Shatter the service quo: service management for the AI era | Team '26 Europe | Atlassian

0:31 Please welcome Atlassian head of product marketing service collection Molly Bronstein. Hello Amsterdam. I'm so excited to be here in a new city for team Europe this year. Thank you so much for coming. Yes. Now, when Carlos said in that video that we just saw that we need to shatter the service quo, he was talking about breaking a vicious cycle because for too long service teams have been trapped in a reactive d reactive doom loop, endless break fix firefighting, siloed systems, lost context, and stale knowledge bases.

1:18 Customers and employees feel it too with rigid, frustrating forms, portals no one can find, status chasing, and ticket limbo. These legacy systems used to bring order to chaos and enabled your teams to scale, but they were built for a pre-I. We're now in the era of AI where the pace of change is staggering. A recent Forester study found that only 21% of companies use Agentic AI for service and operations today.

1:54 But over the next 12 months, that number is expected to grow to 80%. This is an industrywide inflection point. And bridging that gap requires more than just bolting a chatbot onto a service desk. We believe it requires three fundamental shifts in how service gets done. First, service has to be embedded. Support meets people right where work happens. Second, it's got to be proactive, catching and resolving incidents before they ever disrupt the business. And third, service management has to be contextual, drawing on rich organizational data. So humans and AI agents can act with speed and accuracy.

2:44 That's where service collection comes in. Now with service collection service is completely embedded. Whether you're an S sur managing an outage or an employee or customer looking for an answer, help is available wherever they are. Service collection also takes teams from reactive defense to proactive offense.

3:19 things like resolving incidents, identifying stock shortages, or intercepting requests before they ever disrupt the business. And the best part is that everything sits on top of the teamwork graph, which maps the relationships between people, work, and tools across your organization. So that every single service you deliver is actually an act of context creation which ultimately means that service becomes faster and smarter.

3:53 Embedded, proactive, contextual. That is how you shatter the service quo. Okay, now let's see it all come together. Meet Beacon Stone, our fictitious bank. Over the next 30 minutes or so, we'll follow three people who keep Beacon Stone up and running. Hector, the ITSM manager, Austin, the principal S, and Priya, our customer support agent. With that, I'll hand it over to Shemik to get us started.

4:30 Thanks, Molly. Let's take a look how how we're rethinking the employee service experience. The service quo way of building a service desk is far from magical. Configuring request types, a knowledge base, custom workflows. All of this can be really time consuming and all this setup friction just delays business impact.

5:00 With solution composer, any team like HR, legal or finance can describe their needs in just natural language and the solution composer will create the service desk in seconds. It pulls all the stuff that it needs from the teamwork graph. Let's see it in action. Hector's Monday morning starts with an ask from HR. Beaconstone is setting up a new office in Amsterdam and the people team urgently needs a new travel service desk to be set up before the office opens.

5:38 Now, historically, the way this thing would work is that HR would file a ticket with it and would wait for weeks while custom fields, workflows, and permissions are set up. However, with solution composer, Hector can hand the reigns to HR, the HR admin can just tell Robo exactly what they need in natural language and Rovo picks up the right components and in seconds, the new travel service desk is built. It's that easy.

6:15 What's amazing is that Robo draws all the appropriate context. Which teams exist, what are the right privileges, what are the projects that people work on, and what travel policies there should be. All of this information from the teamwork graph, and it sets up all the right components in the travel desk in seconds. And making edits is simple. The HR admin can tell Ro exactly what they need. For example, in this case, a new request type for booking travel request.

6:49 And boom, there it is. The HR team gets a fully functional travel service desk without having to do complex configurations or involving it. Now, getting your service desk up and running in record time is only half the battle. You also have to make sure that the right people are in place to support the service desk. Intelligent workforce management is our new feature. It has schedules, realtime availability, and intelligent routing.

7:23 All built natively into Jira service management. No add-on needed. Let's take a look at how that works. Beacon Stone has announced that a new IT support coverage policy is now going to be in place and Hector needs to make sure that the right coverage in place to meet SLAs's. So he takes a closer look and he spots a staffing gap. Now he can easily edit the shift and assign it to an available team member. This improves coverage and make sure that the SLAs's are met.

7:59 He can also dig into views of his team's capacity and availability to determine how to optimize the team's workload. All of this makes it super easy to rebalance work. For example, when Hector looks at his L1 support team, he notices that one team member is overloaded. With just one click, he can rebalance their workload and reassign that work to others in the team who have less who are more available. Behind the scenes, Robo and smart routing actually looks the at the previous history of people's workloads and their availability and their capacity and the teamwork graph to make sure the work is landing at the right place.

