Welcome back to the Teams Insider Podcast. This week sharing a great conversation with Chandra from Microsoft, talking about agent readiness and your team structure in readiness for agents in Teams. And we talk a little about his role and his perspective on what's going on in the space. Many thanks to Chandra for jumping on the podcast.

Also many thanks to AVI-SPL who are the sponsor of this podcast. Really appreciate all their support. On with the show. Hey everybody, welcome back to the show. Really excited to record this one. Um, with Chandra I keep having massive conversations in the prep calls, uh, so hopefully we'll get some of that on the recording now.

Um, for, uh, for those that don't know you, I know you've been on the show before, but do you just wanna reintroduce yourself and, uh, your role? Uh, yeah, of course. Very, always nice to be on, you know, this, uh, show, Tom. My name is Chandra Chivukula. Um, I work at Microsoft. Um, I work with a brilliant set of engineers and, you know, uh, folks, product and design.

We're building, uh, some of the important pieces that are shipping as part of Microsoft Teams across the Windows client, Mac client, web client, iOS, Android, and all the good stuff. Um, so we, my team and I, we build the core infrastructure, the, the framework, the, and some of the experiences that we are shipping for our consumers as well as commercial users Yeah, we've had, uh, different people from your team on in the past talking about the, the Mac experience, the mobile experience.

Uh, and it's a-- there's, there's a fascinating conversation to be had, just the, the, the crazy scale of Teams and then the innovation over time. Um, but today we're kinda gonna zero in on the, the conversation around, uh, th- kind of the future of Teams, and Teams as a human and agentic platform. Yeah. Like, from, from my point of view, obviously it's, it's really exciting.

When the, um, OpenClaw kind of thing spiked off, in particular last year- Mm-hmm ... um, not that long ago in the grand scheme of things, suddenly it was like, oh, people like the idea of engaging with agents through their messaging platforms. Like, that is a really interesting engagement model, and that was immediately like, well, light bulb, hang on, there's 350 million give or take MAU on Teams, and that's the enterprise platform for collabs, so hang on, there's something here.

Um, but maybe I get your, your view of that. Uh, this is, uh, fantastic. It is phenomenal, very exciting for us. Uh, I am part of some of the crew that is working on this. Uh, uh, we should get some of my subject matter expertise on agents and the agentic platform onto the show. We'll just do this in a, in a subsequent show.

But for, for the most part, uh, it was very exciting to see, you know, OpenClaw, uh, launch and how much of an uptake it has had. Uh, but you know, we have a significant chunk of enterprise customers, mission-critical customers who are using Teams for, uh, their critical ex- workflows. Uh, we gotta be very responsible towards those customers.

We want to bring agentic infrastructure to them, but we want to bring it in a way that doesn't add an overhead, uh, so that it is compliant, it is secure, so you have this, you know, trustworthiness that you have with Teams, that your data is secure, uh, your interactions are secure. We also want to bring the, uh, agents along with that, uh, safety net, uh, in there.

Um, hopefully you all have caught up, uh, at Microsoft Build, the, the announce and the launch of Microsoft Scout, uh, which is Microsoft's on-prem, uh, you know, in, sorry, in-cloud, uh, uh, version of OpenClaw, which is running in the M365, uh, uh, corporate boundary, compliance boundary. Uh, which is a phenomenal tool.

We use it all the time. It's a separate app, but you get an idea of what a taste of an agent with that capability looks like- Hmm ... especially when it has, uh, safe, compliant access to your corporate data, with, to your access data. So the amount, the, the kind of things you can do, the kind of automations you can set up, the kind of skills it lets you create pretty effortlessly is pretty mind-boggling.

And, uh, for folks who don't know Microsoft Scout, it is Microsoft's, uh, uh... You know, we, we partnered with OpenClaw- Um, and there's a fantastic team under Omar Shahin who, um, who brought that to life inside the Microsoft 365 boundary. So they've, you know, done this due diligence of compliance and other good stuff, uh, that makes the Scout app, uh, way more safer, uh, and way more trustworthy to use.

