Turning AI Into Infrastructure with Atlassian’s Sherif Mansour

Hinzugefügt: 2026-05-31

[00:00:00]Anyone who's a builder out there is no short of ideas and trying to work out the quickest way to test them and to turn left or turn right and and the strategy becomes the most [music] important high leverage exercise that we need to do. You could be building a lot of stuff but the boat's headed in the wrong way. [music] Welcome to Prompted. I'm Cameron Adams and I'm fascinated with leaders who are actually using AI to be more creative than ever before. Today we're joined by Sheriff Mansaw, head of product management at Lassian where he leads the AI organization. With 20 years of experience shaping products like Confluence and Jira, he is tackling big challenges across Atlassian's cloud suite when he's not running coaching youth soccering or sketching his next cartoon idea. So Sheree, welcome to the show.

[00:00:47]>> Thanks for having me Karen. Good day everyone.

[00:00:50]>> We have a very unique way of getting to know people on this show and that is through the prompt.

[00:00:55]>> [music] >> This is where we get you to draw something from your childhood that inspired you while you tell us the story about it all in 90 seconds. Are you ready for this?

[00:01:08]>> Let's do it. Sharpies here. Ready to go.

[00:01:10]>> I've also heard you're a comic artist, so I'm expecting big things.

[00:01:14]>> Talk [snorts] it up. So, in primary school a lot, um, while the girls were passing notes in class, uh, about who's got a crush on who, me and my mates were often drawing comics. What I loved about this process of drawing comics was the element of surprise and unexpectedness.

[00:01:29]And so the way we would do it is you would uh draw a bunch of squares on a page. Each square you would draw your thing in the square and then pass it on to the to the next person and then they would draw their thing and then at the end you would get the comic at the end and see how the story progressed and everyone got to change the narrative when it got to when it got to their specific square. So that was definitely gave me a lot of inspiration cuz you you know saw how people change stories and mix things up and gave me the unexpected. Sometimes they destroy your character or sometimes they make your character look a superhero. Everyone always wanted to go first cuz everyone always wanted to set the scene. So I get to go [music] first in this one. Can I show you what I've done?

[00:02:04]>> Yeah. Give it to us.

[00:02:05]>> All right. So it looks usually something like this. And you get to draw the first square. So I've just drawn my little cartoon character there. and uh he's on air and he's looking a bit nervous in a microphone in front of him and uh and you see where the story takes us.

[00:02:21]>> So, I love this. We get to fill in the the rest of the squares. This is really cool. I feel like I should have a marker. You should pass the paper over and I can draw the second square.

[00:02:28]>> I'll pass it on to you.

[00:02:29]>> All right. Well, we'll see what we can do. We've actually got our in-house designer, Avon, who's going to team up with AI to whip your sketch into something new. and we're going to reveal yours, Avon's and AI's creation at the end of the episode. So stay tuned for that if you're listening at home. While Avon works, let's chat AI. Are you ready?

[00:02:49]>> Yeah, let's go for it.

[00:02:51][music] >> So you've been at Lassian now for over 16 years and you've done everything from working on Confluence to driving a product innovation incubator. I would love to know your journey to AI product building and how it's been different to the rest of your career.

[00:03:09]>> It's reset a bunch of assumptions. Maybe if I rewind back, I think like uh throughout the tail end of high school, I became obsessed with building websites in the in the late '9s. It's my first job kind of doing websites for a bunch of businesses around Sydney and then I moved to web apps. And there's a bunch of assumptions you have back then on how the technology works and what that enables you to do. Uh, so I reminded back then of like, you know, basic sites with like then flash intros came along.

[00:03:33]You probably remember that era. Well, some of you intros, >> if you want to age yourself, some of you might recognize that. Uh, and then all of a sudden forms and dynamic apps and then e-commerce and um, and there's assumptions keep resetting on the stuff you can build. So, I went from that to then building a bunch of web apps and then uh, yeah, joined 16 years ago. Gez, it's gone like a flash. But, uh, yeah, and the assumptions have changed. I heard that you wrote a blog post that someone at Alassian read. They contacted you to be a product manager and your response was, "What's a product manager?"

