idea, it can actually execute. Again, because it has, knowledge of your calendar, those who are on this thread, and who need to be included, we make it really easy, to schedule. Which, scheduling for many of us is one of our most cumbersome and derailing tasks of work. You have to look around, switch tabs, ask people if they can move things, find out when they have availability. But lucky for you, your AI is connected to your calendar. And so it checks everybody's calendars and suggests, a couple of times, and picks an ideal time where you can meet next. This is really AI working the way people do. And with Grammarly, it's been like that for more than seventeen years. And now we're wrapping, today's, AI power and capability into that same experience, going well beyond writing, into really automating and assisting you with so many parts of your workflow thanks to Superhuman Go. Go works wherever your teams already work, as I mentioned before, across a million different apps and websites, and it just really doesn't require a behavioral change. It connects your existing tech stack with, nearly a 100 different integrations so you don't have siloed insights. And it proactively helps, you know, come up with ways that it can help you, in the moment, turning, you know, just from a point solution feature into a true operating layer for how work gets done. And, you know, what you're seeing here isn't just another chatbot. It's what we like to think of as an AI native productivity platform. At the foundation is what you just saw there with Go. This is the platform that understands, your systems, can connect data, and works everywhere your employees are working. On top of that is the agents that we bring, Grammarly and our writing assistant being the featured agent on the Go platform. But now there are many others helping you as a sales coach, as a style guide for writing content, as a connection to the various systems that you're working in to get more done quickly. And you can even build your own agents with our agent builder, no code agent builder, and our agent SDK software development kit, for building more advanced agents that run on this ubiquitous proactive platform. And last but not least, we have a number of surfaces where you're working in our apps, for, our docs product, our wikis, databases, documents, our mail and calendar product, and many other custom apps that you can, work with both humans and AI to get work done. And, what we've seen in practice is, companies are recovering this hidden productivity task of context switching, app switching, and really getting more out of their AI investments. Being able to build on this platform means you can take, you know, AI tools that you've built maybe as siloed chat experiences and actually bring them to our platform as a, an agent on, on go. And now it shows up everywhere your employees are working. We can help, suggest proactively when they should use that capability instead of creating a new destination for employees to go. And so this is not just about, utilizing the capabilities that come out of the box, but it actually means getting more out of your existing investments. And adoption starts on day one. Nothing has to change. This is why we see consistently 90 plus percent weekly active users on Grammarly today across organizations. I don't think there's really any other AI tool that can claim that sort of adoption. And, you know, that's because it's not really just a platform that you adopt. It's a platform that can be customized, can be built on, and, really helps you get more than, you know, what's available out of the box. And so the bottom line is, you know, you're not waiting for us to build what you need. Go can help you kind of close that last mile of AI experience delivery, even for workflows, that are very unique to you and to your organization. Awesome. Thank you so much, Luke. Now it's time for our audience q and a. We received a a lot of thoughtful questions, so thank you everyone for writing in. I know we have one question at the top here, so maybe we'll start there. Marissa had a question. For the people who are interested in experience, what have you seen work best to help them build confidence quickly? Curious your take, Luke, and then maybe I can chime in as well. Yeah. For sure. I I love this question. I think, you know, obviously, we're talking here as much as you can try to adapt the, AI solutions to their current ways of working, I think it's ideal. I think as far as getting folks to adopt some of these new tools, we do what we have found work works really well. Actually, we think about a lot of these tools. They're they're kind of inherently, single player tools. Everybody sets up their kind of, like, experience in some very individualized way. And, I think the the the and because of the pace at which things are changing, the value I might be getting from AI today looks maybe will look very different in a month. We spend, here at Superhuman, we have a Friday meeting we call AI Fridays. Anybody can come and give a two to three minute demo, and show what they're working on. And then after the fact, they can share what they've been working on with the rest of the company. Some of the tools we have internally, make this pretty easy to do. I think, you know, that's our benefit of kind of having built a number of these tools. It's these tools are more collaborative, and you can work together with your team. But I think really just showing, having those people who are in that top, right quadrant that you went