00:00good afternoon everybody my name is
00:01Steven Sinofsky and I'm here with
00:03another edition of the a 16 Z podcast
00:06also with us today is Benedict
00:08Evans and also excited to have Stewart
00:11Butterfield the a co-founder and the CEO
00:13of slack and so today we're gonna do a
00:16product focused discussion talking about
00:19what's going on with slack where it came
00:21from and really try to focus on on the
00:24product and and what it is and so first
00:27of all we'll just kind of dive right
00:28into it and it's sort of a dozen
00:30questions for for Stewart
00:32so first good afternoon Stewart good
00:35afternoon thanks for having me
00:36sure so let's uh let's first go back
00:39before you had slack up and running both
00:42the company and the product and when you
00:44were just part of another organization
00:45or a larger organization how did you
00:48personally go about sharing information
00:50there and and I'm guessing it was a lot
00:52like a typical company thing like email
00:55and it was slow and painful or because I
00:58think that's gonna be important in the
00:59genesis of how slack came to be as a
01:01product there's a ongoing joke about
01:03software developers and software
01:05designers then among the very first apps
01:07that they wanted to crack are either
01:09replacement for email or to-do list or a
01:11little bit of both and I think that's
01:12been ongoing in my own career I've seen
01:15you know many times I've stopped
01:17whatever I was working on to work on
01:19some better way to manage the things
01:20that I was working on all of which have
01:23been more or less completely ineffective
01:26and you know a lot of the software
01:28that's been designed for the
01:31collaboration Bartlett speak English I'm
01:33sure we'll get into in a bit has proven
01:36to be ineffective so it's it's kind of
01:38it seems like something that's off for
01:39it software should be able to help us
01:41with and software typically isn't but I
01:44have used ad hoc email obviously quite a
01:46bit by ad hoc I mean manually entering
01:48the two and the CC or the BCC fields
01:51mailing lists wiki's purpose-built
01:55project management tools tasks tracking
01:58bug tracking all kinds of stuff like
02:02the thing that happened for me and for
02:06this company that caused us to come up
02:09with slack was really when we founded
02:12what everyone knows was six years ago to
02:14build a web-based massively multiplayer
02:15game we start which is the natural
02:17evolution for collaborative software as
02:19you first build against yeah like
02:21massively metal play games are really
02:23hard to make and so they're good warm-up
02:24practice for making other kinds of
02:26people because you didn't know what
02:27quite what to do so you thought you'd
02:28focus a little on building some of the
02:29infrastructure to then build the game so
02:33there are therefore co-founders all of
02:35whom were on the original fekra team and
02:37we had two people in Vancouver BC one in
02:39New York and one here in San Francisco
02:40and no one was going to move and so the
02:42natural thing for us was to use a very
02:45old messaging system called IRC to
02:47communicate it was the natural thing
02:49because an IRC the basic form of
02:51communication is to send a message to a
02:53channel rather than an individual or a
02:55group of individuals and the channel is
02:57something that can exist before the
02:58people are there and exist continue to
03:00exist after the people are there no IRC
03:03is twenty five years old actually
03:05predates the web by a little bit a
03:07little bit yeah I know monkey yeah I
03:10mean it's a little bit wonky to look
03:11definitely most IRC clients and servers
03:15haven't evolved much since the early 90s
03:19and you know definitely not widely used
03:22I think the peak of popularity was
03:23probably the mid late 90s something like
03:25that only a few hundred thousand people
03:26left in the world who use it but for
03:28some critical things like you know the
03:30old HTML standards and things like that
03:31so place in that and still popular among
03:34the developer community and in
03:35particular open source projects so life
03:36ethic Linux is still large even Mozilla
03:38is a giant network of IRC channels but
03:43for us one of the critical deficiencies
03:46was that if I send a message to you when
03:49you're not online you'll never see it
03:50there's no store and forwarded messages
03:52there's no archiving so the very first
03:54thing we built was archiving a messages
03:56and once we had an archive so you wanted
03:58to be able to search them
03:59there weren't good iPhone clients
04:00available so the next thing we built was
04:02a way to view the archives through an