[00:04:05]>> Oh, they told you that story. I still have that email. So, I uh I was a bit of an Atlassian fanboy, still am. And I wrote a blog about this unique business model they had created. Like, who could who on earth bought business software on a credit card online with transparent pricing? Like, you know, back then it was unheard of, right? So I thought these guys are either nuts, they're going to go belly up and the business is not going to take off or they're on to something genius. And they reached out and said, "Do you want to come and apply for a product management job?" Yeah. And I at the time I I Google I looked in our uh job seek website in Australia and I t product manager jobs and there was zero results and there was a lot in the valley in California and all that. I remember talking to my wife going this sounds like a career limiting move but I was never a good engineer and like you know everyone I'm reading says like you might want to try product management if you're not a good engineer so I gave that a go see I've been doing that ever since and then just yeah been working on most of apps and we've gone through several technology shifts of assumptions on we used to ship software behind the firewall with no before the cloud with no data then the cloud came and the assumptions reset and what you could do was quite different uh and now a mobile came and then now AI so how we build our apps we're kind resetting our assumptions on what tools we have access to, how smart the app can be. Uh, and if you're starting again, you would build the apps quite differently. So, we're in this evolution phase of rewriting our app slowly with the assumption that um AI is always there and available.

[00:05:27]>> Did you dive straight into being head of AI or was it this gradual learning curve of like, oh, we should introduce more AI into this and suddenly boom, you're you're head of AI now?

[00:05:37]>> Yeah. No, I have I gravitate toward a lot of um 0ero to one projects at Atlassian and so three we've always had an AI team traditional uh ML style stuff. uh but I think three years ago for most of us in the industry there was another I like to call it a GDPR like moment where everyone's uh everyone's road maps got reset with this transformer model discovery and uh and then chatpt and came later and what could we do and so uh there was a group of us that was a lot of tinkering of what's possible what does this mean for us and uh and a lot of learning >> how do you think it's fundamentally changed product management and product development like what does the experience of building products with AI now look like compared to what you're doing 7 years ago.

[00:06:20]>> Crazily different. So there is the whole discovery process which I often describe as a way to just working out are we building the right thing. Uh the cost of that discovery process is quite large uh traditionally because you typically have someone build something as close as possible to the real thing or find a cheap way to build something then you end up rewriting it uh and all that stuff. uh and I think now with AI the cost to discover and get feedback is reduced drastically on the concept or the idea you're trying to test the hypothesis you're trying to validate and then once you've proven it out and you're building it building it's a lot quicker now with AI great at generating code uh exploring different design options that happens as well uh then getting customer feedback you know similar probably to Canva and many of your listeners we get thousands of feedback items a day like tens of thousands and So how do you analyze that stuff? We used to do that with human eyes and whatever and now we have AI take a first pass and summarize that for us. So closing those feedback loops are a lot quicker and then definitely creates a lot of compression but also helps everyone create more uh which I think is the exciting part. I think anyone who's a builder out there is no short of ideas and trying to work out the quickest way to test them and work.

[00:07:32]Do they work? Did I get it wrong or do I need to turn left or turn right? AI is just such an amazing partner in that whole uh discovery life cycle.

[00:07:39]>> As the transformer model started coming on the scene, did you was it existential for you? Were you like, "Oh crap, there's a bunch of stuff happening that might disrupt us or were you like this is fantastic. This is open opportunity for us."

[00:07:52]>> You always go through waves. So you Yeah. And and you go through these waves of emotions, even just what it means for society, right? I think everyone goes through these waves of motions of like and it's really just about thinking through the implications and trying to understand a trajectory and then play that trajectory forward to then simulate what are the possible implications of it. The worst thing you can do is go into this paralysis mode or feel like oh I'm not going to do anything. The best thing you can do I would say to teams and AI is or if you're doing anything with AI and you're not sure what to build or you know what to do next define longterm as the next six months. So, uh, in the next 6 months, is this the next best possible move you could do? And most people like, oh, absolutely. We should do this. I'm like, well, move ahead because everyone's learning at the same boat. Everyone's moving forward.

[00:08:36]Uh, and so far it's been awesome. It's, uh, you know, it's created a ton more opportunities than we could have imagined. I imagine your experience deep in innovation, like actually looking at innovation and understanding how it performs and how you get more innovation out of your teams probably has set you in goodstead for the last couple of years at least. It helps us move a lot faster as the bigger you get, the slower you're worried about getting. But what AI's enabled is accelerated the speed at which people can make decisions. And I use the phrase a lot with my teams. They hate me. I use the phrase shed brain.

[00:09:07]How are we going our shed brain? Is someone lagging a bit or someone ahead?

[00:09:10]And how are we building that shed brain quickly? And to help with that innovation, we need people to when we when you're talking about something, we need to be seeing the same thing in our minds. Uh, and so AI is an amazing visualization partner that helps us get to that shared brain quicker. No, we mean this, not that. Oh, okay. I was on a different wavelength. So, it helps speed up those alignment on the kinds of innovation things that uh that that teams can do just across the world and it's been awesome for that.

