Atlassian Team '26 | Molly Sands, Atlassian

Clips
More from Atlassian Team '26

Molly Sands

Head of Atlassians Teamwork Lab

Atlassian

play_circle_outline The AI Fragmentation Tax: How Faster Execution Raises Coordination Costs

play_circle_outline Fragmentation Tax Score: Global Study of 12,000 Knowledge Workers Finds Six-Hour Weekly Coordination Gap Between Teams

play_circle_outline Estimated $161 billion loss extrapolated from Fortune 500, considered conservative.

play_circle_outline From 24% to Team-Wide AI: Redesign Workflows, Automate Steps, and Build Shared Skills Libraries

Info

Molly Sands, Atlassian

Molly Sands

Head of Atlassians Teamwork Lab Atlassian

search

>> Welcome back to Atlassian Team '26. We're streaming live here in Anaheim. I'm Alison Kosik, alongside Christophe Bertrand, and we're about to dig into some really interesting research.

>> Yes, we're going to talk about fragmentation tax. I'd like to know what that's all about. I thought we had enough taxes.

>> Never enough. All right. I want to bring in our guest, Dr. Molly Sands. She's the head of Atlassian's Teamwork Lab. Welcome to theCUBE.

>> Thank you. Thanks for having me.

>> So everyone is talking about AI accelerating productivity, but your research suggests that's not the real story, that's not the whole story. Talk us through what we're missing.

>> Yeah. So what we're seeing when we look at workers across the world is that things are moving faster. So in that way, productivity is speeding up. People are doing more. But there's this huge challenge because coordinating across teams, within teams, making sure that everyone moving faster is actually moving in an aligned direction is becoming harder and harder and harder as work speeds up.

>> What is your research showing are the biggest challenges?

>> In that?

>> Yeah.

>> Yeah. So what we're seeing is something that we're calling the AI fragmentation tax. More taxes. And this is really the idea that as we can do more, we have more work to do to keep everyone on the same page. And so we look a lot at both coordination and execution, and that's how us behavioral scientists like to think about work. There is work about work, work to figure out what are the goals we're all aimed at, what are we trying to accomplish together, how do we actually make sure all of the right things are being done by the right people to move towards those goals. And then there's also the execution work. And so that is actually doing all of these tasks. And AI is massively speeding up execution, but we still have a huge opportunity in that coordination area.

>> Right. And I was looking at the research you did and, to quote the numbers correctly, 12,000 knowledge workers across six countries around the world, 170-plus Fortune 500... 1000, sorry, executives. Amazing sample. Now, let's talk about the tax itself, the fragmentation tax. How did you come up with this concept, and what's the methodology behind it? And is 161 billion even close to reality when you think about it?

>> It's a good question. So we really start in my team by thinking about what are the problems that our customers are experiencing at work, and how can we help to solve some of those. And one of the things that we were seeing is that people are moving faster, but they were needing more and more support in that coordination and alignment area. And so that's a lot of what we are trying to help our customers do better and better right now. And we really wanted to capture that opportunity cost, right? What does it mean if everyone's not on the same page? And so with this survey, we looked at a bunch of factors around coordination, so things like goal alignment, things about how AI is actually creating value for teams, if they believe that AI is speeding up their work. And we looked at how much time teams are spending trying to reconcile what's really going on inside the organization, and we gave every team a fragmentation tax score by looking at all these different coordination elements. And we compared the top teams to the lowest performing teams in this way. And there's about a six-hour-a-week difference in how much time is being spent simply trying to coordinate. And so we're taking the Fortune 500 population, we're looking at that six hours a week, which-

>> Yeah, it's a lot of time....

>> pretty conservative estimate of what happens in an organization when everyone is moving in different directions. And we're using those numbers to get to this estimate of 161 billion, which is a staggering amount of money. And-

>> Right....

>> there's a lot of other elements contributing to that that may mean it's a lot higher.

>> What surprised you most when you dug into how that 161 billion came up? What surprised you most?