04:04html5 interface and once we had that we
04:07wanted to be able to post messages from
04:08this iPhone app and on and on so over
04:11the course of about three half years
04:12that's just her out there kiss Benedict
04:13here and that I think that's a
04:14fascinating I mean like even though it
04:16was six years ago you sort of were like
04:19mobile what I mean as a bunch of
04:21developers that seems fairly unique yeah
04:26even then it was a requirement because
04:30I'm not unique in this habit and it's a
04:32terrible habit and I'm trying to break
04:33it but everyone does a first thing I do
04:36when I wake up is roll over and pick up
04:38my phone and start reading I haven't put
04:40my contacts and yet so I'm squinting in
04:41my phone and I'm reading these things
04:43that are just making me either anxious
04:45or angry before I even get out of bed
04:48into the shower and for some reason
04:51that's a critical business need and so
04:53we had to design for that particular use
04:55case making me angry before you go to
04:56bed on over the next three and a half
05:02years we built more and more little
05:05features and tools on top of IRC because
05:07what we saw happening was when we hired
05:09the first employees of the number-5 at
05:12the company through to the forty-fifth
05:15employee everyone got an IRC everyone
05:17started using it and we didn't use email
05:19at all and that wasn't ideological it
05:21wasn't intentional even and it was
05:23something that we actually only noticed
05:24after the fact but the reason we didn't
05:27use it it's because everyone was paying
05:29attention to IRC so IRC was the one
05:30thing you had open alongside whatever
05:33your role specific bit of software was
05:35and that were all specific that a
05:36software might have been Excel or it
05:39might be Photoshop or it might be our
05:41because we were making a game or level
05:44building tool but in other examples it
05:46could be Salesforce or it could be an
05:48email client it could be Excel it could
05:50be whatever it is that you do all day
05:52and we found a interesting positive
05:55feedback loop or virtuous cycle whereby
05:58because people paid attention to IRC
06:00that's where we routed information so
06:02when someone uploaded a file to the file
06:03server that got announced into IRC when
06:05their daily stats run was ready we
06:07pumped the stats and IRC when there was
06:09a database alert instead of emailing it
06:11or paging people that waited IRC and the
06:14positive feedback loop was of course
06:15then people paid more attention to it
06:17and we kept on routing more information
06:19so the game didn't work out
06:21obviously we shut it down and at the
06:24point we shut it down we realized he
06:25would never work without a tool like
06:27and instead of being a hacky series of
06:29kind of jury-rigged fixes on top of IRC
06:32and all these kind of poojas we should
06:33start again with what our dream system
06:35look like for this kind of thing and
06:37build that and that's luck so okay so
06:41here's a easy one then so that describes
06:44the how you got there's actually a hip
06:46sting as a story for me personally like
06:48you you ended up with the same way we
06:50were trying to build the spec for office
06:52and of course we didn't quite know what
06:54to build so we built like a spec place
06:56to put all the the Word documents and
06:58spreadsheets and that became SharePoint
07:00and it was like 90% of the time on 10%
07:02of the problem building a tool and I
07:04think a lot of the tools in the space do
07:06you arrive arise from that um here's a
07:08pretty straightforward one that I think
07:10a lot of times people always find tricky
07:12I think it's really important for for
07:14particularly for enterprise products
07:16like to be able to describe them
07:17succinctly and tell people what it is so
07:19what's your your longer than a tweet
07:21less than a blog post version of what
07:23slack is today and where it's heading
07:25well you know I think I can actually fit
07:26it in a tweet and then there's two
07:28alternatives the half tweet length
07:29version or the 45 minute version so I
07:31personally don't know how to do any kind
07:32of tweets on offer you give it a shot we
07:35say all your team communication in one
07:38place instantly searchable and available
07:40wherever you go and I can unpack that a
07:41little bit but the first part is all
07:44kinds of messages that you might send so
07:46they might include an announcement about
07:48a policy change for HR the spec or the
07:52you know the goals for q4 it might be
07:54I'm five minutes late for this meeting
07:56can you guys please apologize you know
07:58any kind of messaging that sort from
08:00person to person or person a group and