8:50 Talking about right place, increasingly work doesn't get done in traditional ticket cues. It's right inside the help providers everyday workspace. It could be things like Slack, MS Teams, and now increasingly clawed so that Hector and people like him can manage their work right where they want. Let's take a look here. Hector asks Claude, "What are the most urgent things I need to do for today?"

9:24 Claude reviews all the open requests in Jira Service Management and provides an assessment of the priority that depends on the SLA risk as well as the potential impact of that particular ticket. Among the list of urgent requests that Hector sees, he sees that his company's CEO has a Wi-Fi problem. just before he's about to go on stage. Well, that seems pretty urgent. But it also gives him a particular resolution plan directly inside Claude.

10:02 If Hector so chooses, he can action upon that resolution plan right from within Claude. But in this case, let's imagine that Hector wants to take a closer look at the suggested plan and see whether it's the right one. So he goes back into Jiraa service management and he can look at the resolution plan that has been created by the request resolver that comes with Jiraa service management. This is one of our out ofthe-box agents in the service collection.

10:31 He can look at the request and the proposed step-by-step act actions that it's proposing. All he needs to do is to approve the plan. The work gets done and the CEO's Wi-Fi is back. Now, resolving a request in seconds is obviously a massive leap forward. But what if we took it one step even further and we could deliver support to employees before they even know that an issue has happened?

11:03 With proactive service management, Hector's team doesn't just respond to requests when they come in. they can actually stay ahead of them when an application crash for example starts hitting laptops across beacon stones. JSM's new realtime endpoint telemetry system detects these failures. It automatically triggers self-healing steps in the background and resolves the issue before a single employee even has time to even file a ticket. Problem solved with no action required from any employee.

11:45 Behind the scenes, however, Hector's team gets full visibility into device health. They can see how many issues there were that were created per device type, the trends over time, and the remediation success rates, all in one single place. With proactive service management, device issues are detected, resolved, and routed to the right people before they ever become a request. That's pretty amazing.

12:19 We have seen now how it service can deliver frictionless employee experiences. But whether it is opening up a new office or managing critical infrastructure, success often comes down to one thing and that's the right context. Let's talk about how your service asset and configuration management data can change from just being a static graveyard which you put data in and then you look at in your dashboards. From that it can become a foundation for new kinds of agentic workflows. I'm sure everyone in this room is familiar with the pain of traditional CMDBs.

13:03 But with service collection assets is now connected right with the teamwork graph. This means that every asset, every CI, every custom object is linked to the teams, to the projects, to the the people that are connected to those assets. So when you need to change an asset or you need to take it offline, you can know who's going to get impacted and proactively inform them. So how do we get all this information into assets?

13:36 It all starts with discovery. Now, we know that discovery is not a one-sizefits-all kind of solution, which is why we provide multiple options for customers. We have enhanced native discovery that comes inside JSM. But for complex enterprises, we've also deepened our strategic partnerships with both Flexera and Lansu. These discovery engines scan your entire network and all your entire estate to find every device, the OS versions and even shadow IT applications that are there across your network.

14:20 And they stream all of that attribute data for these assets directly into the teamwork graph. This is all done in real time. The best part is you don't have to start from scratch. All this incoming data is mapped into our common data model which is now generally available. This gives you enterprise structure from day one with all the flexibility you need to customize it and extend it with the right attributes that you choose.

14:56 Now all of that clean data and the automated discovery is now coming together in one particular application that we're launching called hardware asset management. Let's see how Hector uses the hardware asset management console to manage Beacon's Amsterdam office. Now as new team members come into the Amsterdam office, Hector doesn't have to wait for equipment shortages to escalate. right in his hardware console, Robo is flagging that there is a low stock alert.

15:32 So, Hector clicks into the alert and instead of pouring over spreadsheets and trying to connect them to one another, Robo analyzes the stock room telemetry data across all the offices and pinpoints that a critical laptop shortage is about to happen in Amsterdam. It also recommends the next steps. So, Hector asks Ro to generate a device fulfillment request for this laptop.

16:04 Instantly, Robo will draft a procurement request that has all the details, the hardware specs, the vendor details, the delivery address, the budget, etc. In the team's shared workview, this new procurement request appears and Hector can click in. Sorry. And Hector can click in, verify the purchase against the budget that he has and approve it with a single click. Now the order is dispatched and the complete life cycle tracking is established within this hardware console right from the moment of purchase all the way till completion.

