Task Intelligence: Doing More with Less
All, right and so let's go ahead and get started and, I've. Been told there's two and. Mic. Runners. In the back so if you have anything to. Say please wait, until they get. To, you think you just pop up your. Hand. And they'll come over and. So. This, is the task. Intelligence. Breakout. And I. Didn't. Add the, doing. More. With less part, so I'm not really sure what that means but. But. Basically it's about and. Tasks. Is a way to. Sort. Of. Represent. What, we do and tasks are obviously extremely. Important. And. Tasks. Are defined. As pieces. Of hope and work. In the. Range. And scope writes from like extremely. Small. Scale things like, an email for example all the way up to very, broad, and large scale things like trying to plan an event. And. Tasks. Are extremely important obviously they span kind of everything that we do and. Task. Intelligence. Is is, largely, about. Technology. And, experiences. To sort of excite. To understand and support the completion, of tasks right so and when, we're talking. About task intelligence. And this afternoon, that's kind of what we are. Talking. About. You. Know in task intelligence. Forms part of part of lots. Of different applications ranging. From. Digital. Assistants, productivity. Applications and. So, on, and. I don't expect and. Read. All the text here that's not the point, and it's. More just to show the. Brights. Of different things that we, and, think. About when we're talking about task, intelligence. Right so it ranges from task. Extraction, task understanding. With support, tasks, and. Progress. Over. Time and. Also how do you. Evaluate. Support. For tasks as well and. We've done a few pieces. Of. Work. In this. Area. I just wanted. To draw, your attention. To them so and, we've, looked at things like. Trying. To estimate how. Long a task will take you and, so, based on. The. Text of the task is input, can. We predict, the amount, of time that you may need to, block. To complete that task right, a sort of. One. Application there. I've, explored, and. Contextual. Information. Around, tasks. Tasks, taxonomy. There's been a bunch of working tasks and, in. A lot of different. Places. As well obviously and. And. Lately. We've been looking at and. Can, we predict when a task is. Completed. Right so and can, we infer from what, you do after, you start, the task or create a, task come, in further the task has. Already. Been. Done and. Tasks. Appear in. A lot of different Microsoft. Products. As well these. Range. From. Cortana. To Outlook being. A. Project. Planner, and. To. Do. Either. There. You. Know and. These are just some examples. That. I picked based on who we have here to, speak, speak. To us so. We, mean I guess the point is the tasks are. Everywhere. And. We, do also have the. Partnership with. Academia. Around, tasks as well and. A. Couple of years ago we started this. Venture. Called, the Cortana. Intelligence, Institute. Which. Is a, targeted. Partnership, we have with our. MIT, where. We. Sponsor. Research. There, around. Tasks, right so as, I'm. Started, a couple, of years ago it's going to conclude in September October of this year and. And. There the emphasis is really on tracking. Tasks, over. Time. Right, so can. We predict. How, far into a task and. The person, is and. Then we'll. Learn. More. During. The panel and. Speaking. Of the panel, and I'd, like to introduce and. The. Panelists, and I. Don't know if you guys would like to but they come up and sit here would that be I don't, want to sit there is up to you, can. You set up here a bit good actually yeah the. Decision right and. So. We. Have. Five. Really. Fantastic. Panelists. From. A. Range of different, places. All, you. Know each of them is going to speak, to you in a bit to sort of introduce. Themselves. And, what they do and their. Backgrounds. And. So. On, if. Florez. An, associate. Professor at. RMIT. In. Australia. She's, one of the key partners in our and, Cortana. Intelligence, Institute. Collaboration. And, Keith. Is a general, manager for Azure, and, developers. Services. Killens, the principal p.m. and, to. Do and. Shh. Reason. Circle p.m. and Bing and wen, Hao is a partner, GPM, on project. And planner so. There's a broad and, range. Of expertise, here and lots. Of different microsoft.