08:04communication that happens outside of
08:06the context of your regular messaging
08:08system and so that sounds a bit weird
08:10but what I mean is in our case every
08:13time someone tweets at us at our company
08:14that goes into our slock instance every
08:16time a the customer creates a ticket
08:19through Zendesk that goes into our stock
08:21instance every time someone signs up for
08:23the service through our own internal
08:24integration that gets pumped in and
08:26every time a new team is created we use
08:28pager duty for monitoring and alerting
08:30and that will send notifications in the
08:32slack as well when we create a bug
08:34that'll get pump into slack I can keep
08:35going like this Ford doesn't have a
08:36couple dozen examples but whatever tools
08:40people use that email or can be made to
08:43amend some kind of message that can go
08:45into slack as well second part of it was
08:47instantly searchable so we put a lot of
08:49assisting a search so first of all just
08:52you know the quality of the search is
08:53very good and we have a unique approach
08:54to message search which we can get into
08:56it's interesting but I can also search
08:59for things like specific phrases that
09:01people have used when they're tweeting
09:02at us or keywords inside of help tickets
09:05that have been directed us when I'm
09:07looking for someone slack account
09:10because they've mentioned a problem or
09:11or something to me I can search with the
09:13email address associated with their
09:14slack account through slack itself and
09:16all those things rather than going to
09:17some separate tool and then the last
09:19thing available where you go again is
09:21just mobile yeah it's interesting I mean
09:25quite a lot of what you're talking about
09:26is things that are not actually IRC and
09:29aren't really chat and not necessarily
09:31things that you would do with email
09:33that's to say you're kind of finding a
09:34new sort of interaction model for
09:37solving problems it would previously be
09:38done by going looking at a database or
09:40going to a website or editing your
09:41document or going to a spreadsheet or
09:43Google sheet or something and you're
09:45kind of going down to what's actually
09:47the thing that we're trying to achieve
09:48here and say well actually that could be
09:50done through this entirely different
09:52model you know we don't need to put a
09:54spreadsheet here we could do it in a
09:55conversation or through an API feed or
09:58something like that and then all of a
09:59sudden that doesn't need to be a web
10:01browser or doesn't need to be Microsoft
10:02Office or Salesforce it can be this much
10:04lighter weight more kind of universal
10:06more portable interface so suddenly you
10:08can kind of go and see your search all
10:09of your help tickets in an iPhone app
10:11rather than having to have an iPhone app
10:13from that particular provider so it'd be
10:16kind of it makes it much more kind of a
10:17universal interface in some ways yeah
10:18I'd even be willing to admit that in
10:20many cases using a purpose-built
10:22specific to offer each of those items
10:24might be the better experience but the
10:27value of having them all in one place
10:28just so completely overwhelms any
10:31benefit you can get by spotting them
10:32apart and this is maybe a slightly
10:34different topic but one of the shifts
10:37that we're trying to take advantage of
10:39is in contrast and I had this
10:41conversation with Stephen before but in
10:43contrast to when I got started making my
10:45living with computer stuff which was
10:46that a mid late 90s what Microsoft
10:48completely dominated everything now for
10:51his customers chose Microsoft yes it's
10:55active domination well you know because
10:57it was just it was the it was the
10:59environment in which we swam to a
11:01certain said you know I've worked at a
11:02design agency and we used windows on the
11:04desktop we use windows for workgroups
11:07for our local networking we used
11:09Microsoft server technology for the
11:11sites we developed we use office the
11:14project manager used MS Project and we
11:15had custom written some VB script let's
11:18connect that to Outlook for tasks the
11:20original full spec yeah and but I mean
11:22there was ie and there was active acts
11:24and there's Ola and all these things
11:25that made Microsoft stuff work well
11:26together and now for every or nearly
11:29every product category that Microsoft
11:31once dominated there's a dozen good
11:32vendors most of whom are cloud-based and
11:35there's you know there's categories that
11:36are almost unrecognizable compared to
11:38their late 90s form like CRM host
11:41salesforce it's a totally different