16:49 Now when the shipment actually arrives in Amsterdam, the inventory is updated automatically. Hector can filter his view and he can see every laptop has an asset tag. It's mapped to the Amsterdam office in the common data model and the inventory levels flips back to green. Now all of this information also lives as context in the teamwork graph. So now your hardware is connected to new incidents that might happen, new requests that might come in and to the right teams and projects that that particular asset is being used for.

17:27 This ultimately helps IT teams deliver better, quicker support. Now let's shift gears to operations and let's meet Austin, Beacon's principal S. Historically, operations has been purely reactive. An alert triggers at 2 a.m. An on call engineer wakes up, joins this particular conference bridge, and Slack channels fill up. And teams spend hours guessing what just broke.

18:03 The reality is 80% of all unplanned outages is because of poor change management processes. Now, when development is moving fast and operations teams are left in the dark, that's where every deployment is a roll of the dice. Anything can happen. What if your platform could predict the risk of a change before the deployment happens, before it touches production?

18:37 Now, that's why we have built AI change risk assessment. Now, let's see it in action. In this case, Beacon Stone's promotions team is releasing a new welcome offer for new shoppers. Now, Austin looks at this change request and kicks off a risk assessment directly inside JSM. Rovo draws upon the teamwork graph and evaluates this change across four risk factors.

19:12 Technical risk, operational risk, compliance risk, and business risk. In this particular case, it comes back with red high risk across all four categories. Now, Austin now needs to dig deeper before he can approve this change request. Now with his assessment, Austin inspects each of these individual risk factors to understand the details of what's risky here. Now the teamwork graph pulls together context and shows that the playbooks past PRs coding standards what are the problems with this particular change. It even identifies gaps in testing or the lack of roll back plans.

20:00 It also gives Austin suggested remediations, actual concrete actionable mitigation plans that can derisk this release. With one click, Austin is able to kick off that work for the promotions team to handle. And once they have addressed these issues, then only would the change get approved. So now in this case the promotion team has come back with the changes and Austin can rerun the risk assessment.

20:34 These mitigations have actually done the right thing. The risk score has now dropped to low and Austin can feel much more confident about making this change deploying this change. This lowers the risk and leads to fewer incidents thanks to all the context that Robo pulled together from the teamwork graph. Now, I'm sure that any one of you who's been on on call can agree that the best incident is one that never happens.

21:10 But in every organization, right, unexpected failures still happen. Now, when an outage does happen, how do you resolve it as quickly as possible? What we're doing is we're bringing the entire incident response life cycle directly to where your engineers already collaborate right inside Slack. Bad news here. Beacon's payment system just went down. Now, because this is a major incident, JSM automatically spins up a Slack channel and packs it with content, an incident summary, on call responders, service dependencies, and a timeline.

21:54 Instead of spending hours trying to hunt across and gather all this information from different tools, Austin just needs to click on Robo to investigate. Rovo taps into the teamwork graph which unifies data from all across Atlassian products as well as third-party monitoring and observability tools. It pinpoints the root cause in just seconds. it was a recent PR. Now, Rovo can answer follow-up questions and even be tasked to drive the mitigation plan based on the details of the investigation.

22:33 All of this work has happened without anyone ever leaving Slack. But let's say that Austin actually want to take a deeper look into the suggested mitigations that Robo had suggested. With one click, he can he's back in JSM and details from the investigation and RCA that came in SLAP have been automatically synced back into the incident command center that's in JSM. Austin can access the full service graph and confirm which are the affected components.

23:09 He can see the full mitigation plan, the confidence score from Robo that this is the right steps to take. And once he feels good that what he sees is the right mitigation plan to execute, he can actually invoke a PR in GitHub. Now Robo can do the work instantly, creating the draft PR for the team to review and to publish. This is how modern service operation should work.

23:39 Predicting risk before you ship and swarming where you work and resolving outages quickly. I want to close with a couple of announcements that are particularly important for the larger enterprises. The first is AI governance. In order to make the most of AI, teams need visibility and control into what AI is doing at every level, even down to individual spaces and individual service desks.

24:17 In the founder keynote, we introduced two new governance capabilities in Atlassian Atlassian administration and and within Atlassian studio as well. We are also adding some new features for Jira admins directly in Jira service management so that each admin can see, manage and control AI for service and operations work as well. And finally, we know that meeting your compliance obligations is a big part of deciding which software you can trust.

24:54 Everything you see behind me is available for Jira service management today with more compliance features coming soon across the rest of service collection. I have just stepped through all these latest advancements for employee service and for operations teams. Now I'll hand it over to Edwin to talk about the people that we all care the most about, our customers.