[00:09:35]>> I know you embrace it wholeheartedly and you've even brought it into your personal life. You've done some amazing creative projects with your kids. Can you tell us a little bit about those?

[00:09:44]Honestly, something Cameron I tell a lot of leaders to do which is you know um leaders often like ah one of my teams to use AI and you're like and you're like how are you using AI personally and uh you know it's a start most people will start with like I asked Chip questions and or like I stopped using Google and now I'm chatting but it's the same thing and behavioral change is hard um so how I've been trying to model that behavioral change in my family just tinker a lot with with my kids I'm fortunate my kids still think I'm cool although the older one is just at the tail end of >> they're always going to think you're cool.

[00:10:17]>> Oh, we'll see. Um, but my son, he loves coding. So, he uses Scratch, a visual coding language. So, doesn't like writing code syntax, but he's using is in the age where he's using the visual language. So, he's been man going wild with his games, his tower defense games and all that kind of stuff. He uses actually Leonardo to to make his sprites for his images for uh his little games.

[00:10:39]Uh most recently we vibe coded together, built a little moneymates tracker kind of app for within our family. Track that pocket money. My daughter, she's on the creative side, so she loves music and arts and drawing. Man, I don't know if you've played with oro services, Cameron.

[00:10:57]>> Uh she haven't got too much love out of them because I like going deep on music stuff, but >> uh she goes nuts. So, she uh she texted me the other day with uh she had a neighbor over and um I could play the song for you if you want. And uh they were they wanted ice cream from the fridge and I kept saying no. And then they came back with this AI generated song of like pop style. Uh can we please some have some ice cream in the fridge and it was just very catchy. So I mean after that effort I had to had to give her some ice cream. But yeah, we do a lot of stuff together. On the normal side we we have been using AI a lot for homework help. So, you know, I've got an AI that uh I've just said, "Hey, you're E6 teacher in New South Wales. You know, the New South Wales school curriculum syllabus. You teach math very well."

[00:11:42]>> And I'll put it in voice mode. I'll say, "Don't give away any answers. Always help whoever's talking to you to try to uh get them to work out how to get to an answer." And I'll put on audio mode and we'll walk through some math questions with my son cuz my math is not great as well. So, it helps me a lot when we go on long drives. We just went on a long drive uh down the south coast and um we'll put AI audio on the in the car and we'll do the old Do you remember as a kid doing the old um choose your own adventure books where you're like you get to page seven you're like if you want to do this go there and if you want to do this go you know turn right or whatever.

[00:12:15]>> Yeah. Um, so we'll we'll make we'll say we're heading down to Berry in the South Coast. Use some real uh life um streets and um sites and uh museums there or whatever it is. and um my name's Sharief and my super characters abilities to be able to jump jump through between buildings and my son and my daughter will say their characters and then it'll make an adventure story where we're all superheroes heading down the south coast and it's just like a really cool way to discover the town and everyone's yelling out no turn left or no jump out the window and like you know to the AI and uh it's pretty insane. Yeah, lot lots of fun stuff.

[00:12:49]>> That is super creative. I'm going to have to try that.

[00:12:51]>> Yeah, works really well. I was just actually chatting to someone from OpenAI about how study mode evolved in their product and I'm not sure whether you've tried it but study mode it helps you explore topics and and learn more about them rather than giving you the answers.

[00:13:06]>> Sounds like you've kind of concocted your own study mode but it's deeply ingrained in your relationship with your kids and and being part of their learning. How does something like study mode just in chat GBT working with your kids with no intervention feel to you?