11:42thing and there's product categories
11:44that didn't exist then like application
11:45performance monitoring is is a whole new
11:48yeah and then definitely mobile so in
11:51one sense from from the perspective of
11:54the business customer at the world is
11:55better because the software is is easier
11:57to use it's much much cheaper it's much
12:00much easier to manage and deploy it's
12:02more powerful it's simpler you know in
12:04every respect it's better except that
12:06nothing works together and this isn't
12:08anyone's intention but if we manage our
12:11customers issues in Zendesk and we
12:13manage our developer issues in github
12:15those are two things that ought to be at
12:18least accessible in the in the same
12:20forum but there's siloed and again this
12:22is not it helps attention it's not Zen
12:25desk intention it's just they want to do
12:26the thing that they do very well and
12:28they do the thing that they do very well
12:29and that is not just for those two
12:33product categories but for the 30 or 40
12:35or 50 tools a company might use across
12:37the board you know everything from
12:38marketing analytics to bi document
12:41creation editing data warehousing and
12:44there's just there's so many services
12:45these days and we keep on inventing new
12:48categories and it's just so scattered so
12:50there's huge value bringing it all
12:52wow I think that's super acing I that
12:55while was genuine because one of the
12:57things that like I was thinking about
12:58like a the typical ITU answer to 50
13:02different silos and then overlay the
13:05mobile one because every
13:06is like I got to do all this in the
13:09browser and then by the way I have to
13:10figure out that whole reflecting it in a
13:12mobile as well but the typical answer is
13:15to make them all work together is sort
13:17of this very heavyweight integration you
13:20either go to all the vendors and try to
13:21get them to agree on some high fidelity
13:23data interchange which is impossible and
13:26the cloud almost says well that's sort
13:28of not the way that we would think about
13:30solving the problem or you you try to
13:33think of it from like the the dashboard
13:35client perspective and get everything to
13:38integrate at some deep deep level like
13:41okay I'm a Salesforce person and we now
13:44need to have you know Zendesk stuff show
13:47up so now I'm going to write a bunch of
13:49code to get Zendesk to be displaying in
13:51the middle of my salesforce display but
13:54the problem is is that that solves the
13:55sort of the fringe case of the person
13:57who's like constantly immersed in both
13:59but there a lot of what you were saying
14:01is also this well there's a lot of uses
14:03to connect all these things but you
14:04don't have to be so heavy about the
14:05whole thing because so much of the usage
14:07is not this massively weighty challenge
14:10is that something yeah no I think in one
14:12case you end up with a very lowest
14:14common denominator kind of integration
14:16which is smashing down whatever kind of
14:19data into a message but I think that's
14:21saved by the fact that almost everything
14:24has a URL these days so you might just
14:27get the short summary of the item you
14:28might just get a little bit of the
14:29information and obviously not have the
14:31full functionality of that third-party
14:32application available to you inside of
14:34slot but if you have URLs you can do
14:36anything so you can just go back to the
14:38original source and that's this you know
14:39very very lightweight but very powerful
14:42the thing that strikes me listening to
14:44you which is if we think back to our
14:46computing experience 15 years ago you
14:48had a network drive with 150 folders and
14:50you had no idea what you folded the
14:51files for it but once you found it you
14:54double-click on it and then the
14:54application would open and then you go
14:56to the cloud and you've got 150
14:57different applications and you've got no
14:59clue which of those applications the the
15:01meeting notes you're looking for and
15:02might be and they might be in Evernote
15:03they might be in Salesforce so they
15:05might be in vend s it might be in this
15:06or that of the next thing um and
15:08theoretically you could have gone
15:10searching network drive no matter how
15:11well that would have worked but and that
15:13would have again that would have kind of
15:14given you that lowest common denominator
15:16and a project file that said this is the
15:18thing for this client and then you'd
15:19open it and you're almost kind of giving
15:21that kind of experience again that thank
15:23you you can find all the stuff and then
15:25you open it in the thing that created it
15:26yeah we actually we often use the