So you can actually do some real productive, um Uh, workflows that you do in your normal day-to-day, um, you know, work experiences, you can kind of defer and bring Scout along to do more of automation. Now, that's just one example of, uh, an agentic system that is at play inside Microsoft 365. Clearly, chat is the way you interact with the agent in Scout, uh, but in its own app.

Like you said, we've got, uh, 300, you know, 40 something, uh, million people or higher, um, using Teams on a monthly basis, and many of them, many of them using mobile clients, uh, desktop clients, all in different perspectives. You know, how do we bring agents to you, uh, as a mobile first user, um, is going to be a different track than how do we bring agents to you in the conversation, in the flow of your conversations.

Hmm. So it feels like a, you know, a, a supplemental somebody who's helping you with the work rather than somebody who constantly jump- jumps in with, you know, random piece of information that sometimes is useful, sometimes is not. Uh, and so that's one of the things that is going on. You might have seen, uh, our audience might have seen some recent announcements from Microsoft.

We talk about Facilitator, which is a meetings agent, uh, which is a phenomenal value add. Lots of people who tried Facilitator love it. Um, and it's- Yeah, we had Han Yu on the, on the Fireside Chat, and it was one of the ones afterwards I got the most messages from afterwards being like, "This is really exciting.

He explained it really well." And like, like he gave a bit of a hint about where it's going as well, and it's a, it's a- There you go ... problem space that everybody immediately understands. Like, I have lots of meetings, it's a multiplayer game. How do we get AI to help us in that scenario? Yeah. Uh, we should definitely get, uh, some of my other folks, uh, my other colleagues on the show, uh, to see how this entire agentic platform, plus how you structure your teams, your chats, your channels, um, become a critical investment in decision-making so that your agentic systems can make better use of it and help your engineering- Hmm

help your product, uh, help your teams when you start employing them. Um, it's very hard to surgically implant agents in the way we work and expect to see some significant growth. Uh, that part is clear. And one of the things that my team is doing, uh, since the past few, uh, let's say, uh, quarters, is, uh, you know, I'm on the engineering side, so I, I, I'm really fascinated with how coding agents have come, um, on board, have become matured significantly in the past, um, eight to 10 months, uh, especially with Cloud Code and Copilot CLI becoming really state-of-the-art.

Hmm. Uh, we had access to Cloud Code, we have access to, uh, Copilot CLI. Now we are all in on Copilot CLI because for all practical purposes, we're, we're very happy with CLI. And the Uh, the, the buffet of models it has access to. It's not just Anthropic models, it's also OpenAI, it's got MEI, uh, new coding model as well, and any new model that you may want to set Yeah, the, the- there is a real benefit to having model options and having models check, check each other.

Like, I'm not- Exactly ... a d- developer, but I, anything I do now, I usually have, like, if, if I'll have the, the Claude models create it, I'll have the OpenAI models check it or vice versa. And that's brilliant for me as a non-developer 'cause I'm like, I've got two different sets of intelligence checking over the work.

So having a CLI that can mix and match models is, is really beneficial. Uh, you hit the nail on the head. Like, that is the biggest value we see, uh, especially when you want to create, um, a feature change, a pull request that is nearly flawless. You know, uh, no engineer likes reviewing someone else's code and finding all kinds of, you know, defects or, you know, uh, problems in there.

Mm. It's just not good use of their time, just wasting of resources, time, and, um, effort. Um, and we've gone through that phase. In the past two and a half years, all kinds of different coding assistants and coding agents have come and gone. Uh, but with these new models plus the, the harnesses that we have built on the engineering side made it very, uh, very easy for our engineers to, you know, quote, unquote, "harness the power of these harnesses" so that they are more effective.

Uh, what we have seen in business land and, uh, as well as in engineering land is, uh, there are, uh, unique segments, unique profiles for people who are really successful at using AI Uh, in, in engineering land that I'm most familiar with, uh, we have got some really successful, highly qualified, highly subject matter expertise, uh, principal engineers, uh, who are excellent, who are already excellent and who are now getting, you know, 5X, 10X productivity, uh, through using this coaching agent can harnesses.