[00:13:23]>> I get quite excited about it because it's hyper personalized. So, example is we did a quiz the other day where uh my son was into Fortnite and I was like, "Hey, if you get three answers right in a row, give him a Fortnite trivia quiz answer." Right? So, you can use it as a motivational tool. One thing that is often not talked about is how AI is a tool for just education and it's such a and you can personalize education for each person. Everyone learns differently. I was always a visual learner. I always struggled with textbooks. Some people are good at math and sciences and so you can adapt the learning. So honestly, I think it's incredibly exciting the what that will do to a next generation of kids who can better adapt to different styles of learning. AI can better adapt to the different styles of learning to accommodate them. One of the things that most AI tools at the moment feel like they're in is single player mode. It's you talking to an assistant. You query it. You get an answer from it. You copy and paste that answer to some Slack channel where you're talking to your teammates or you put it into a document or an email and then you send that off to someone else and it is full of friction. What's your view on where AI needs to go in terms of collaboration and making it a true multiplayer experience? I I have this framework internally always thinking pyramids. Uh must be the Egyptian me or something. Uh but like which is like just the maturity like if I talk to a customer about how they're going in the maturity curve and you always hear like three rough stages of of the story. Stage number one is like I type things into the AI and I get answers to my questions. Stage number two is the aha moment that wait it can help do things for me. It can generate that media for me. It could give me a draft idea for a social media post. he could um you know help design the website for me whatever it is sort of the second stage. The third stage I think is really to answer your question is like when we start to deploy these we call I call them virtual teammates these new teammates with us in teams workflows every day you start to remove that friction of like you know it's just a from a personal productivity gain me chatting to a personal tool which I think a lot of us will have in our day-to-day personal AI assistants but in a team in a work context you're often trying to work out how to deploy AI to a group of people and how to make the use of that and so at that level you're getting an AI agent and deploying it in some sort of business workflow. Uh a workflow could be as simple as creative marketing team writing a blog series and they're tracking their uh the different ideas they want to track and they might go back and forth with the AI on uh different topics and and different content and all that kind of stuff.

[00:15:51]People are now feeling more comfortable deploying agents in a collaborative workflow with multiple humans together uh collaborating together to to get to an outcome. to show share examples where it's been working really well where teams have business workflows already modeled in a system or documentation. So like you know this is what we do when uh what's our our procurement team at Atlassian review a bunch of contracts all the time that's their full-time job and they have a set of knowledge they review the contract with there's certain clauses they look for etc that knowledge is already documented in a tool and they already got the workflow of someone raises a ticket and it goes through a workflow of reviewing it etc coming up with a new proposal sending it back to the vendor etc that's already mapped in an actual system in a workflow so the companies that are doing amazing ly well with AI and deploying agents and AI already have a workflow modeled in their business. You know, uh famous app store uh tracks a lot of app reviews through a workflow system like you know the apps come in and needs to go through process whatever and so at each step of the workflow where they've got a business workflow modeled they're looking going oh I could add AI here to either reduce time or maybe increase the creativity or give me different options uh at this step of the workflow. And so they're adding these virtual teammates at steps of the workflow for specific tasks within that workflow. And that's where we're seeing the most leverage happen.

[00:17:13]Uh you know, for me myself, I use a lot of agents in different steps of my workflow. Uh I have an agent that summarizes all my customer feedback about all our AI products in into themes on a daily basis and I read it and makes some recommendations and then it goes and tracks all the work in Jira and all that kind of stuff. And so it's about finding in your daily workflows as a team, the team workflows, the shared workflows, not the personal ones. Where can AI jump in and take some of that repetitive work that I can clearly describe and give you the knowledge for so that you can speed this thing up or give me better ideas for this particular step in the workflow. You've actually done a ton of research in this area and you recently uh launched the Atlassian AI collaboration index which is a research report based on surveys of 12,000 knowledge workers across six countries and hundreds of different companies. And you've analyzed how AI affects teamwork and business outcomes.

[00:18:07]Seems like most companies have adopted AI, but your research shows that 96% of them haven't seen any major transformation yet. So that means only 4% are reporting real change in efficiency, innovation, or creativity.

[00:18:22]Is that an accurate representation of what the industry is currently like?

[00:18:25]>> I think it's pretty spot on, dude. I I think you'll we see we see a lot of that companies taking the first step where they're saying there is a lot of individual productivity, but it's quite hard to measure. it's like fluffy people are just getting stuff answered whatever but uh that 4% really is is where there's a big difference between those organizations that do that and what everyone else has been doing but yeah I would agree with your your statement absolutely yeah >> is that just a matter of time like people get used to it organizations will catch up or you think that there needs to be some very directed leadership and learning that needs to happen to actually bridge the gap >> yeah time helps education always helps when any technology shift has happened.

[00:19:07]I mean, if you think of yourself personally and and when mobile came out or when e-commerce came out, how long did it take before you felt using that thing in a work context and what happened before it and most of the time you use that thing in a personal context? Like we became comfortable with online shopping personally before it became comfortable in a work context.

[00:19:26]We're comfortable with personal mobile phones before we use mobile phones in a work context. So I do think some of that is personal modeling of that new behavioral change with how we use AI personally that helps at that time. But I think applying it to the specific leaders that are deploying AI in those organizations in that large percentage that fail I I think Cameron what I often see is like they're looking for a killer use case and trying to discover a new set of problems with AI. They're like okay let's see what we can do here.