>> When I first saw that number, I thought, "Okay, well, are we estimating that everyone's wasting their entire week or are we estimating really high cost of employees and we're not?" And so what really stood out to me as I dug deeper into how we got there was, "Oh, these are fairly conservative estimates. This is very realistic, and this is certainly what people are really seeing."

>> Yes. And if you were to apply this to government, I imagine that the numbers would be staggering because there is quite a bit of inefficiency across a lot of organizations. And these are supposed to be obviously profit-driven organizations. You would expect maybe more efficiency. So I'd be curious to see how that would play out. So let's double click on something else, which it's a question I'm asking everybody. So is AI a friend or a foe?

>> I think it's a friend, but you need to be intentional about how you use it.

>> Let's talk about that.

>> Absolutely.

>> Yeah.

>> What do you mean by that?

>> So I see AI as a huge accelerator for teams, I really do, in a positive way. And when we think about what makes a difference for the teams that are really successful with AI, part of that is about their mindset. They are treating AI as a true collaborative partner. So not simply, "Hey, I'm going to send a couple of tasks, getting a little bit of tooling help here." It's actually about sparring with AI, thinking through problems with AI, using AI to expand the set of knowledge that you have access to as an individual. And when you start to bring AI into that fold in that way and really think of it as a true thought partner, that's where we start to see people getting an immense amount of value.

>> Do you think companies maybe, though, are overestimating how much AI is actually improving performance today?

>> Today?

>> Yeah.

>> Yeah. I mean, our data definitely suggests that people are using AI more and more, but we're seeing that only 6% of executives can really point to clear ROI from those investments. And some of that is about where we are in this transformation, that we're early on. There should be a pretty expansive set of things that leaders are exploring right now. We don't want to exploit before we explore, right? You want to see all the places and figure out where is AI really creating value. But I would like to see leaders across the board pointing AI at real problems more and more. So thinking about what is actually holding us back from achieving our bigger goals and then trying to use AI to solve some of those things rather than simply for everything all the time.

>> Thanks.

>> Right. And there's another interesting stat from your research. 24% or only 24% of executives focus AI implementations at a team level, even though teams perform 80% of the work, another 80-20 rule here, which I think is sort of what you were saying, and here we're quoting the numbers. So what are these executives getting wrongs?

>> So with team-level AI implementation, what we really want is for everyone to think about how we redesign full workflows and systems. And I hear a lot of people saying workflows and systems, and I think a lot of people are like, "What is a workflow? What do you mean? What do you actually mean by that?" One of the things I did with my team with one of our customers recently was sat down with all of us in the same room and we talked through some of the biggest challenges that hold us back at work, things like getting alignment, things like fractured knowledge in lots of different places. And we started to talk through what would really let us achieve something that we couldn't before. And even within that day, we started to think through those end-to-end processes, what are all the steps that happened today, and found places where we could automate several steps or where we could augment our own work by using AI as a collaborative partner in those moments. And I think more teams need to be doing this because we shouldn't all just be off working by ourselves with AI. We want it to be part of our team and to help everybody on the team. And we don't need each person creating a totally new way of doing everything. So the more that can become a shared resource, a shared system, shared sets of skills libraries, that is going to be very, very positive for accelerating change.

>> Yeah, that's where the attempt to coordinate becomes chaos and everybody comes to the table with just so much more. At what point do you say, "Okay, we're good with what we have"? Or do you just keep... that radius just keeps growing and growing?

>> Yeah, I do think part of where we need to be is culling things over time, right? It's okay to try five different implementations of something. I think with this type of technology, having lots of people try something and seeing what is actually working is a really good way to go about building the strongest solution, but you need to have those points within your teams and within your company where you then reevaluate, "Hey, we've got six agents now that do this same thing. Is there one that could meet the need better for all of us? Should we all create this coordinated skills library? What should that look like?" And the teams that are doing that are doing a lot better, and they also have the right culture for that to happen. And I think that's something really interesting, too, that we see in this research, is you do need high levels of psychological safety. You need good levels of trust. You need leaders that are out there asking really silly questions about things that they're running into with AI that are confusing to them.

>> What silly questions do you mean?