Products. Plus. Some external. Perspective. As well and. Just a little bit about how, we're, going to structure the. Rest of the time and. So. We'll, do a panelist. Round-robin. Reach. And, panelists. Will speak for a little bit of time about and. Their. Background. Tasks. As it. Pertains. To that they. Will have some discussion, between. The panelists, and Q&A. With. Y'all. So. Um. We. Should just kill it and go ahead and get started. I. Think. I should probably go in this, order. So, Laura, would. You like to start. Thanks. Yeah. I. Don't. Think I'll be able to finish ten, minutes. So. An overwhelmed, user. Who. Is one of the puppies. It. Wasn't, sorry now can you hear me all right okay right so. I mean on data with Microsoft, every day look at my phone with all these notifications. Emails. Text, chat, and appointment, reminders I wish. There's a way to aggregate and help me to prioritize them so this, was one, of the real comments, from our, participants. So how many of you can identify yourself with this. Great. So our meter now, the question, is if you and. I are going to learn, more about people's, tasks and being able, to generate only, notifications. And recommendations, that are useful to users. How. Do we actually do this how do we generate relevant, timely. And useful tasks notification. And recommendation. That would be really helpful. For. For. The users I think. A lot, of you in this room, have. Been working on the search. Problems, right, so. Basically, a lot, of the tasks, in you know information, behavior trying to. Understand. People looking for information is almost. Perhaps. Solve. Right. So when, you actually say you. Know I want people trying to search. For something you, always try to infer the intent now, the thing is this, information, behavior, these stairs are very, much complex, because. People looking for information at, different, contexts. They. Might be going through different apps what. Trying to while completing the task and they, might be doing this at a comfort, of their, sofa at home and this. Might be a personal task this, might be work tasks we don't know in. Fact I did something, I tried, to do some comments, on my students work on my, phone, last. Night in, the hotel room and it looked like this it was actually a work task sometimes very hard to say whether this is a personal or test. Well. Somebody, might be. Looking. For information while. They're in a shopping mall and. This, could be a mom who's. Got a task that I need to find. New. Shoes for my son, before. The school start so, that's at us and. The. Information behavior, again is quite different where. You have an intent, on looking, this, and you know it may not be a search. Activity. But it's actually going through, you. Know different, comparison. Websites. But. Somebody. Working on a task could, look like this too so. The context, is you know this user might. Be you, know are sitting in. Their office. Juggling. Three tasks on multiple. Devices and. In. In front of him might be. Azure, DevOps, or maybe, it could be a. Planner. While actually communicating. With his project teams on teams on the phone so, right. Now what, has, look like in very complex, and understanding. This is not an easy challenge so. I said to Ryan you're going to be featured in my slide and there you go so, Ryan, recently. Had. An article in ACM communication, that says as a community, we need to invest in evolving. Search interaction, to. Support, the often. Underserved, last mile in search interaction, and that is, task completion, so, it is very hard to to get there to, understand, when, do you know this somebody, is doing, a certain task and that chaos has completed. Now. For. Me, understanding. Task behaviors, has. A lot of facets to it, first. If. You're trying to develop, a task assistant. That, can automatically, understand, what tasks users are doing, right now. Where. Do you start where, do you get the signals from you. Might have these you, know your your, information. In being in Microsoft, you got lots of information from your productivity Suites you've got office 365, or, the emails data and all that but there's a lot more information beyond, that, you. Know people asking, you requests, but and. As, I said before you might be in different places and how. Do you know if somebody's progressing, a certain task and it's. Very important as well that attach. Systems, needs to celebrate, when. You. Know help to celebrate, a milestone as achieve so a task completion, now. A lot. Of my work even before I was working with us, understanding. Just, the last two years. My. Research has been mainly on context, and behavior modeling so, I'm. Very much interested in understanding user. Behaviors. And. How they relate, to multiple, different contexts, they might be in and this.