25:31 Thanks, Shamik. Now, we've all been there as customers trying to get help. You're stuck on hold for hours or forced to fill in giant forms just to ask basic questions. And every time you bounce from one agent to another, you have to repeat your entire life story from scratch. At the end of the day, all you want is to get your problem solved painlessly. But for so many organizations, this seems like such a daunting task. Why?

25:58 Because we treat customer service as an isolated island, expecting our frontline teams to work miracles without the full context. But this doesn't have to be all on our customer service teams. The whole organization actually has context that can help. Bug fixes, incidents, features, documentation, release notes, all shape your customers experience. And these are all parts of Alassian's teamwork graph, which gives our solution the context it needs to be truly intelligent.

26:32 Now on top of this foundation, we bring together your support teams with workflows, cues and insights. And layered above that is our AI agent which meets your customers across every channel to provide immediate and contextual support wherever they may be. Now let's dive in and look at our customer service management app in action live here on stage. So to do this, I'm going to play the role of a customer of our fictitious bank, Beacon Stoned, that you've seen earlier. And I have a problem trying to log into the Beacon Stone app on the phone. So what do I do? I'm going to call up Beacon Stone support. Let's see how we go.

27:26 Hi, you have reached beacon zone support. How can I help you today? >> Hi, I can't seem to log into the app today. Is something wrong? >> That's frustrating. Let me look into that for you. Yes, Beacon Stone is investigating an issue affecting some customers ability to sign into the mobile app. The website, payments, and card services are operating normally. Anything else I can help you with?

28:02 >> Actually, yeah. I have recently ordered a replacement card, but I think I might have just done it twice accidentally. Can you check for me, please? >> Sure. Give me just a moment. I can help with that. Let me look into it for you. I can take care of that. First, I need to verify it's you. What email is on your account? I've checked. The mobile app has a partial outage and Beacon Stone is still investigating failed signins for some customers. Your account, payments, and debit card services aren't affected. And you can use the website while the app issue is being worked on. Anything else I can help you with?

28:44 >> Yes, I was trying to check my duplicate card order. Can you check on that? >> Sure. Give me just a moment. >> I can check whether you have more than one replacement card request. What email is on your Beacon Stone account? So, it is princess_buttercup@hotmail.com.

29:18 >> Let me check that for you. >> I heard princess buttercup at hotmail dot. what comes after dot such as com? >> You're right. Dot.com. >> Sure. Give me just a moment.

29:49 I've sent a six-digit code to princess buttercup at hotmail.com. Please tell me the code when you have it. I can't place an outbound call. Please say the part after hotmail dot so I can send the verification code. >> I have the verification code here. It's 483920.

30:25 Thanks. You have two replacement cards in progress. An expedited one requested on October 5th and a standard one requested on October 1st. Shall I cancel the standard replacement? >> Yes, please. Done. I've canled the standard replacement requested on October 1st.

30:56 Your reference is CX914. The expedited replacement requested on October 5th is still in progress. Done. I've canled the standard replacement requested on October 1st. Your reference is CX990. Your expedited replacement requested on October. >> It did it. >> Most of it. I promise it worked a second ago.

31:27 before everyone else everyone is going to email me a princess buttercup. let's actually dive in now and look at what the actual app did on the back end. So let's switch over now to look at the CSM app. And what you see here is our con conversation review screen that allows us to look at the transcript of every single chat and every single conversation that you have had with CSM. So as you can see here, this is my conversation that I just had three minutes ago, which looks about right.

31:58 And as you can remember, the first thing I asked about was the fact that I had a problem logging into the app. Now, in this reviewer, I can actually look at every single step that AI has taken. So, here I can actually look at the steps that it took. And what it did was it actually used context from the teamwork graph to figure out there was a live incident and it was able to query status page for me. And you can see that on the right hand side there. Next, I asked about duplicate card orders, and it actually required a little bit more involvement here. And this is where skills come in to our AI agent. With skills, we can actually automate more complicated tasks and even chain some of them together to form like a mini workflow.

32:41 In this case, it was able to actually take me through a verification loop with a bit of persuasion, and it asks for my email as well as that two-factor authentication code. And once it's done that, it then finally got me into actually pulling the details. So it able to continue that workflow to pull the details about the card requests and ultimately cancel the card for me after I confirmed that I wanted to do that. So here we are. Here's the step that actually cancelled the card for me. And you can actually see if you jump in here the details that it used including the email address and actually the request ID I have.