>> Oh, like, "Hey, I have no idea which of these tools to use for this thing." I've asked my team that several times, "Hey, we have this and this and this, and which thing... What's working best for you all?" And you get amazing insight from everyone. Lots of people within most companies are actually trying this stuff and they're trying to make it better. And asking simple things like, "I don't know how to do this," or "Has anyone set up a way that they're automating moving around things on their calendar based on priorities that are emerging in this ticketing system?" All our systems are very connected, and so usually there's some great insight your team can give you. But by showing that level of humility and that you do not have all the answers, because truly none of us have all the answers in this moment, that helps a lot with getting people to have that conversational flow so they can figure out team systems.

>> Right. And it's interesting when you don't get the same answer to the same question. You get different answers from multiple people and multiple agents. Everybody's trying to create their own answer. There was another data point that was very interesting to me, which is the fact that leaders are 84% more likely to invest in new tools and tech rather than developing people and the skills to use them, which feels to me like we have this sort of trigger-happy set of leaders. "Okay, let's use some tech. Here's new tech." We already have great tech. Maybe we should learn how to use the existing agents we've built or improve on existing applications or processes that work. Was that number surprising to you, that the investment in tech would be something that would be more natural for leaders versus actually the people?

>> So this did not surprise me. I wish that it had, but people are harder to change. And so I think that's part of it, is figuring out how do you really redesign work and how do you equip everyone in your organization to redesign work is a hard problem. And it's going to be where we get stuck in this transformation, though, because it's certainly not the tech. The things the tech is capable of today are mind-blowing. And the more that you play with it and the more you get embedded into it, the more you see what is really possible. And so I think leaders can help their teams by making the tech accessible and also by thinking about new ways to teach and upskill teams. So at Atlassian, we're doing some interesting things where we are really enabling teams to have some dedicated time to redesign work processes. And we're trying to do that at the team level so it's not just every individual redoing their own workflows. And then we also have a series of projects that we're undertaking to rebuild systems and workflows that touch thousands of employees to make them AI enabled. And I think a lot of organizations will probably move towards that, and that becomes a very natural way to get people to use AI that is less about me writing individual prompts or needing to build my own agents or tools, but simply AI enabling the bigger systems that I work within.

>> Yeah.

>> Oh, go ahead. Follow up.

>> I was going to follow up with a question that is not necessarily related to Atlassian, but related to your background. You have spent quite a few years studying, and so have I.

>> To like-

>> ....

>> be here, it's like, what have...

>> .

>> Maybe too many years, okay? And I'm looking at this feature this morning from things in the New Yorker saying, "Well, will we need universities in the future with AI?" So I'm curious, as an academic, literally, what is your take on that? Because it seems to me that this is a convergence point also with what's happening in organizations.

>> Yeah. Okay. So learning is incredibly important, and learning to think is the skill that is going to differentiate people in the future.

>> Right.

>> And I think you learn that. You can learn that from university. You can certainly learn that in other places in the world as well. But we cannot pull back from experiencing the friction that comes with doing really hard things. And what sets us apart as humans is, honestly, creativity. We are not the aggregate of everything that has happened before us. And I very much want to see universities growing and see organizations investing in younger employees, too. Yeah.

>> Hear, hear. Me too. All right, we're going to close out. One more question. If you could correct one misconception that leaders have about AI and teamwork right now, what would it be? And is there a metric that's not being followed that you think could be added to the mix?

>> Yeah. Don't lose sight of the problems. What do you really need to solve? Think about how you would measure success in those areas. There's never going to be a one-size-fits-all AI metric. I hate to disappoint. That would be truly amazing. It's much easier to measure usage. Doing more things is always easier to measure. This is an evergreen problem in the psychology of work. But think about what you are really trying to solve and for whom, and how you would know that you are making real progress on that at a faster pace. And those are the kinds of metrics I want to see people leaning into more so than simply we're using it ubiquitously.

>> All right. Before we go, one more question.

>> Yes.

>> So you're going to be here at the conference. What are the conversations you're looking forward to having?

>> Yes. I am really excited to hear about what people have been building. I can't wait to talk to more of our customers. I have seen some pretty fantastic new things in Rovo. And so hearing from them all about what have they created, what problems have they actually solved, where are they seeing pockets of success in their organization, it's always one of the most energizing parts of team for me.