And likewise, on the complementing on the other side of the scale, we have fresh out of college, you know, one or two years in the industry who, uh, engineers who are absolutely fascinated with this tech, and they go research not just the how to use the model, but how to best use the model to find out- Hmm

where the model is great and more importantly, where the model is not great, so that they can mix and match various models to get the best of all rather than, you know, you get averaged out to what the model can do in all scenarios. So it's kind of a profile of, of people that are developing. It's not really, uh, the age in the industry that, you know, maturity in the industry that counts.

What really matters is are you able to latch onto the wave of these new models. Yeah, it's kind of adaptability, isn't it? Like, uh- Exactly ... and I guess like en- energy that we were talking before on the prep, like there's been some massive jumps in the model's capability. So if you checked it out nine months ago, it, it's a completely different story now with the newer- Yeah

with the newer frontier models and with sub-agents and, and the, the paradigm of how you use the agents has, has changed as well. Uh, absolutely. Uh, exactly right. There's deep research that is needed to understand the, the models, what they're capable and what they're not capable of, and then be prepared to do this in six weeks again with the new batch of models that come out.

Uh, you know, there's heavy competition there, which is great, uh, for the industry, but it also means it's a different line of thinking. It's a different line of optimizing for engineers or product makers, uh, to see where these models are capable of. We have vast majority of our product team Um, who have, uh, who are using AI tools to envision what a, a, a feature looks like, uh, for the, for the future.

How do we reimagine what feature making looks like? You know, product making looks like. Same thing with designers. Uh, we want to have a, a friendly rivalry between our product design and engineers where, uh, the design says, "Hey, I've got an idea. Just dreamt about it last night" And then- Mm. ... you know, four hours later we have a, a thing that they can click on in product- Yeah

that they can evaluate, is it a good idea or not so much? Yeah. And we can learn from it, right? Yeah. Uh, so that loop is- And we, and, and we see the flip, the flip side on the kind of, uh, PM product owner side, which is actually they can, like, mock things up and be like, well, like, like, like previously it would've been, like, engineering, like...

You know, it's a lot of work to prove that out. Yeah. It's like, well, now I can proof of concept it myself and, and bring it to engineering and be like, "Here's a, here's a... It's not production ready, but, like, here's a mock-up." Perfectly said. Uh, exactly right. And that's happening on the engineering scale, and then, uh, I think the same loop is gonna play out in other fields as well.

Uh, but I see it very clearly raising, uh, the quality of our output, uh, in the past- Mm ... I'd say four months or so. It took a while for us to build our harness With our own knowledge base, with our own self-running loops, uh, you know, recursive le- self-running loops so that every bug fix, every feature we do adds to the corpus that helps make it easier for the next feature to be built, next bug fix to be built.

Um, so we don't create the common patterns of mistakes. We, uh, the hardness helps discover potential issues. Um, and that self-learning referential loop is now expanding onto other fields. You know, what do you do when you have easy-- in a meeting, uh, when people are discussing a new con- three or four concepts, would you-- how would you invoke Facilitator so it can spin off doing a thing, uh, that previously people would have taken a follow-up on from the meeting?

Hmm. Uh, so that while you're in the meeting, you know, a, a minute or two later, Facilitator comes back, "Here is my short report. You wanted me to go check on that." Like, it's not just the availability of the tool, it's also the invocation of the tool that is necessary from, from humans. Yeah, I had, I had a wow moment, I had a wow moment with Facilitator, which was, um, with its, its voice modes.

Uh, we were doing a, a dev backlog call on our, our product stuff, and it was like, "I bet Facilitator can't write a PBI yet." And it was like it, it jumped in, in voice mode and wrote the PBI, pinged it up in Loop. Like it couldn't yet pull it into the, the backlog, but I was like, "Oh, this is-- You can see where this is going now."

Like a, a permanent expert PM in every session, involved in every call, constantly monitoring the backlog. Like the potential productivity is really exciting. Yeah, exactly. And just keep things a little light, uh, one of my favorite movies is Matrix. It's like in one of the scenes where somebody wants to fly a helicopter, and they're like, "I don't know yet."