33:24 Moving on. Now let's take a more complicated example. This time I'm going to play the role of a merchant. and I am going to be the owner of a coffee shop. actually using Beacon Stone's point of sale terminals. Those things where you tap your card on to pay. Unfortunately for me, that terminal's down during the morning rush hour. And because I have my hands full making coffee, I'm going to contact Spacon Stone support via WhatsApp. CSM's AI is able to respond in exactly the same way with the same context and knowledge you would expect from every other channel. But this time, money's on the line. I'm losing money.

34:05 So, I want to talk to a customer service rep on the team. And the AI agents is able to take that up and actually escalate that for me. and right away bringing in a team member on live chat. Now, the beauty about this is I stay in the same WhatsApp chat. So, I don't need to go somewhere else or find another form to fill in. And when Priya, who is our customer service rep on the Beacon Stone team, picks up the live chat, she doesn't have to start with, "Hey, can you explain your whole life story again?"

34:38 She has the full transcript, the customer profile, and all the relevant information right in front of her. Now, if you look on the right hand side of this conversation, you'll see something really cool. Priya can actually look at that point of sale terminal in this coffee shop because it's tracked as assets on our platform. And because assets is really like our living context engine, she sees everything that's happened to this particular terminal. In this case, we've actually found the likely culprit. A maintenance update was actually pushed out last night, and I bet you that's what caused the problem.

35:16 Now, to help our coffee shop get going again, Priya can use Rovo to pull an emergency runbook for the device, which gives her instructions on how to actually restart this terminal in a safe mode. Priya can actually also just with one click share these exact same instructions back to the customer right in the chat here. Now because this is going to affect lots of terminals, we also want to ensure that we create an incident for the dev team so they can patch this bug quickly.

35:48 Priya can again use Rover to do this with one click and all the context from this conversation will be recorded with this incident. This is true endtoend service resolution. You can chat with your customers. You can look at your hardware through assets all the way to an engineering fix all tracked on a single platform. All the capabilities that you just saw demonstrate the breadth of our solution we have already and how we can help you solve customer problems faster with the power of teamwork graph and our AI.

36:23 But really, none of this is set and forget. You can't just launch it once and expect it to work. Continuous improvement is essential to any great customer service operation. And when we set out to build CSM, we made a conscious choice. We weren't just going to build another static agent. Rather, we've built a system of continual improvement that keeps learning and gets smarter with every single interaction that it has. And the first step to this is to observe what's happening with the system.

36:57 We bring the data and the signals that help you observe what's going on. The traffic you're getting, the containment rates, the resolution you're getting from AI. And through the conversation review that you've seen earlier, you get complete visibility into every interaction. You can drill into any chat to see the exact reasoning chain. So you know exactly why AI arrived at the responses it did. and we'll check on that later about the one that we just had. So nothing's ever a black box for you.

37:26 And once we've identified the gaps, we can actually improve. Now for many traditional teams, this is actually one of the biggest problems I keep hearing when I speak to our customers. Where do you start? How do you even do that? This is where our optimize feature comes in. Optimize will detect gaps for you, analyze successful resolutions that are made by your team and use those insights to draft new knowledge articles. These then can go through a human in the loop approval flow before it's published. And once it is, your AI agents can then ingest this new knowledge to improve its performance.

38:03 And this is great, but how do you really know it has gotten better and more importantly that it hasn't gotten worse? Well, of course, you test it, right? We can all do manual testing, but it's really slow, tedious, and chances are you're going to miss out on a whole bunch of stuff. In CSM, before any change is made for a real customer live, you can run your agent against an evaluation data set in a safe sandbox environment. And then you can compare that performance and its accuracy against the benchmarks that you've already established.

38:37 And finally, once we're confident on our improvements, we can deploy it live to our system. And this is super easy in CSM. We have a version control system that allows you to manage this. All the changes are made in a draft version of your agent. But when we are ready, we can simply push this publish button here. And all your AI agents will receive the latest updates across every channel, live chat, voice, email, or social. I've just showcased many new and exciting features that can help you deliver a frictionless and connected customer service experience.

39:15 The CSM app is available to all customers with a service collection subscription and all the capabilities that you're seeing today is available in the app. So, please give it a spin and see the improvement in your customer service. And that wraps it up from us. I hope we have given you some ideas for how service collection can help your teams reimagine service to be more embedded, proactive, and contextual. And for everyone ready to shatter that service quo, scan our QR code here to find out more about our vision for the future of service. Thank you for your time today and enjoy the rest of the event. Thank you.

39:57 >>