15:29analogy of what it was like for us and I
15:32think this is true from for many many
15:34people and certainly most listeners of
15:35his podcast and what it was like to
15:37switch to Gmail from another email
15:39client where it wasn't just the fact
15:40that you could search for the messages
15:42and find that again after which the
15:43obviously is very important but the
15:45cognitive cost of processing incoming
15:47messages felt at the floor in comparison
15:49to a world where you had to manually
15:51choose the folder in which you'd file
15:53something away so suddenly it didn't
15:55really matter you could just throw it
15:57you could archive messages willy-nilly
15:58because he knew that if you ever needed
16:00to find them again the search was good
16:01enough that you'd be able to i'm this is
16:02in contrast to my experience and i think
16:06most people's experience and I'm trying
16:08to I'm hedging so much because you're in
16:10the room but uh feel free you know got
16:13it you know what you want make a clippie
16:14joke i've using outlook which you know i
16:18always had a search feature but it's
16:20more or less a nominal search feature it
16:22yeah okay where I promise didn't work it
16:23would take four or five minutes with
16:24results to come back they weren't good
16:26and so you might as well have not had
16:28the search so the experience was exactly
16:29like you're saying benefit of having to
16:31like try to remember what folders
16:33something is in and we have this all the
16:35time now you know there was a McKinsey
16:38study and I can make any consultants he
16:40come up with a study this has anything I
16:41want them to say however I think this is
16:43it's probably pretty accurate about the
16:45amount of time people spend using email
16:47and one of the other aspects of it was
16:49the amount of time people spend looking
16:50for stuff yeah and what I need a stat
16:53because I want to put it in a
16:54presentation and it was in an article
16:56that someone sent to me I don't remember
16:57the publication I don't remember the the
17:00the title may be I remember the person
17:02who sent to me but I don't remember the
17:04medium through which they sent it
17:05whether it was a text it was a Google
17:07hangout message of him email it was in
17:09any one of these you know there's a
17:10comment on a task in a sauna or who
17:13knows we're having everything come
17:16together in one place and making it
17:17searchable has incredible value well
17:19it's fascinating case you know the this
17:20whole one of the biggest transformations
17:22that the that that the cloud enables is
17:25sort of this infinite you know secure
17:27reliable storage kind of thing
17:29of course not technically infinite but
17:30but that what happened is is there's
17:32transformation from you know everybody
17:34on their device being an Mobile is a
17:37huge part of this because you your
17:38mobile devices don't have the storage
17:39that you can save everything and so now
17:41everybody's trying to figure out like
17:43which part goes where when really you
17:45just leave it all on the cloud and then
17:47you never have any of it and that's like
17:48way more secure the Sony breach doesn't
17:50happen and all this stuff but people you
17:52are used to spending all of this time
17:54like being filers like instead of a
17:57Pilar like that was the acronym we
17:58always a little saying we did without
18:00look was all your a Pilar or a filer and
18:02and the problem was and I so this is the
18:06question really is the typical folks in
18:09IT they are trying to figure out an
18:11information sort of management solution
18:14for the for the enterprise or the
18:15company that actually supports filing
18:19like if you so I'm curious how they how
18:21some of your early customers would slack
18:23and how they see it because their their
18:24their first motivation or their first
18:27responses okay first I need the finance
18:29department that I need the legal
18:30department then I need the accounting
18:31department within finance I need
18:33corporate development and I need
18:34accounting and I need tax and then
18:36within tax I need domestic and Internet
18:37and they they have these hierarchies
18:38which often look just like the org chart
18:40when they're done even though they
18:41they're not and and really what you're
18:44saying is that's the part you don't eat
18:45like because you spend all that energy
18:47and if you just keep pushing it will
18:49we'll find it and if the weird part for
18:51me is always people want to do that or
18:54that people don't want to do that and
18:55then they go on use Google and then
18:57Google has no big giant hierarchy in
18:58fact Yahoo sort of lost that battle over