Is Done through user centric sensing, so what I mean by user centric sensing, is. Basically. You know you, know that IOT. Data, and you know everything. Is all sensed about our life, and smartphone, smartphone has a lot of sensors in them and basically. Can we understand, people's activities. Can. We understand what user drew and, the activities in there when they go about their daily life. Maybe, it's a Saturday do they start by, dropping. Their kids in the childcare our, nursery. Before, in school before. Getting. Trying. To get a car park in a train station before, committing, to, downtown. For. Work and then. Maybe. They go through meetings, of the meetings of the meetings and then, they might have their money to have a project fight crime meeting somewhere else it, may not be at that workplace so, try to understand these physical, movements, and behaviors, mobility, from mobility. Behaviors, to, infer, what. Looks like an, you know user, mobility. Behaviors relating to toss but. In on top of that another. Source of users and central sensing data are social. Media data so we. Know for example when people checking ins because every morning, I just have to grab my coffee then, I know that is part of you, know regular. Tasks and. And. These social media data actually, gives you semantics. This give you content, and semantics, into the. User behaviors, now. Where. Do you start with all this behavioral, data it's, very important, to always start by understanding and recognizing context, because. Context, characterize, the situation of a user and. Context. Itself is. A strong, influence for. Users for taking decisions so I just give you an example, when. You get an incoming, call. These. Days maybe, this is rare right you, might be getting notifications, about, you on your whatsapp, or your team's now. What. Would user do with it this, is very important, in. Order for us to understand user behavior, we need to know the context, is this, person, in a meeting if. I'm in a meeting, will. This user answer the phone well it depends on the social relationship, it if, it's my mom who calls and I'm. Expecting, her, call because she's looking after my kids then. I might be excusing. Myself from the meeting and I have to leave but. If it's incompletely unknown, calls I'm not. Actually ignored. And in in, some of the actual call. Log studies we've been Luke studying. We. Look, at, behavioural, very, strong. Signals between context, and behaviors, so, even for example if somebody had their boss calling sometimes. Completely ignored so. These. This, was very important to understand how. Context. In, what certain context. This. Uses. Will you know how will you respond, to all these notifications. And. That's, actually really. Largely dependent, on the context and. We. Between. Couple of years we've, been studying heterogeneous. Data sources to understand, all these different contexts and behaviors now I. Want. To emphasize again that really. Context. Can. Reflect on different behaviors, of users. When. Whether. Or not they're doing, a certain task and the. Type of tasks, that they do so just. Giving you another example. Now. We. What, about, you. Know indoor behavior, activities, in a shopping mall so this is completely, different scenario. When. People go to mall what do they do what sort of ties they might do can I ask you some questions some answers from the audience eating, good what, else. Shopping. That's obvious. Hanging. Out. Hey. Okay. What else anyone. What. Meeting. Friends yep they actually a lot of instances meetings project, meetings in shopping malls and and. Apparently. They're. Also a lot there's also task, of caring. Like, parents. Or carers with, young kids when, they just want, to. Shut. Off and let the kids just run, around at an indoor playground, and they, can, actually then work on, their, email at least you, know so there there are a lot of cases like this now. So. We did a project with, Westfield. This. Was completed in 2017. It. Was a large-scale in the movement, information, behavior tracking. We. Had these. Wi-Fi access locks for more than one we, also have the content, data so from you, so we also had access to the ISP logs of. More. Than hundred twenty thousand anonymous, users of the public Wi-Fi, it. Was a large mall in the city Sydney, so it has, more. Than seven story and, what. We found out is we actually come up with these terms. Of cyber physical social context. These. Cyber physical social context. Will. Largely determine how.
We Understand, user behaviors. Just. Give you an example here. So. When. Somebody is in, when. Somebody wanting to buy and. Searching, for who I honest, if, you're not sure why this - so, those. Are - and the, the, context, when the issue the search queries are very. Similar. To the context of the shopping mall so where. They are at in the access point so they're. Very close to the shoe stores so, this is of course not not. Not. Surprising, but. The thing is when we start mapping. When. We start moving movement. Loads across. Floor. So the, the darker colors showing. The higher floors you. Can see for example some, strange very quick, movement, I'm. Sorry I think the arrow didn't point card correctly, but there are movements for example from level 1 to level 5 directly, why, would people move directly from level 5 from level 1 right, so. If. You haven't been to Sydney. Level. 