>> Oh.

>> Great.

>> Really enjoyed talking with you real-world experiences and examples. Thanks so much.

>> Awesome. Thank you.

>> And you've been watching theCUBE, the leader in live technology coverage and enterprise tech analysis. We'll be right back.

Molly Sands, Atlassian

search

>> Welcome back to Atlassian Team '26. We're streaming live here in Anaheim. I'm Alison Kosik, alongside Christophe Bertrand, and we're about to dig into some really interesting research.

>> Yes, we're going to talk about fragmentation tax. I'd like to know what that's all about. I thought we had enough taxes.

>> Never enough. All right. I want to bring in our guest, Dr. Molly Sands. She's the head of Atlassian's Teamwork Lab. Welcome to theCUBE.

>> Thank you. Thanks for having me.

>> So everyone is talking about AI accelerating productivity, but your research suggests that's not the real story, that's not the whole story. Talk us through what we're missing.

>> Yeah. So what we're seeing when we look at workers across the world is that things are moving faster. So in that way, productivity is speeding up. People are doing more. But there's this huge challenge because coordinating across teams, within teams, making sure that everyone moving faster is actually moving in an aligned direction is becoming harder and harder and harder as work speeds up.

>> What is your research showing are the biggest challenges?

>> In that?

>> Yeah.

>> Yeah. So what we're seeing is something that we're calling the AI fragmentation tax. More taxes. And this is really the idea that as we can do more, we have more work to do to keep everyone on the same page. And so we look a lot at both coordination and execution, and that's how us behavioral scientists like to think about work. There is work about work, work to figure out what are the goals we're all aimed at, what are we trying to accomplish together, how do we actually make sure all of the right things are being done by the right people to move towards those goals. And then there's also the execution work. And so that is actually doing all of these tasks. And AI is massively speeding up execution, but we still have a huge opportunity in that coordination area.

>> Right. And I was looking at the research you did and, to quote the numbers correctly, 12,000 knowledge workers across six countries around the world, 170-plus Fortune 500... 1000, sorry, executives. Amazing sample. Now, let's talk about the tax itself, the fragmentation tax. How did you come up with this concept, and what's the methodology behind it? And is 161 billion even close to reality when you think about it?

>> It's a good question. So we really start in my team by thinking about what are the problems that our customers are experiencing at work, and how can we help to solve some of those. And one of the things that we were seeing is that people are moving faster, but they were needing more and more support in that coordination and alignment area. And so that's a lot of what we are trying to help our customers do better and better right now. And we really wanted to capture that opportunity cost, right? What does it mean if everyone's not on the same page? And so with this survey, we looked at a bunch of factors around coordination, so things like goal alignment, things about how AI is actually creating value for teams, if they believe that AI is speeding up their work. And we looked at how much time teams are spending trying to reconcile what's really going on inside the organization, and we gave every team a fragmentation tax score by looking at all these different coordination elements. And we compared the top teams to the lowest performing teams in this way. And there's about a six-hour-a-week difference in how much time is being spent simply trying to coordinate. And so we're taking the Fortune 500 population, we're looking at that six hours a week, which-

>> Yeah, it's a lot of time....

>> pretty conservative estimate of what happens in an organization when everyone is moving in different directions. And we're using those numbers to get to this estimate of 161 billion, which is a staggering amount of money. And-

>> Right....

>> there's a lot of other elements contributing to that that may mean it's a lot higher.

>> What surprised you most when you dug into how that 161 billion came up? What surprised you most?

>> When I first saw that number, I thought, "Okay, well, are we estimating that everyone's wasting their entire week or are we estimating really high cost of employees and we're not?" And so what really stood out to me as I dug deeper into how we got there was, "Oh, these are fairly conservative estimates. This is very realistic, and this is certainly what people are really seeing."

>> Yes. And if you were to apply this to government, I imagine that the numbers would be staggering because there is quite a bit of inefficiency across a lot of organizations. And these are supposed to be obviously profit-driven organizations. You would expect maybe more efficiency. So I'd be curious to see how that would play out. So let's double click on something else, which it's a question I'm asking everybody. So is AI a friend or a foe?