[00:19:58]the ones that are getting the best impact for maximizing creativity or speeding up a workflow or saving time, they're looking at what are my teams already doing like they're coming at the problem from a quite a different angle like what are my what's my business already doing and where can I improve things so signals might be like tasks that people are doing repetitive work um cues of work is often a really good signal miscommunication or standards enforcement is pretty big in big companies they're always like there's standards for risk standards or road map standards or whatever it is. Uh and that's often a lot of stuff. So the ones that look at what are the you know behavioral change is hard. So like I don't go to the gym but people tell me if you go to the gym it's a pretty hard thing right. So behavioral change in a team is even harder. So to get a collective group of people to change behaviors is even harder. Right? So and so what can we do to reduce the friction to for that behavioral change? Let's look at the behaviors that our team's already doing. So start with that lens, not come coming up with it some new thing and then go all right for those existing behaviors that they're doing.

[00:21:01]Where are our teams hitting walls? Is it the creative side of this step of the process or is it the time spent on these reviewing contracts or uh aggregating customer feedback and trying to come up with proposed next steps for it etc. So I think teams uh the organizations that seem to do well are the ones that find existing things that are already happening in their business and audit them and then try to look at them and go okay I don't need to do a crazy amount of behavioral change if I just in this step of the process I can add an agent to do this or I can add an AI flow to do that. That's really where the biggest gains are getting and the sum of feeling the impact of incredible transformation is you know we believe this as organizations it's through our people and what our the our people and what decisions they make and what what output they produce and so what better way to get to that sum total than helping the groups of people the teams that do that uh and that in aggregate has a pretty compounding impact across your organization.

[00:21:58]>> How do you make that change happen? Like do you just call in all hands and say hey everyone start using AI look at your workflows or do you have a team of AI detectives that goes and sniffs out these problems and helps people automate them like how how can you move it forward the um hey folks do more AI I'm going to set a a goal or OKR for AI that helps I think in in terms of direction and we're going to mandate AI or whatever it is what I've seen have the biggest impact even internally uh we ran an AI builder week where we had a thousand people uh stop what they're doing. So what what did we do there? We created some psychological safety in that one it wasn't some random person in a team waiting waiting there to find some spare time to geek out on something then work out then come and tell the team is it worth their time or not. Uh it was everyone at a synchronized time we're stopping so you don't feel guilty about taking time. Uh two is we're giving people space and we're giving them space to say that success of this week is is learnings. It's not an outcome. You're probably going to get lots of failure. It's just learnings.

[00:23:03]We're just going to celebrate what you learned. When I pick up a new task, I can try a new way of doing it, which I might not complete the task, but I'll learn a lot. It might take me longer, or it might save me a ton of time, or I can do the old predictable way, but it's very predictable and I know exactly what I'm going to get. We need to create incentives that let people try the new way. Uh and so safety is one. Making sure the time is synchronized across all your team members. When I say synchronizing, so example for us is we had product managers and designers and engineering leaders pause for a set period of time to tinker. If it was only one of them, there'd be an incredible amount of guilt that you know you're blocking the team and all that kind of stuff. So I think that's another big tactic that works really well. And what you find within teams, people want to rethink how they could do their job better. they just haven't had the space and as leaders our job is to work out which constraints we want to relax and be okay with that and encourage that behavior and reward learnings. So we would you know at the end of the week would celebrate the biggest fail as a joke but also as a celebration of the most ambitious project that didn't work.

[00:24:03]You know people tried building a bunch of agents to do a crazy amount of stuff and it didn't work but they learned a ton of stuff and we actually changed a lot of in our processes and how our agents work as a result of it. So lots of examples like that. Yeah, that is a great approach. Creating that safety net that everyone feels they have the time and the space to to experiment is fantastic. We actually ran something similar a couple of months ago that we called just AI week. You know, exactly what it says on the tin. And literally everyone at the company got that week.

[00:24:32]We set up a whole bunch of learning workshops. They got the entire week to just try out whatever they wanted to look at their automation workflows, prototype a product that they really wanted to see and just go crazy on it and they absolutely loved it. We're actually scheduling the next one very soon.

[00:24:48]>> Yeah, that is awesome. It it is and I'm super glad you did it. We actually had someone from your team come out as well say share some of that learnings with us too. doing it in in that synchronized way where everyone's doing it together creates a massive encouragement for people to just I use the phrase tinker like we need to tinker and when you tinker it sparks ideas and then you can work out the best ways to improve a workflow or explore a concept or whatever it is. The other thing that stood out to me from the Atlassian collaboration report was that people are reinvesting their AI time savings into collaboration, creative thinking and innovation, not just efficiency. And I'd love to know an example inside Atlassian where you're actually seeing amplification of a team's creativity instead of just automating a task.