And then now they know a couple of seconds later. Yeah. What you know about Facilitator, you know, is probably dated already. It probably knows it has downloaded and can-- is capable of doing more skills than when we last tried it. Uh, so that's the beauty of this. Like, once we create this new invocation pattern with agents, and it is in the flow of your work, you're not going to another app, you're not going to a different surface area.

You're doing the work like you have been doing before. Hmm. Um, and these agentic tools, they're really capable, the more mature, more, uh, uh, you know, easy to invoke, simplified, uh, seamless, easy to invoke tools. They just come in and amplify your production. They speed you through, uh, what used to be previously, you know, sequential tasks.

They let you speed you from one segment of the task to the other segment of the task to the other segment. Because if you look at Amdahl's law, you know, uh, or Amdahl's principle, it basically says the system as a whole is not going to see the improvement unless every segment sees the improvement. And it, it basically states a corollary.

You-- The only improvement you see in the end-to-end system is the, uh, speed up you have done to the biggest, uh, long pole in the system. Hmm. And so it's one of those where if, if, if I can get to my next checkpoint faster and you can get to your next checkpoint faster and someone else can get to their next checkpoint faster, all in the tools that you use, that's how your system of work, which is, you know, we like to think of, uh, Teams as a system of work.

Like, y- your system of work can help you achieve more. So th- that's, that's really interesting because I, I think, I feel like development is six to 12 months ahead in this AI thing constantly. Like, like, like, like the developers and, and, uh, kind of IT people are more willing to experiment. There's some more measurable outcomes.

So I feel like the productivity we're seeing in development now, there's a six, eight, 12, depending on the organization, depending on the vertical lag for organizations to get it. But th- one of the key things you highlighted there is you need a, a, a, a, a usable, scalable enterprise system around it, and that's where Teams is really interesting, right?

Because with developers, you can be like, "Well, we'll use GitHub for the repo, and we'll have it pull into this system, and we can create a file share." Like, like you can pull together your own thing. Yeah. In an enterprise, like, Teams has the potential to be that surface for a lot of people, particularly in the sense of true teams in terms of channels and files and apps, not just chat.

Uh, you said it very well. Uh, we see this responsibility very clearly. There are a few announcements we have made. Uh, in the past we talked about Facilitator. We also had an announcement around Channel Agent where, um, uh, a durable... Think of channels. Let me, let me spend, spend a few minutes talking about channels and chats.

Like, Teams is fantastic for chats. Lots of people, you know, millions of users every day use group chats and one-on-one chats, DMs, so to speak. Uh, and it's a fun way to, you know, correspond with a small group of people on topical items. Uh, let's catch lunch. Hey, what's going on for, you know, uh, what's going on on Friday?

What do we do? Et cetera. You can also use group chats for some project work. Hopefully, those project work are pretty, uh, short, uh, timeframe. They're, uh, very pointy, very unique, and then they have a short life, shelf life. We view channels as a longer duration construct wherein, uh, team members, squads, uh, come in and go.

They operate on common artifacts, um, which are all in a SharePoint that is mapped to that channel. Um, so you have the governance capabilities of SharePoint. You know exactly where files are. Um, you know, the all users, current and future members of the channel, can easily onboard to the channel and figure out where things are and how things have been.

And so bringing an agent into a channel was a logical first step for us Because we knew people who are using channel as their ways of getting work done will have a common location for all the relevant pieces of their information for that project. And so we did this channel agent, it's in preview, um, uh, and it's, it's pretty exciting.

We've also brought n- uh, Copilot directly into chat. So in Teams, if you've got the Copilot license as well, you will be able to add Copilot into a group chat and, uh, you know, kinda converse with Copilot and say, "Can you do this for me?" As a, kind of a digital worker in the chat itself. I, I, I love that as an adoption feature.

Like, we've been doing a lot of in- internally, like, like bring Copilot into the chat, and the questions that people ask, you're like, "Well, just ask Copilot in line," just to kind of re-cement, like the obvious thing to do here is to ask the AI. That, that's one element to it. But the element to it is none of the other platforms I feel like have nailed multi-play- player AI yet.