>> I think it's a friend, but you need to be intentional about how you use it.

>> Let's talk about that.

>> Absolutely.

>> Yeah.

>> What do you mean by that?

>> So I see AI as a huge accelerator for teams, I really do, in a positive way. And when we think about what makes a difference for the teams that are really successful with AI, part of that is about their mindset. They are treating AI as a true collaborative partner. So not simply, "Hey, I'm going to send a couple of tasks, getting a little bit of tooling help here." It's actually about sparring with AI, thinking through problems with AI, using AI to expand the set of knowledge that you have access to as an individual. And when you start to bring AI into that fold in that way and really think of it as a true thought partner, that's where we start to see people getting an immense amount of value.

>> Do you think companies maybe, though, are overestimating how much AI is actually improving performance today?

>> Today?

>> Yeah.

>> Yeah. I mean, our data definitely suggests that people are using AI more and more, but we're seeing that only 6% of executives can really point to clear ROI from those investments. And some of that is about where we are in this transformation, that we're early on. There should be a pretty expansive set of things that leaders are exploring right now. We don't want to exploit before we explore, right? You want to see all the places and figure out where is AI really creating value. But I would like to see leaders across the board pointing AI at real problems more and more. So thinking about what is actually holding us back from achieving our bigger goals and then trying to use AI to solve some of those things rather than simply for everything all the time.

>> Thanks.

>> Right. And there's another interesting stat from your research. 24% or only 24% of executives focus AI implementations at a team level, even though teams perform 80% of the work, another 80-20 rule here, which I think is sort of what you were saying, and here we're quoting the numbers. So what are these executives getting wrongs?

>> So with team-level AI implementation, what we really want is for everyone to think about how we redesign full workflows and systems. And I hear a lot of people saying workflows and systems, and I think a lot of people are like, "What is a workflow? What do you mean? What do you actually mean by that?" One of the things I did with my team with one of our customers recently was sat down with all of us in the same room and we talked through some of the biggest challenges that hold us back at work, things like getting alignment, things like fractured knowledge in lots of different places. And we started to talk through what would really let us achieve something that we couldn't before. And even within that day, we started to think through those end-to-end processes, what are all the steps that happened today, and found places where we could automate several steps or where we could augment our own work by using AI as a collaborative partner in those moments. And I think more teams need to be doing this because we shouldn't all just be off working by ourselves with AI. We want it to be part of our team and to help everybody on the team. And we don't need each person creating a totally new way of doing everything. So the more that can become a shared resource, a shared system, shared sets of skills libraries, that is going to be very, very positive for accelerating change.

>> Yeah, that's where the attempt to coordinate becomes chaos and everybody comes to the table with just so much more. At what point do you say, "Okay, we're good with what we have"? Or do you just keep... that radius just keeps growing and growing?

>> Yeah, I do think part of where we need to be is culling things over time, right? It's okay to try five different implementations of something. I think with this type of technology, having lots of people try something and seeing what is actually working is a really good way to go about building the strongest solution, but you need to have those points within your teams and within your company where you then reevaluate, "Hey, we've got six agents now that do this same thing. Is there one that could meet the need better for all of us? Should we all create this coordinated skills library? What should that look like?" And the teams that are doing that are doing a lot better, and they also have the right culture for that to happen. And I think that's something really interesting, too, that we see in this research, is you do need high levels of psychological safety. You need good levels of trust. You need leaders that are out there asking really silly questions about things that they're running into with AI that are confusing to them.

>> What silly questions do you mean?

>> Oh, like, "Hey, I have no idea which of these tools to use for this thing." I've asked my team that several times, "Hey, we have this and this and this, and which thing... What's working best for you all?" And you get amazing insight from everyone. Lots of people within most companies are actually trying this stuff and they're trying to make it better. And asking simple things like, "I don't know how to do this," or "Has anyone set up a way that they're automating moving around things on their calendar based on priorities that are emerging in this ticketing system?" All our systems are very connected, and so usually there's some great insight your team can give you. But by showing that level of humility and that you do not have all the answers, because truly none of us have all the answers in this moment, that helps a lot with getting people to have that conversational flow so they can figure out team systems.