[00:25:34]>> It happens a lot. Uh I mean the most recent example from that AI builder week we had was the explosion of prototypes that we had on concepts. I think what blew me away the creativity that came from people that wouldn't normally have had the opportunity to express that concept. So people that may have come from outside the traditional product team or the design team whatever it is they now have the tools to express their concepts and it's been fascinating and they come with a diverse set of different perspectives cuz they don't sit in a traditional team and so like well hang on a second no I think if we change Jurro to do this or whatever whatever it is that the concept they want to do that was incredibly enlightening so to see that happen to see that people who didn't have access to specializations or specialist tools now with AI can uh are becoming creatives themselves. When you see AI used for um you know we have a marketing AI hacks channel where a lot of marketing folks are showing the stuff there when you mix and match the uh quirkiness of uh different cultures. You know we have a lot of like agents inside Alassian that do funny stuff like translate Aussie slang for non Aussies uh or rewrite content in really heavy Aussie slang to the point where I don't even understand it. I'm like what is this saying?

[00:26:49]They'll use AI to be apiring partner to mix and match multiple uh concepts in this case uh different geographical languages to explore ideas for a creative campaign they might launch in an area to try to make that campaign more contextually relevant in a particular region. And so it's that mixing and matching exploration where you would have to have the domain knowledge of those regions or have someone from that region to be able to be come up with that spark. But AI can help you help feed that information for you to help do that.

[00:27:18]>> I think agents can mean a lot of different things to different people and you've mentioned that you use them internally quite a few times now. Can you actually describe some of the more pivotal AI agents that you've deployed inside Atlassian?

[00:27:32]>> Yeah. Yeah. Absolutely. We um by the way I define them really simple as an AI instruction and AI knowledge and AI skills. And the three things combined you would call that an agent like you you've specialized uh that thing there.

[00:27:43]Sorry. I I mean for me personally um or our product teams I should say there are three or four big ones. So for example we have anyone meeting with a customer uh interacts with our customer 360 agent. We have hundreds of weekly users that use that. And what that does is it's an agent integrates with a ton of internal systems to give you one picture of the customer. So we have systems that have the customer sees another system that tracks proposals and RFPs from that customer. Then we have Slack channels dedicated for each customer as well. And there's conversations happening there.

[00:28:12]So when I'm about to meet with a customer, I'm like, "Give me the brief.

[00:28:15]Who's talked to them recently? What are the top things I need to know and it produces a report for me, attaches it to my calendar, you like I can come in prepped and doing that." We have a a customer feedback agent for analyzers. I think I mentioned that one earlier. Uh a lot of the teams here deploy their own in-app feedback contextually. So on any feature, someone might give a thumbs down or a thumbs up or they might hit a feedback button. We get th tens of thousands of these each week. And so how do we find the big trends, identify them and prioritize them on our backlog that is largely now an agent that helps humans facilitate throughout that whole process automatically posting a summary to the team every day in their week in their chat room uh creating a report in Confluence putting new ideas in their backlog in whatever Jirro whatever there is they're using etc and so that happens there as well.

[00:29:02]>> That customer 360 agent sounds really cool. Was it what was the development process for that? Like was it a big here's all the data we're going to pull in and we need to map out the workflow and we need to pull all that together using code or was it just one person randomly firing up GPT or something?

[00:29:21]>> Yeah, it's definitely a little bit on more the complicated side because of the amount of internal systems. So we have you know one of the things we have in our our offering is just customers could connect their data. So the thing we we all customers need to get right or is get your data in shape. AI is only as good as the data has access to. So you need to connect your data central system. So people connect it all the way to Robo and then the agent sits uh the agent uh can now have access to that data. What's amazing about that is that's permissioned per user. So if I don't have access to some certain Slack channels or some uh Google documents where this customer information was discussed, I won't I'll get a tailored result which is quite different to your result. And we had our IT team work through that and and set that up for all the account managers and the customer uh facing teams, the product managers at Alassen and they all have access to now doing that. Some have put it in their automated workflows as well. So they'll they'll put it in as part of their calendaring process or their or some of them will just chat to it ad hoc which is like I'm about to meet with this customer, give me the brief and all that kind of stuff. So getting the data in shape is a big part of that ingredient to I always say it's two things. It's data and workflows to make agents successful. Not every agent needs access to all the data. You just need to be clear on who needs what you need access to what. But it's really those two things are the two big ingredients and the workflows and how you deploy on your team is is the one that gives you the the the operating rhythms and the regular frequency to get the most out of those agents.