over, Colin, around, like, those that are really engaged and really leaning in, having them show what they've been working on. You know, one example is an employee came and showed this, chief of staff, executive assistant agent that they built. They talked about exactly how they built it, and they showed people and even after the fact sort of shared all of the instructions on how to build such an agent, and, for for yourself. And it really sparked a lot of people to say, you know, oh, that didn't seem so hard to build. And so I think, you know, showing the way like that, having that top, that top right quadrant of folks really show the way. And you can do that in various ways. You can do that asynchronously and sharing in Slack or Teams. But we've really found that this, like, demo day every Friday, it's optional for employees to attend. We really encourage employees, to attend. You know, I'd say the other thing is as managers, as people managers, we need to make sure we're making time for people, to continue to experiment and to join meetings like that and to share what they're working on. So those would be two, quick tips kind of off the top of my head. Love it. Love the plug for AI Fridays too. It's always fun to see what people are doing with their work. I'd say, you know, on the research end, you know, the instinct, especially for enterprises, often to send people off to more trainings. And I think we hear a lot from users and customers that they're burned out already and overwhelmed. And so it becomes a big even bigger burden to go off and go to more trainings for a different tool, and then maybe the organization shifts tools for a different tool and they're doing different training. So I think the instinct there is maybe misguided a bit. And to Lew's point, word-of-mouth is incredibly important, especially with kinda realizing the value of these tools and how it integrates into people's workflows. And from a product perspective, really having easy onboarding. There's some AI tools I'm sure we've all experienced where the onboarding is confusing. It's not clear how to integrate the tools that you're already working with. That's hard to figure out where things are. And so having products that are really simple to use, the onboarding is quick, it's clear. It's really easy to see kind of that value and help people get that value right away, especially for this group is is incredibly important. Great question. And then second question, Luke, who would love to hear take again on this. How much time do you think a pilot phase should be, and and how do you tell the difference between an AI pilot that just needs more time versus one that's failing? Love this question. We have spent so much time, trying to figure out, all the answer to all these questions. You know, I would say a couple things here. Number number one, we live in an incredible time where, you know, it used to be not that long ago, actually, where you would you would get demos of software. You would, you you know, see someone who's a professional in that, you know, in the in the tool, kind of show you what's possible. You kinda never knew if those demos were real or not sometimes, and you ultimately had to take a leap of faith. Maybe then you would buy some huge professional services contract and, you know, just hope over the next three, six, nine, twelve months or even longer, that you would realize the value. And we live in a very different time where, you know, if your AI vendors are not offering you pilots or they can really experience the software, that's a huge red flag. And, you know, what we see works best, is really to treat the pilot almost like a precursor to your actual implementation. And now that does take a little bit more time upfront. It usually means you can only maybe bake off two tools, maybe three at most. So it's important to do your research upfront and pick those, you know, those tools for the job that, you know, a a a short list that, you think could work. Maybe you just even pilot one for the job that you're pretty confident will work. But get, spend time upfront designing the pilot. Be very clear on what success criteria wants to be. That's the hardest part, actually. It's not, adopting the tool. It's not trying the tool. It's not using the tool. It's actually coming up with the metrics that are gonna help you decide whether this pilot is effective or not. That could be basic things like adoption and maybe surveys, you know, and looking at trends and patterns in the data. That's sort of table stakes. If you can go further, we've had customers, say, like, a customer support team who has rolled out our AI assistant. And they have said, I wanna actually see that our first reply time, our total handle time of a case, and, our customer satisfaction is higher in the group that is using this assistant. Right? If this assistant is actually helping you write more confidently, it's helping you get more, information to your fingertips in the moment. You should be responding to those chats or emails faster. You should be responding more accurately with, better information, and we should see the results in the numbers themselves. Right? But this is not easy. This is not a lightweight pilot. This is a little bit of a heavier weight pilot. We run them for fourteen to thirty days depending on the level of depth that the customer wants to go into. But usually by the end of that period, you know, thirty days kind of at the most, you have very clear signal how this is going to work. Maybe you've only rolled it out to a small percent