And it's still a work in progress on the Microsoft story, I think, as well. But like there's a, there's a huge unlock if we can all be working- Yeah ... together with AI. Absolutely right. Uh, and this is just the first incarnation of all these, you know, Copilots and agents in Teams. We are constantly iterating, finding a better model than what we have, uh, to make sure that we remove all the seams.

And as we learn from profiles of usage, uh, usage profiles, you know, this is where things are working fine, here's where two extra clicks are happening, which is not in the flow of work. We'll get to optimize it and, uh, and make it faster, easier. We just don't want you to think of Copilot as its own thing.

We want to think of Copilot as another, uh, virtual coworker. You know, maybe a limited capability, but definitely a limited coworker who can do, uh, small things, but reasonably well and reasonably reliably and robustly. Uh, and I think that's where Teams as an AI agent surface comes to life. Like there's a huge habit change to go to the Copilot app or, you know, other apps.

Like, like I go over here and do some AI things, and I think that's what we saw with OpenClaw and Hermes and, and Scout in the enterprise is like, a, a lot of people think of it as a scoped virtual coworker. That's a paradigm they understand No, 100% right. Um, and then we should get somebody, uh, some of my colleagues on the show who are subject matter experts in this agentic platform because we have fantastic amounts of work that is going on right now.

Hmm. Uh, humans and agents are building the next evolution of Teams with agentic flow. So, uh, it's pretty exciting time for us, and it's gonna be very exciting time when we get to announce it. Uh, but the right people to talk about it are some of my colleagues. We'll get them on the show and we'll have a- Yeah, yeah, yeah.

We've got, we've got a thread going ... much deeper I'm keen to get them on the show 'cause that's, uh... Yeah, definitely it would be good to talk about that. And, and that's where I think the channels thing is interesting as well. Like, you, you mention all the benefits of channels, and they're very clear, but adoption is definitely challenging 'cause for the individual, like the new Teams UI has changed this quite a lot and obviously with this in mind.

But, like, chat was, I understand what WhatsApp is or what iMessage is. I understand what chat is. I don't quite understand what the team is and what the channels are. I feel like this is a compelling event to be like, the reason we're all using channels is because the AI channel agent or facilitator, whatever it ends up being, has great context of the files, the previous meetings, the history.

Like, like there's a compelling reason for the entire group and for the individual to be like, "We want to keep all the knowledge in this team's container for the benefit of the project." That's absolutely right. Uh, uh, there's, there's, there's time and place for using chats, and we want people to enjoy chats.

There's lots of improvements coming into the chat infrastructure as well. But channels is where the companies, organizations, uh, think about large companies, they want a standard way of working. Hmm. There is a way of working for every organization. Uh, it could be ad hoc, but it's still there's a way of working.

Uh, for large companies, they will, uh, get a much higher ROI if they structure how their employee workforce work. Not like telling them exactly what to do minute to minute, but telling them, "Hey, here is how you discover other information." So the discoverability of channels is way higher, uh, than group chats, which are by design limited and, you know, access controlled.

So the chats and challenge, and chat, sorry, chats, messages that I send in group chats, uh, they're not searchable by anyone not in that group Uh, but if I do something in channels because it is a general purpose information that other people in the company might also find beneficial, then my public channel is indexed in search engine and, you know, in, in, in M365 Search.

Yeah. And you get to find out even if you're not part of the, the channel that, that my team is working on in the channel, and you can join us, or you can borrow from knowledge that we have gained, um, and then reuse it for your areas. There are a lot of trickle down effect. And, and people cha- people change roles over time as well.

That too. So they come into a project that has two years history, that chat isn't any good for that, but they can join and actually then they have permissions to the channel data, so then they can ask AI about all the history. Yeah. Yeah, exactly. And so what I'm going to... I don't mean to say, or I don't mean to imply that organizations just should define a rigid code of conduct and, you know, rigid, this is exactly how thou, thou shall work.