5 is. The. The food court so. You've got the shortest path possible from. From. The ground, floor to level, 5 and these. The, fiber content during, this is also very, very. Specific. To. Work-related. Tasks. So they might be doing meetings, there, might be on their productivity, apps, while. They're sitting in the food court so. We. Can actually recognize task. When. We combine cyber, physical and social, context social context, means. They are at and, therefore. What. We did further was. You. Know this public Wi-Fi data are not annotated, so it's, very hard to be able to validate anything, so that's what we then release a, survey, to. Public. Wi-Fi users we. Could actually then analyze if. They're cyber physical social behaviors, I. You. Know using, those traits, from cyber physical social behaviors, in their Wi-Fi locks can. Reflect on the demographic, data and apparently, they are very strongly, correlated, so. Um what. We use with this data and we can then come, up with some kind of like try product, recommendation. Of where people will be interested in or. Locations. They might be interesting or maybe some, items. Or products they might be interested in in, in, the mall, now. Another. Project I'd like to mention before I mentioned the Cortana intelligence Institute, is, we've. Also done some work on a physical, and social sensing at work for workplace analytics, so. We. Know that, research. On hot desking hasn't, been quite positive. So. But we there's not much evidence about that so what, we did with with, Eric was we, instrumented, some, of the workers carrying. Around our, mobile sensing app and we also put some indoor sensors and we, collect also their project, data so the background, data. And project data and we, want to be able to see if. If. We can infer. How. Much a worker. To. Concentrate. Depending. On multiple factors including whether how, many projects they have right now, whether. They actually sit, at their, preferred, seat or you know they just sit in a place where because everyone else has taken his favorite seat then, you know I have to use an existing other seat or and. There's, a lot of other data we collect including CEO to act which means air quality and we've. Done some, analysis where. Apparently. Concentration. The. Ability the perceived concentration, reported by the user really. Has a strong correlation with, the. Nam of meetings, when. There are no more meetings. Of, the the less they can concentrate, and. Apparently. Although, the reported. Concentration, they actually, say well. Co2. Or, air quality and, temperature, doesn't have an impact on my concentration in, fact when we look at the data it does have any he has a very strong relationship and, when. They actually sit at the preferred zones, they actually, also. Have a, better, productivity. Pursue, productivity, so. Here. I'm not going to talk more about that now, when, you look at these two data, sets that we've been talking about the mall, and this. Workplace is so, very specific, to that those. Two buildings now, the challenge, is how how do we recognize task, behaviors, in daily. Life from. People. That. Can work at multiple, places not, just Microsoft, not just. Universities. But, maybe people who are working in marketing agencies, people who are working in, all. Kind of we're all kind of professional. Roles. Really, so, there's a really challenge challenging, thing, another. Challenge, is there's. No data, set that will give you there's, a lot of data on. You. Know you've been collecting a lot of data on.
Online. Activities, for example and email, activities, project, activities. But. You know and also there's another group, on ubiquitous, computing I come from you be competing, research field collecting. A lot of mobile sensing data. But we. Don't have the task annotations, and this is what we need so, how do we construct such a data set for test intelligence, and this, is why the, RMIT. Court I mean my career so far MIT Cortana until I just Institute was launched. So. Ryan, just a correction it was launched last year February, 2018. So. It wasn't that long ago so. I is this is just a, just. A bit more than a year ago and. The the main aim of CI I was. To, build. A bench my data set for task behaviors. Now. How do we do that first. We need to collect rich, annotations. Of. What. People do and, we do that through sprint, sampling and end of day survey so, we collecting, title, notations, it states and the progress of every task and the. Activities, the sub activities, of those tasks and. While. We do that we continuously, lock the cyber physical and social signals, so, again, what is the cyber physical social so this cyber, signals means the online activities including the browse web domain app category app usage. Time the, physical signals, including movement, and social. Including basically. Calendar. Information. Whether. Somebody's busy in a meeting or and all that now. What. Makes it harder though. The. Existing, categorization. Of, tasks are very broad, according, to American, time use survey. So, the, American, time you serve it has these, attacks, on the US personal. Tasks work related to social, exercise entertainment. Caring and civil obligations, but, work related tasks has no subcategory. Well everything, else has subcategory, so, this is why then, we, before. We cited the logging we, send out a survey to. More, than 400 participants in, english-speaking. Countries