of your organization. You have very clear signal if it's going to work or not. And by the way, you've already done a lot of the hard work on implementation that you can turn into your production rollout. In fact, it totally removes that risk of, like, hopefully, this will work once we buy and we, you know, we pay you a lot of money. Hopefully, you'll get this to work. You can actually really see it. And that for us has been the most successful engagements with our our prospective customers. And there's also been times where they say, you know what? Actually, this tool didn't exactly do what I wanted to do. And then we've either pivoted to trying something else or, you know, we found out that maybe it just didn't work. But it puts that it takes that burden off of this implementation period, and it just dramatically increases the chance of success, post purchase. But really to I guess to succinctly answer the question, I think, you know, running the actual thing for for two to four weeks is probably, enough if you've done your homework upfront to set it up for success on really knowing what you're trying to learn. Love that. And I love how deeply that you and the team have thought about ROI too and how to measure that. I think that's sort of an amorphous term in the enterprise space where no one everyone talks about ROI, but they don't really know what it means and how to measure it. I thought that we're able to take a data driven. approach to it. Yeah. We're in an interesting time, you know, with some of these new tools where there's all these people talking about, you know, just token maxing. Just like, hey. Every one way to get people to adopt AI is to not put limits on it. And I think there's actually something to that, which is which is exciting. It's just to let people play and try. And, you know, what we do is I get a report about how much I'm spending across our internal and external, AI tools, but I'm not really capped. But I think we're gonna have a shift, especially, given that many folks are in, early access kind of, pricing with some of these AI model vendors, and they're gonna be increasing changing their prices and continually to, you know, relying on a usage based model, and new models will continue to be quite expensive. And so I think we are gonna get back to a place where, this idea of just use as many tokens as you need is is not gonna work. It's not gonna be practical. It's not, you know, we are gonna be asking we are certainly asked today to justify ROI. Adoption is probably the best proxy for ROI today, but it has to go deeper. You know, that example I gave of customer support teams, seeing the impact on a group of people when they roll out a pilot, that to me is where we really start to get at ROI. We know how much it costs to serve a customer, to serve a case for a customer. And if we can change those metrics, then you can really point to hard dollar values that are accruing to your, you know, to your business. And then that helps you understand how much you should be spending on these AI tools rather than just blindly hoping that that, you know, adoption to value gap is being met. And it's something we think about deeply. We're we're hearing more and more from our customers to justify, and that's another great part of these pilots is coming up with what those ROI metrics will want to be that goes beyond just how many people used the tool. 100%. Sort of related, I'm I'm curious your take on what's the most common mistake you see enterprises make when they're rolling out AI tools? Yeah. I think the risk is, oh, taking a one size fits all approach. You know, what an engineer needs, is very, very different than what someone on the HR team needs is very different than what someone on the CS team, you know, customer success, customer support team might need. And, I think the the biggest mistake I see is not being really thoughtful about the one to three things that each of those different groups, can get from the tool that you're rolling out, and to really make it clear, that, you know, those workflows are available to be improved, say, out of the box. So in the example of superhuman, you can configure by kind of department, you know, which, agents the, those different departments see on the platform. And so, you know, a sales, sales team member will see the sales coach, and will see the, you know, the, the Salesforce, and Gong and other sort of integrations out of the box, you know, whereas someone on the product team will see a different set of agents and capabilities. And those are really, we really try to tailor those to get the things done that are the most high priority. If you sort of run a one size fits all approach or a, hey. We've delivered you a generic tool, and, you know, start using it, you know, that that is where we run the risk of what we talked about in this webinar where you end up hitting just, you know, a fraction of the adoption and usage that you were hoping for and not really getting the organizational transformation, that you were, you know, hoping for. For sure. Awesome. So that brings us to the end of today's discussion. Again, a reminder that a recording of today's webinar will be shared in a follow-up email. Before we wrap, we've launched a poll, which you can find the poll section in your chat box. Please let us know if you'd like to hear from us and keep the conversation going with a superhuman product expert. Thank you so much again for spending time with us. I hope you have a great rest of the day. Take care, Thanks, everyone. everybody.