[00:30:40]>> You can now get a whole lot more done.

[00:30:42]Like you got access to way more information. It's going out. It's doing its own research. It's bringing it in.

[00:30:48]Sometimes it feels like now we're the bottleneck. Like all this information is coming in. You got agents coming at you.

[00:30:54]they're feeding you all these signals and now you have to make a decision about it or you have to take some action from it and the queue is just you know queuing up. How how do you deal with that?

[00:31:05]>> It actually it's relates to the topic of like people concerned about what about AI jobs as well and there'll be a lot of job changes all but no but you're spot on. There's more work to do. Uh there's a lot more work to do. I was talking to a friend of mine who runs a radiology clinic uh in Sydney and they've just upgraded their systems to be geni driven. the throughput the system can get through now to do a prioritization order of all the uh the health care folks folks seeking healthcare coming in it's getting through a lot more the result he's had to hire more radiologists uh to put them on on board to review to review that etc because there's always a human in the loop and uh and do that but AI does an amazing job and help us prioritize so there's more people coming in and doing that but the same thing you're spot on with our product teams so we're getting a lot more coming in so then what's then more important then for product or software teams to focus on really it's around directing the ship and and the strategy becomes the most important high leverage exercise that we need to do. Um and so if you're getting more and more stuff coming through, you need to make sure that everyone's rowing the boat the boat's you know headed in the right way.

[00:32:08]You could be building a lot of stuff but the boat's headed in the wrong way. And so you would you you know I always tell teams here you're going to your product development life cycles are getting faster. Therefore, your strategic direction, you know, before you had time like as in you had a bit more time to be like, oh, we'll, you know, few months we'll try this and then we'll turn left and then we'll turn right. That's been compressed a bit more depending on what you're doing. And so ensuring that we're course correcting on a regular basis, which is again the importance of having humans in the loop reviewing the work becomes a critical part in in helping making sure we're directing our virtual teammates and our human teammates in in in the right thing. Will we be a bottleneck? Yeah, we will be a bottleneck. But the reality is we want to make sure that the ship is, you know, headed in the right direction. Uh rather than just, you know, heading the worst thing we could do is feel like we're a bottleneck and keep scatter gunning everything without any sense of direction and then we'll just be going around in circles.

[00:33:04]>> Feels like a level shift in the type of work we're doing. And I think you referred to your whole career as moving from building things to architecting things. And it feels like everyone is now being pushed up more towards the architecture end of the spectrum. What do we need to enable our teams to become architects and to learn the proper skills of architecture?

[00:33:26]>> Yeah. Uh great question. Yeah, I would say teams move from doing the thing to architecting the thing. And so what does what makes a good architect? Well, one is that uh they can describe how they would like something done and they could describe why doing that thing is important. uh the why becomes more important because if we all if we're all describing the same thing we end up with what we call AI slop everyone just ends up getting you know the classic example if you if you type the same prompt in the five six different LLM uh tools out there you'll get almost 80% the same it's and it's really the effort and how well you articulate what it is you're after uh and your taste comes through in that so the architect's not only great at communicating they're great at expressing the why and that their their taste be behind that especially In the early days of agents, they're the ones that are are very good also in in showing the agent how to do something.

[00:34:18]You know, it's one thing to say, go and analyze a thousand customer feedback tickets and then tell me the top themes, but you'll get a much better outcome for like here is what I mean by analyze the thousand customer tickets. Step one, I want you to read the title and the summary and then I want you to create a classification system. And step two, I want you then to to then group all the common things into a classification system and I want you to rate them by high severity. like that's the nuance that matters for those architects. And so it's really around getting the teams again to document their workflows uh like how they do their work and how that works well. And those teams that are already doing that um whether in documentational whatever already in a fairly good step ahead in architecting AI uh to work for them.

[00:34:59]>> Looking to the future in your own report executives predicted that by 2030 only a third of work will be fully done by humans. So what does this actually mean for creative industries for for product building industries? Is the future about fewer people or is it about new kinds of creativity?

[00:35:18]>> I don't know about you Cameron, but I'm guessing the team at Canva has a lot more ideas they want to build than actual people do. Is that a fair assumption?

[00:35:26]>> It would be a fair assumption.