so that includes us, and. The outcome is a new work work tasks are so taxonomy so, this was publishing. In. Cheer, so. The work the sub category and, we call this as micro tasks. Include. Communication documentation. Planning, and so. On you can you can see the list there now. Now. That we've we've, come up with this way of annotating, tasks. We. We. Then collect the data from. A. Range, of professionals, and non-professionals so, non professional mean students, and this, is what make that makes a day unique as well because. A. Lot of data, has been collected mainly, on students, so, and. What make it challenging, is we need to, collect. Data, we. Challenging ourselves that, we want to collect data over four weeks so we can then start observing. Long-term behaviors, on longer. Longer term behaviors, on tasks and. Of. Course to. Get someone. On a manager level a business, owner to. Be able to be involved, which. And then we actually have a weekly meeting every, week to, actually make sure the. The. Data that we collecting the annotation of collecting actually makes sense so there's, a lot of work on on that part as well and weekly meetings so, we actually pay the participants. You. Can I'll tell you more later about it if you interested how much we have to pay eight participants to. Keep them engage over four weeks. So. What do we get in, fact before. I go there we. Did this over seven batches and we've, just completed the data question last month so because, it these every, time we do this because of the weekly meetings we can only. Onboard, ten maximum, because, of the number of people that are available to do the interview every, week so. We can we've, we did this over 12 months data collection and. The. Behavior, locks we've, got more, than 200. Now, sorry nearly, 200 row features I need to get it correct so nearly, 200, refused, including, calendar. Events, movement. Data so we even get the transport. Hit map so we're. Where. They are likely. To, depart. Even potentially, we can learn their transport, modes. Important. Locations, in this person's life. Cyber. Data including, background. And foreground app and mobile, also on desktop, if they install, the the. Application on, their, desktop or laptop is their work allows for it then, we'll be able to no less than somebody's on Microsoft, work or if somebody's on Project, will be able to get that and. If they also install. The. The. Browser plug-in, then, we'll also be able to collect the website. Now. Along. With that there's, a lot of sensor. Data channels including facility, Wi-Fi, cell tower and all that so, and the report, the test data, there. Are some partial snapshot, of that for example there so I'm gonna report that to us let's. Say I'm. Affected investigating, on email this project. A or, I'm looking compliance, governance so you've got a lot of we haven't done any NLP, at all on this data and all some it's, pretty, pretty.
Noisy So so and along with that people actually report, whether they think each, of these tasks, require. Cyber or physical or social context, and we. We like we, did some. Initial. Analysis, on the data as well. We found that median saturation, is just about a half an hour except personal tasks. They're. They're, much shorter. Work-related. Task participants only, the work once you. Can see it across the users. Work-related. Tasks has different. Duration, so, that means every, and. These are people, from completely different work, domains so, that means work, tasks are very personalized. You. Cannot really learn. One model and train it across, other users. Especially. If they're from different disciplines, or they've from, different, industry. And. When. We look at the long tail a, lot of these long tail consists of the micro tasks related caring. Personal. Social exercise and entertainment, and. When. We look at the reported. Cyber-physical social, context that are required during tasks we, found that. A. Lot. Of these especially, work-related. Tasks. Actually require cyber physical social. Context so you can see the dark blue ones are, basically. The. Number of tasks where. They think requiring, cyber physical social, context, and and. If, you look at this for example a. Lot, of the work done here for example Microsoft our cyber, only, if. You look at the aggregate of this that would be just representing, maybe a quarter of the, tasks so. That's. Why it's very important, to, to, be able to collect and understand. Behavior, from, different. Signals. Because. I don't have much more time I'll skip this some, of the initial, analysis that we did. Basically it also shows that. When. We combine these signals cyber-physical, social, we can identify tasks better. Still. A long long, way to go because we still have misclassified. Tasks I don't, have time to talk about that we also look at feature, importance, analysis, across. Different tasks. Some. Initial insights are for example we could look at. Communication. Related. Tasks really. Can. Be. Some. Of the important features including social signals. And. This is an you. Know common sense of course some. Of these like when, you move you know you can actually recognize travel, ties and all those things. There. Are more detailed findings to come and happy, to share more as well in the workshop this Friday and there's. Still a lot more to be done so we, need to recognize tasks, in 10 recognized tasks progress from this.