19:01having a heart and so how does that
19:02conversation with enterprise customers
19:04go in terms of thinking like don't spend
19:06all your energy doing that like just get
19:09well I'm not sure if we actually have
19:10that as an explicit conversation yeah
19:11yeah tends to be very bottoms up but
19:14that's actually I've never heard that
19:15pilot versus filers then that's that's
19:18sort of perfect from my perspective
19:19because the degree to which you can
19:22remove friction even in the tiniest
19:24little instances makes a world of
19:27difference you know that this is a
19:30slightly different example but one of
19:33the very first integrations we built for
19:35our own uses and slack wasn't bug
19:37command so I would type the slash
19:39and the word vug and then at somebody's
19:43username and then a bunch of text and I
19:45hit enter and a bug will get created
19:47assigned to the person whose name I
19:49mentioned and then all the tax will
19:50become the subject of the bug and then
19:52the URL will get echoed back to me in
19:54slack where is the bug go the bug goes
19:56to our bug tracker so it's its own
19:58internal tool but it would work you can
20:00do the same thing with asana now and I
20:02think you can do the same thing with
20:03Geryon if you can't yet then it's coming
20:05soon the DOL gets echo back to me so I
20:09can click it and then add some more
20:10detail but the difference is and this is
20:12the stuff that you lose the friction
20:15that you lose and it doesn't sound like
20:17much but it ends up being a lot and
20:18that's switching away from slack to my
20:20browser command T to open a new tab
20:23starting to type the URL the bug tracker
20:25waiting till I've typed enough
20:26characters that I recognized it hit
20:27enter wait for the pages load hit the
20:28new bug button and then start to well
20:31tend to factor off and like oh yeah
20:32exactly but that little bit of friction
20:36meant that we were a lot better and more
20:38disciplined about the creation of bugs
20:40because when we had some conversation
20:42then we realized okay I see that is a
20:44bug and now someone needs to report it
20:46it wasn't this giant ordeal and I mean I
20:49say giant horde yellow that's what it
20:50feels like just to have to switch
20:52applications and open a new tab and I
20:54think I think this is there's a strand
20:55running through all of this which is a
20:57lot of the kind of the heavyweight
20:59things that you one would do with a
21:01mouse and a keyboard and a web browser
21:04or PC application and a file browser you
21:07kind of need a PC so you want to do that
21:09full book bug tracking thing you're not
21:10going to do that on a smartphone
21:11standing in the street because you've
21:13just seen an issue um
21:14and so you'll then you'll say well you
21:16know I need a PC to do this job you know
21:18I need a big PC and I need a big
21:20keyboard and a big screen but actually
21:22that's just because what the terms the
21:23tool the tool are you using required
21:25you'd have a big screen in a keyboard
21:27and a mouse and actually the underlying
21:29task is I just kind of need to send this
21:30little packet of information or make
21:32this decision or give this piece of
21:33information and pass it to somebody um
21:35and so when you boil it away from what
21:38was the tool that we were using - what
21:40is the underlying objective then all of
21:42a sudden the device you can change quite
21:43a lot yeah and again it does and that
21:46lowest common denominator so if you use
21:48something just to extend this is the
21:50example if you use something really
21:52complicated like Bugzilla or Jarrow or
21:54some more sophisticated bug tracker
21:56which might have 14 fields that you're
21:58supposed to fill out for severity and
21:59priority and reproduction steps
22:01inversion it applies to and a screen
22:02shot and all that kind of stuff you
22:04won't get to do all of that but it's not
22:06something that you're gonna do while
22:07waiting in line at the bank anyway
22:08you're just gonna type out the very very
22:10bare bones basic but it seems to me I
22:13mean that that metaphor though is sort
22:15of consistent with like keeping the
22:17organization going because the
22:19alternative is you know like I
22:21experienced the bug and now it's on my
22:22own personal to-do list and I'm probably
22:25gonna drop that whereas at least if I
22:27put it on with just the title and the
22:30minimal information now the organization
22:31knows it and someone can help me and
22:34keep sending me the URL saying will you
22:36finish filling this in because you know
22:38you were online at the grocery store
22:39with the bug yeah so that it seems to me
22:42that that again the this benefit of