Uh, not at all. But there's a certain conventions that can become easier, that can make it easier for humans to communicate across their projects, even when they don't know about the projects right now- Mm ... that will enhance discoverability. Um, that's because channels are public or private. You could have a public channel, and you can expose your knowledge that you have gained, your expertise to other people in your company.

Or you can keep your channel private and say, "Nope, this is a pretty significant secret project, so I'm only gonna do this once. And then once my project launches, then I can, you know, make it public inside the company." It's not public to public, but it's public to your company workforce. Um- Yeah ... and that enhances the knowledge base for the entire organization.

And so your agents and AI infrastructure works better, uh, if the agent that you are invoking has access to all the right information. And this is, um, Sati has been talking about it a lot recently on, uh, LinkedIn and his post about, like, that enterprise knowledge is where the value is, and you want to maximize it, control it, protect it, like, organize it.

Yeah. So, like, like, I, I feel like with adoption, there has to be a compelling reason beyond the individual person's motivation. That's where the channel adoption has fallen down in the past. Like, we all know it's the right thing to do. Mm-hmm. But someone important on the project falls back to chat or falls back to email, and the project just...

The conversation naturally moves. But if- You lose a piece of information, yeah. Yeah. Yeah, exactly. But if, if everybody's like, "Look, stop." Like, we need the knowledge in the, the, the agent knowledge base, or we need facilitator to be up to speed on this decision, it will keep, I hope, keep pulling us back to let's have the conversation in the right place.

So it totally makes sense. In, in fact, one of the under- uh, understood, uh, benefits of channel, uh, this public knowledge base or the searchable knowledge base is recall. Uh, when everybody's working actively on a project, most of the information that you need that is relevant for the project, for its progress, is already in people's heads.

You know, it's, it's top of mind, uh, because that's your day-to-day thing. But imagine a scenario where an important feature shipped, and now people, the same set of people are working on the V next version or some other feature. Um, and then you need to have this corpus of knowledge of why did we decide a certain thing in a certain way that we shipped?

So clearly it was important for us- Yeah ... to ship it like that. Uh, but what made us... What were the decisions that led, what were the inputs that led to the decision, and what would happen if we changed that decision, right? Uh, and that recall of decisions and inputs to the decision, uh, are very easily queryable using Copilot if your information is out in an accessible form, uh, like in a channel.

Uh, in a group chat, you can certainly use Copilot to query that, but you only get information from chats you have access to. And which is a different thing than public channels inside your company where people could have done very similar addition. You know, product launches in certain regions with a certain con- constraints, uh, that people have figured out for one product line.

Another product line, a completely different set of people could borrow from that intelligence and then leverage that for their areas. They just need to search for it. So would it be, uh... We need a little bit of a foresight here for people who are planning information management in the companies. Do they get to d- you know, do they get to not dictate, but they get to bring their people along for, "Hey, normally this is something you could do in a group chat.

We would totally allow you, you know, you should totally do that. But if you do it in a channel, then your contribution to the company's bottom line is ever, uh, growing- Hmm ... o- over time." So it's not a point, one in time or, uh, it's not a point in time value that we can give the company. It's an accruing value we can give to the company knowledge base.

I think that's a- And that's a hard work ... I, I think that's a, that's a compelling story. But the other part is, and selfishly for you, you can ask the channel agent- Yes ... team agent facilitator any question about the project any time of day or night for the whole entire duration of the project. Like, like, the, the selfish motivation for the individual and for the project is like, this is gonna be a much better project for you because the AI can correctly reason over the full data set.

No, exactly right. Uh, imagine setting up a scheduled task in Cowork, in Microsoft 365 Cowork, right? Um, saying, "Hey, I'm working on this project." Um, you know, every morning, you know, 6:00 AM scout... Not scout, but, you know, search, uh, my Microsoft 365, you know, access, um, uh, search searchable knowl- knowledge base to find out if somebody else is working on a similar initiatives.

Why? Because I want to borrow from the knowledge and I want to learn from that. Hmm. Imagine this happens. That search, uh, that scheduled search from Cowork will give you no results if the entire company is doing group chats. Yeah. And, and that, and that's a, that's a major motivation in large organizations is, like, everybody, particularly in this day and age, everybody's running at 100 miles an hour.