Long-term Tasks and. How. Can you even analyze team based toss. Sorry. Thank you. Hello. My name is Keith Ballinger I'm the general, manager for, Microsoft's, developer, services organization. I won't, use slides all this talk very briefly so, we can get through the other participants. As well, generally. Speaking in, developer, services we, focus. Less on the. Individual. Tasks, and work that a developer. Is doing, while. Coding. And creating. A feature those kinds of things it's important, but. It's generally important because we're trying to find ways to build, tools and technology, to automate, what, we're trying to do and. That kind of goes to the heart of what we're trying to do with developer services which is essentially, for. Both Microsoft's, own engineers. And. For all of the what, we call third-party, engineers in the world we're. Trying to build those services, that make them more successful, and more productive. What, that means though in practice is it's far, less about helping. Someone's type. Faster, so to speak in. General, developers. Especially, professionals. Who've been added a while find. Ways to kind of create context. Create, flow. You know they lock themselves in the room they do the things necessary to, kind of hammer, out a feature so, one of the areas that we, focus a lot on is, what, does it mean for a team to, complete work and in particular, if you think of a task for instance as a. Particular. Bug that needs to be fixed or a feature that, you want to ship in your service. Or in a mobile application or even the launch of a new product a team. Needs to work together to do that and there's, a lot of different techniques that you can look at in order to make a team, more. Effective, and. In general as, you notice one of the products I owned, is called Azure DevOps, the, kind, of the DevOps world is a set. Of tools technology, is really, like a kind, of a philosophy, of process, around. How you can shape a team to. Do certain things and how you can get them to work more, closely together a lot, of that comes from you. Know various, management, science from. The 20th century so for instance Elihu gold Bratz Theory, of Constraints, is around, taking any particular, system and he was looking. At manufacturing systems in the 70s, and looking. At you don't optimize, you, optimizing. Any particular, step in a manufacturing process, is generally. Useless and it's you're just creating waste and you're just piling up inventory, which was a is, a real deal killer in. The manufacturing, world unless. You're optimizing, the. Part of the process the task that is actually a bottleneck, and the, Theory of Constraints is. Completely, focused on this idea of reducing. Waste and reducing inventory in order to by, optimizing, finding. Your biggest bottleneck and fixing. That and for, whatever reason, when explained, it sounds quite normal, but if you look at how software engineers in particular work. In teams this, can still be a problem where, a developer, may want to optimize a part of their process or, a team they want to optimize a part of their process but, and they're missing the forest for the trees in that again, you can code as fast as you want you know I could, put all developers, through typing.
School. And make sure they all can type 100 words a minute it, doesn't really matter whether you type 100 words a minute or you type one, word a minute as a developer, that's not where the bottleneck, is and so, that's kind of Li to the other area, of interest. With me. And my team is well. What does it mean to. Reduce, waste right, using, that manufacturing, trim what does it mean to have you. Know piled up inventory and in, the world of, services. Today and apps and things like that what it generally comes down to is the fact that how quickly you ship something doesn't matter how. Quickly you fix a bug for a user or ship a new feature doesn't matter if, no one's using it and if you actually haven't solved, the customers problem so, a lot of our focus has been around how do we spend more time understanding. What the actual. Customer problem is and what the value that the customer, needs to have is and then. Once. We know that we know we can ship the right thing and. This becomes very interesting when you think about the. Idea of tasks, in assistance, that help you with tasks in. My world of development there. Are certainly some. Great examples of where we can help developers, in doing. What they're doing on a day to day basis, in particular I think there's. A lot to do around say for instance code review and helping people make and more intelligent, decisions around the, code they just wrote and is it really good and. There's certainly a lot of automation, we can do if you look back 20 years ago when I was a developer 20 years ago there, were a production, systems I was working on where I would literally take a floppy disk and I would take, it from my computer to another server, somewhere, and then I would load new code in that world's, gone today, people, everything is automated and for, the most part developers, are already really good at this they're good at automating things that are drudgery, that they that are very repetitive in, fact, you could really, look at the entire in just software, industry, as just, a progression, of automating. More and more things that are otherwise. Manual. Tasks that people would have to do which. Leaves us to this place where it really, becomes well if you want to be innovative and you want to be a a company. Or a team that. Is outperforming. Your competitors, you. Have to kind of focus much. More on, rethinking. What the task is and the task goes, away from being here's, a feature that I need to code and then deploy to a server to, being I have. A customer, problem that, aligns, with what my business is trying to solve and therefore. Trying to break through with what that means and in those cases often, times the, real bottleneck, and the world the place where the most help is needed is in, reducing the risk and uncertainty of, is, the code you're about to write actually, useful to anybody. Wow. And so with that I will pass it off to the next person Thanks. Hi. I'm. Caitlin Hart I work on the Microsoft, to do team more, generally we have the ecosystem. Of tasks, lists, and reminders, at. Microsoft. It might be outlook. Tasks, where. Cortana reminders, are stored, it's. It's. An experience that we're we're pretty proud of but we're continuing to improve. We've. Been looking a lot lately at and and by the way excellent, selection of panelists cuz these are all like I'm just, so thrilled at the topics that we're talking about today these, are all the things that keep, me up at night we're, trying to help people to. Do. More and to be able to achieve, more. And I. Guess. I would say more the latter not. Do, more tasks, not, just.