[00:35:27]>> Yeah. Yeah. Um we we are no short of things we want to do. Uh we continue to grow. So um yeah, I think you're spot on. It it it absolutely means but that um uh the companies that want to continue to grow will continue to grow because they've just got so many opportunities ahead of them. Um there are many companies that will be created because of these opportunities like a whole bunch of new industries that will uh be established as a result of it. Uh and at the same time um AI will help train the next workforce as well. Like we often we often go into this cycle of like oh okay so then what what happens with junior folks like oh do I only need senior architecty folks like no actually AI is an amazing teacher as someone who hires a lot of product managers at Alassian the our hiring profiles you know I'm thinking of like shifting how we hire where you know to put quite bluntly I want the kids that have been cheating on their on their assignments at university with chatbt like they're AI native like my son who I was telling you ear earlier who was using scratch coding he's gone from scratch straight to vibe coding. He he hasn't had that window of like coding in actual syntax. He's going to grow up in that just just that that native world where natural language is the coding language, right? So, I think that whole workforce for us is is exciting because one, those that have high agency and are willing to learn are in an amazing position uh to make the most of it. And for businesses out there, you're going to be able to get a lot more uh with the folks that you have.

[00:36:56]>> Well, you heard it here, folks. If you're cheating on your exams, apply for a job at Atlastian.

[00:37:02][music] I think that's enough diving deep. I'd love to change gears with you and talk about your allstar creative team. So, in this challenge, we get you to pick four people. Any era, any universe, dead or alive, put them together, build your dream creative team. Who would you pick?

[00:37:23]>> Oh, four people I would pick. Uh look uh my wife is is we you know anything that I try to do some any sort of interior design or landscaping whatever I think I have taste and then I chat to her I'm like oh man I like your taste better. So I'm going to throw my wife in there. I was watching the Spiderverse Spider-Man movie with my son. Uh man I loved whoever animated that. It was just just a different style of animation. I wouldn't have a clue.

[00:37:51]>> It was mold raping. It's a brilliant piece of work.

[00:37:53]>> Yeah. Just and the music was off the charts too. Like it was just one of those ju just incredible work of art like uh that movie. So whoever worked on that from bringing people from the dead.

[00:38:02]I'm bring back Jesus Christ. He can uh he can pop some miracles and uh you create some beautiful scenery for us.

[00:38:08]I'll bring him as well. Oh Johnny Iive.

[00:38:10]I mean if I can pick anyone bring Johnny as well. Good old mate Johnny. He's done a couple of things in design as well.

[00:38:14]I'll throw him in there too.

[00:38:16]>> All right. I don't know how Jesus and Johnny IV would get along but I'm fascinated to see to see what the workflow looks like.

[00:38:22]>> With a lot of love. A lot of love. Well, thank you for outlining your creative team. That was some good picks.

[00:38:28][music] I think it's time to bring back Avon.

[00:38:34]She's had a bit of time to work with your little comic sketch and I am keen to see what she has filled the other five boxes with.

[00:38:40]>> Okay. Hello.

[00:38:42]>> How did you go?

[00:38:44]>> I just finished it. So, thank you for sharing your story.

[00:38:49]I hope this video, this creation does it justice. I thought about the story you mentioned about your daughters wanting ice cream. So, there's some elements of that in this uh very interesting world.

[00:39:03]>> You paid attention to everything, Avon.

[00:39:05]I don't know how you worked and listened. Let's see what you've made.

[00:39:09]>> All right. I hope you enjoy this.

[00:39:37]Ivonne, that is amazing.

[00:39:43]Wow. That's it.

[00:39:45]>> What a spectacular piece of storytelling. You had everything in there from little share crafting away in classrooms through to the worlds coming to life and just brought it home with the ice cream. Incredible.

[00:39:57]>> That is amazing.

[00:39:58]>> Has to end with an ice cream monster, doesn't it? For the >> strawberry vanilla it looks like.

[00:40:03]>> If you're listening on the podcast, definitely head over to the show notes.

[00:40:06]will link Ivon's uh amazing creation describing basically all of Sharief's life in 30 seconds. We'll give you the link and you can check it out yourself.

[00:40:16]Thank you so much Ivon.

[00:40:17]>> Thanks Ivon.

[00:40:21][music] >> Thank you Sheree for joining us on this episode of Prompted AI people in the creative spark. It has been amazing having you here and talking about all things collaboration, ice cream and scratch coding.

[00:40:34]>> Thanks Cameron. Thanks everyone. See you folks.

[00:40:36]>> And for all those listeners at home, please subscribe wherever you love to listen or head on over to YouTube. Until next time, keep chasing that creative spark. [music] [music]