Like, Microsoft is a great example. Like, there will absolutely be people working on similar and possibly parallel initiatives, right, in a global organization like Microsoft or any other global organization. So being able to join those dots at the intelligence layer is really interesting. Yeah. And, uh, and so, uh, it's not just channels, it's also the, the backing SharePoint, it's also the backing platform apps that you can add.

Uh, and I feel, I feel like Teams has a compelling story, and I want to echo back what you said earlier that, um, it's really, uh, an exciting time for people on Teams. Kinda you should expect, uh, you know, uh, folks in our side, uh, to bring you more and more compelling use cases and easy-to-use scenarios for agents.

Um, and then we'd love to hear any feedback, like in your podcast or in response to your YouTube, uh, and, and, uh, podcast post, or send feedback directly to Microsoft Teams, um, through the in-product experience. Send us feedback. We'd love to understand where you feel the existing facilitator and channel agent is working for you.

We'd love to also understand where it's not working for you, where you want us to evolve that. Awesome, Sharad. Well, thanks so much for hopping on the pod. Um, we're, we're already in the chat pulling on a few of your, your, uh, team. But yeah- Right ... as you said on the podcast, when the, when the time is right, we'd love to get those different SME areas, uh, onto the pod to go deep on the, the agent story and as the new stuff rolls out, 'cause I think it's just a really- Yeah, we, we had-

really exciting time. Thank you so much, uh, Tom. Always a pleasure here. And I feel like one of the agenda items you wanted to talk about, uh, I wanted to bring here, uh, is some new enhancements we did for Catch Up, uh, on mobile. So if you're a Teams mobile user, whether it's Android or, um, or iPhone, um, you should use Catch Up.

Catch Up is a phenomenal, easy to use, delightful way to, you know, catch up on, uh, what's important for you. It's not everything in your activity feed, it's the @ mentions, uh, it's the mentions you have, it's the unread messages you have, and it's a pretty one-handed way of, you know, figuring out what's going on.

And in the card-based view of Catch Up, um, you would be able to quickly triage. And my hope is that, you know, this is what I use it for, my hope is that most people who, who kind of check their phone, uh, getting to work, um, they're able to do this, mm, you know, five times faster than they typically do right now, where they have to navigate back and forward.

Uh, I, I, I love that feature. I think I said at Summit, like, that is one of my favorite features because I travel a lot, so I mobile first a lot, and that's one of the things that pulls me back into channels very successfully. Like, I've got so many channels, I wouldn't naturally necessarily open the Teams and check every single channel or go through every unread.

But the, the card surfacing, this is what we, we the system think is personal to you. And my excitement about that is, over time, the AI can get cleverer about, "Well, I know what Tom engages with or what he cares about." So I, I'm assuming over time it will get better at, like, "Here's the things that really should matter to you."

Uh, we don't discuss future features. Uh- Here- here's my hope ... continue planning it for them. Here's my hope. Yeah, but, you know, it seems a lot of logical things for us to do. Uh- Yeah ... but first we would like people to use it, see how valuable it is already, and then there'll be increasing, you know, uh, fine-tuning of the algorithm to be used to show you the right things, not just based on chronological order, but based on the things that are important for you.

Hmm. Uh, we have other features which are rolling out, which are, uh, you know, messages you have saved, uh, your VIP users and things like that. So all of these are going to, uh, make things, make Teams the system of work where you can quickly catch up on what's relevant and important for you, and then dig into things when, when you need to get some work done.

Awesome. Well, Sharad, thanks so much for jumping on the show. Uh, we've definitely, by that conversation, we've definitely got some more things to, uh, to follow up on. So we'll, uh, have, uh, you and all the team back, and hopefully we're gonna get, uh, still to lock in, but a fireside chat later in the year as well, where we can go live with everybody and, uh, talk about some of these features.

Absolutely love to. Uh, always a pleasure being on the show. Thanks a lot for having me. Awesome. Thank you. Cheers.