Get Through your to-do list but. Actually to be able to end, the, day feeling accomplished. And. That requires a lot it requires. First. Of all you. To have a sense for everything, that you have to do a collection. Of everything, we aim to. To. Serve that by having a collection, in, in. The ecosystem, that we support all of your tasks that you capture an explicit, to-do list show, of hands how many keep a to-do list in a digital app. Yeah. Pretty common piece of paper. Yeah. You're. Not alone it's. It's. The. Tests that we capture explicitly. People. Emailing themselves, tasks. Emailing. Themselves, reminders, a lot of tasks come in through communication, and, often. Tasks are just following, up on communication, channels. Email. Is not a great to-do list it's very reactive. You just wind, up playing whack-a-mole with incoming, things all day and. It's, even worse now that we have more. Chat channels, and communication. Becoming fragmented. Means, I have to follow up on communication. Tasks, from everywhere but that's, not the end of it I have to save other things that I need to follow up on I have documents, that I need to finish or big projects, or goals and I need to capture those somehow so I have to get a sense for everything, that I need to collect. We. Also have tasks. That are assigned right, assigned, through projects, through. Any sort of team task, tracking experience. Either. Digitally, or you know the post-it, notes that you have for, tracking. What assignments you have out of your team meeting this week right. So. Having a sense for everything that you have to do is. Only the step one we also are, really concerned about helping people get an actionable, plan to, feel. Accomplished, at the end of the day not, just to do all those tasks, but to be able to have the healthy habits of being, intentional, about how you spend your time to, be able to feel, like you did something. I'll. Paint a picture of two different days. This. Is me a lot of days maybe it's familiar you. Start, the day phone, in hand trying to figure out what's up for the day coffee, and the other trying. To figure out like Oh somebody's, emailing me I'll follow up on that when I get to my desk and then, maybe, come, in try to start a, something. So I start a document start, a some something, that requires individual, effort. Go. Off to another meeting, maybe. Have some follow-up, so you have to take care of and then, come back to a project and. Then. Come back to emails and then come back to oh and you get chat messages, and then it sort of gets kind of.
More. Of, reactive. More of a. Fragmented. Set, of things that you just have to react to you throughout the day I, don't. End those days feeling very great about what I do I'm feeling very accomplished, it feels like I'm just battling. Things. Another. Type of day one, that I would hope more. People could have and I feel, really, accomplished, at the end of these days. Know. Exactly. My goal for the day right like start, the day knowing what the the, meaningful, thing is that you can accomplish the. Like one. Two three three two five whatever, tasks. That I there. Are actually. Actionable today, and they are going to get done and spend. Some focused time on that and then. Maybe. Go off and, do. The collaborative, exercises, to get tasks from other situations, and, know. That, you won't forget them capture. Them in a way that you won't forget them and that you're gonna follow up on them and. Then. Actually get, to you, know a different type of task, time, a, time, where you can just tick, off a whole lot of things right like the, micro, micro. Productivity, session that we were in or, some of us were in earlier today you. Know getting lots of little things done get that little energy boost right, get a lot of little. Things that are very actionable and require a different type of focus. This. Kind, of day doesn't just, come by accident it comes from intentional planning and it's hard it's really hard, I. Was. Just having dinner earlier this week with a friend whose kid is going off to college and he. Was talking about how he was struggling. With just developing, the like executive, functioning and skills to take a. Big. Project right like a thing I have to do a huge, deadline or something and know that there are intermediary. Steps, the, the, deadline to you. Know really get started otherwise I'm in trouble the deadline to produce the first draft the the, things, that I have to do to be able to accomplish. What. I need and not be rushed. These. Are the skills that are required for planning, my day for estimating. The time that something will take and these are all things that can be assisted especially, with computers. Being able to estimate task, duration, but. It doesn't come in isolation, it's hard to estimate, tasks. In isolation. Tasks, or more than tasks, they are those, things. They come with goals that come with the the. Cut the full context. Of a project, that they relate to and. So. I, want. To be able to support, people in, those. Healthy, daily, habits, those those. Things where at the beginning of the day. They