>> Right. And it's interesting when you don't get the same answer to the same question. You get different answers from multiple people and multiple agents. Everybody's trying to create their own answer. There was another data point that was very interesting to me, which is the fact that leaders are 84% more likely to invest in new tools and tech rather than developing people and the skills to use them, which feels to me like we have this sort of trigger-happy set of leaders. "Okay, let's use some tech. Here's new tech." We already have great tech. Maybe we should learn how to use the existing agents we've built or improve on existing applications or processes that work. Was that number surprising to you, that the investment in tech would be something that would be more natural for leaders versus actually the people?

>> So this did not surprise me. I wish that it had, but people are harder to change. And so I think that's part of it, is figuring out how do you really redesign work and how do you equip everyone in your organization to redesign work is a hard problem. And it's going to be where we get stuck in this transformation, though, because it's certainly not the tech. The things the tech is capable of today are mind-blowing. And the more that you play with it and the more you get embedded into it, the more you see what is really possible. And so I think leaders can help their teams by making the tech accessible and also by thinking about new ways to teach and upskill teams. So at Atlassian, we're doing some interesting things where we are really enabling teams to have some dedicated time to redesign work processes. And we're trying to do that at the team level so it's not just every individual redoing their own workflows. And then we also have a series of projects that we're undertaking to rebuild systems and workflows that touch thousands of employees to make them AI enabled. And I think a lot of organizations will probably move towards that, and that becomes a very natural way to get people to use AI that is less about me writing individual prompts or needing to build my own agents or tools, but simply AI enabling the bigger systems that I work within.

>> Yeah.

>> Oh, go ahead. Follow up.

>> I was going to follow up with a question that is not necessarily related to Atlassian, but related to your background. You have spent quite a few years studying, and so have I.

>> To like-

>> ....

>> be here, it's like, what have...

>> .

>> Maybe too many years, okay? And I'm looking at this feature this morning from things in the New Yorker saying, "Well, will we need universities in the future with AI?" So I'm curious, as an academic, literally, what is your take on that? Because it seems to me that this is a convergence point also with what's happening in organizations.

>> Yeah. Okay. So learning is incredibly important, and learning to think is the skill that is going to differentiate people in the future.

>> Right.

>> And I think you learn that. You can learn that from university. You can certainly learn that in other places in the world as well. But we cannot pull back from experiencing the friction that comes with doing really hard things. And what sets us apart as humans is, honestly, creativity. We are not the aggregate of everything that has happened before us. And I very much want to see universities growing and see organizations investing in younger employees, too. Yeah.

>> Hear, hear. Me too. All right, we're going to close out. One more question. If you could correct one misconception that leaders have about AI and teamwork right now, what would it be? And is there a metric that's not being followed that you think could be added to the mix?

>> Yeah. Don't lose sight of the problems. What do you really need to solve? Think about how you would measure success in those areas. There's never going to be a one-size-fits-all AI metric. I hate to disappoint. That would be truly amazing. It's much easier to measure usage. Doing more things is always easier to measure. This is an evergreen problem in the psychology of work. But think about what you are really trying to solve and for whom, and how you would know that you are making real progress on that at a faster pace. And those are the kinds of metrics I want to see people leaning into more so than simply we're using it ubiquitously.

>> All right. Before we go, one more question.

>> Yes.

>> So you're going to be here at the conference. What are the conversations you're looking forward to having?

>> Yes. I am really excited to hear about what people have been building. I can't wait to talk to more of our customers. I have seen some pretty fantastic new things in Rovo. And so hearing from them all about what have they created, what problems have they actually solved, where are they seeing pockets of success in their organization, it's always one of the most energizing parts of team for me.

>> Oh.

>> Great.

>> Really enjoyed talking with you real-world experiences and examples. Thanks so much.

>> Awesome. Thank you.

>> And you've been watching theCUBE, the leader in live technology coverage and enterprise tech analysis. We'll be right back.