can come. Up with a plan they can see, what the most meaningful, things are for them to focus on today and they, actually have the tools support, them in doing that. Hello. That's. Loud. Hi. My, name is Shree I work in being. Specifically. I work in whole page. Organization. And answers, and entities as we call it in our internal, lingo. What. It means essentially is I, think from a search standpoint we are always trying, to solve. Micro, tasks as the word has been used or, sometimes that is d thing you're doing. Search. In many ways I think we're evolved to the speed of, things. And so both sides both the user and us as a search engine live, like no tomorrow which. Means the, user types in a small query. Text to us like whether and they, say give, it to me and I'm done and, from our side we're like oh you said weather okay. I just need to give you something in few milliseconds, and that's, it. We. Generally don't either in the habit of worrying about oh what what, we asked me whether about are, you thinking about what to wear today I you. Have, a picnic plan like. What is the larger theme that you're asking so in historically, speaking search engines don't get that much context, because, nobody has time to tell all the context and writing. His effort, so. In. General search, has been started, that way but more and more we see it going towards, getting. Other signals, so, for example if, you say like Homes for Sale of. Course. It's a very big test you're not gonna buy the house online, you. Are been thinking about data you just heard a conversation the, market is hard you, know buy something soon so. What do we do with this so the sessions are drawn out long for. Us like. Buying a house, so. We, we. Try to figure out if what is the first way we can satisfy that current answer so we always think in like what is this how can I give the delight right now right. So if I can bring you a better experience I. Can, hopefully say okay you, got me a collection of homes, come back tomorrow and then I can, probably find you the next thing here. We. Try to hook the people in a sense so once, you say I'm looking. For homes you, probably are curious about following, if there's new homes coming in the market god you're tough doesn't end with just homes so. There are these other ways.
The Engine search engines are now trying to understand. The user better so we can satisfy them tomorrow better now. Something like spider-man you know if. You just search spider-man, normally. And he just given the fictional character everybody. Knows about it but. When you search today, you're. Probably looking for the movie that. Just came out you, know so we need to understand you just had spider-man I need. To understand okay, this is 2019. The movie has been released there. Is a movie theater near you it's. Four o'clock so better show me shows, for six o'clock or after so, all of these implicit things, you just said spider-man I need to figure all of this out and give. You the best answer at the same time so. From a search standpoint the tasks are evolving to, understand, as much signals as possible, and bring together a better experience that. Satisfies pretty much not. Just oh you got the movie I don't need a character I already know the character here. Is the showtimes near you here at the top theaters near you not. Some random theater, so we, are bringing together a better experience as we've all tasks. And. To push further we are also now understanding. The fundamental, biases, in queries and, you know it biases, a very big thing today in the world of news and information so. When somebody says is coffee, good for me right. So, I'm really trying to say give me all the good stuff about coffee but. The reason you type that query in was you. Heard something that there might be something bad about it so, we are trying to now understand, the semantics, behind it there why did you say good for me so we try to bring both sides now. So if you try like in, your phones or laptops its. Coffee good for me in Bing you. Will see both the good and the bad sides of it and that's, because we are understanding. Behind why, did you issue the query good and good has two meanings right. And. Lastly in terms of like intelligence. We, have if you do an enterprise search and I say like, Simon's. Office right. We. Historically, would just, be a web results coming for I don't know what Simon's office means today but. Now because you're an enterprise you understand, that oh there's, a Simon you've been having a conversation with you're. Probably meeting we don't know that but, I can quickly pull up the office a map a layout exactly point where it is so like all, of these, signals, implicit, are, coming together to push, the experience, forward and if, you think about it it's all like adding intelligence, and, again, living. Like no tomorrow giving. You in a few milliseconds, the answers. But. Kind of pushing the boundaries of the web or, rather the text box what you gave us in the text box and. Advancing. It. That's. Like my brief kind. Of intro but. Happy to talk more so I'll, get off to Howard thank. You.