22:44having this is it's not just that it's
22:46mobile it's just that it's connected to
22:48all of these line of business systems
22:50that keep the organization rolling yeah
22:52and it doesn't slow down the mobile
22:54versus desktop distinction is kind of
22:55interesting for us too because we
22:58explicitly said at the beginning and
23:00this is stilled in our experience that
23:01slack is not really designed for a
23:03mobile only workforce and not designed
23:06for jobs where people are going to have
23:08a computing device in their hands most
23:09of the time so we automatically
23:11disqualified say retail food service
23:13healthcare 70% of the u.s. workforce but
23:18that the other 30% are people who will
23:20at least at some point during the day
23:22use a desktop so I think that the slack
23:25is probably not use or if one person on
23:28the team is mobile only that's fine if
23:31660 percent of the team is mobile only
23:33that's fine if a hundred percent of the
23:34team is mobile only it might not be a
23:36great product for them because there's
23:37that little bit of utility that that
23:41it's difficult to deliver on a phone we
23:46have I think I don't have the latest
23:48stats on this but let's this is me close
23:50enough 99% of our daily active users
23:53will walk into slack at least once on
23:54that day that directive on the desktop
23:56but only about 65 percent will do the
24:00yeah that we should look at our
24:02andreessen horowitz members because I
24:03think we're all like we might be an
24:05interesting use case because that
24:06because we're very mobile and I think a
24:08lot but what is it let me just ask this
24:10in terms of the usage then what the
24:12average customer who gets the person the
24:15bottom up deployment who signs up what's
24:16the first thing people do it's like
24:18where it feels rewarding and it brings
24:20in more people it's a good question and
24:22it's hard to answer because it's so
24:24variable you know we look through in
24:27there's obviously of critical importance
24:29s to figure out what the patterns are in
24:31successful implementations versus
24:33unsuccessful ones but there's so much
24:36variability we can take a whole year in
24:37some cases for the team to really
24:39convert there's a couple of initial
24:40evaluations and then it'll die off and
24:43it'll be a few months before they get a
24:44small team using it and then the small
24:46team will use it for six months before
24:48the whole company switches over the this
24:54is not a Eureka moment for people this
24:56is kind of a slow boil but I think the
24:57ultimate difference is and this is the
24:59big contrast between slack and an
25:02organization that's primarily driven by
25:04email usage or communication primarily
25:06happens by email where in the email case
25:09everyone has a small slice and I'm
25:10gesturing with my hands to make a very
25:12small slice of a big wedge of the
25:15communication that's happening around
25:16the company and that all the rest is
25:18completely opaque to them and when they
25:19leave the company their slice just
25:21disappears it's gone forever and even
25:23more important when someone joins they
25:26start with an empty inbox they start
25:27with literally nothing despite the fact
25:29that it might have been tens of millions
25:30of messages passed back and forth in
25:33contrast when someone joins a team
25:35that's using slack and this is why I
25:36think it might take a while for them to
25:37get the the most important bit of the
25:40value they can scroll back through
25:42everything they can search through
25:44everything and of course they don't want
25:45necessarily to get everything in our
25:47case on our team we have three and a
25:48half million messages in the archive but
25:51they can get a sense of what's going on
25:53now what's important now they can get a
25:54sense of how people interact with each
25:56other and they can get a sense of who
25:59knows the answers to all kinds of
26:00questions and who really makes the
26:01decisions and of course they also
26:03through search have access to every link
26:05that's imposed at every documents and
26:06share at every decision and every
26:08discussion and you can go back and
26:10figure out why why do we do this crazy
26:13see the origin of that conversation you
26:15can see what happened the last time this
26:16issue arose and how we don't with it and
26:18all that kind of stuff so it's a lot
26:19like you know the sort of a Twitter
26:21phenomenon that you see which is
26:22somewhat new follows you or you follow
26:24someone new and then you go to their
26:26history mm-hmm and then all of a sudden
26:28that's when you start seeing the you
26:29know December 14 you know 2007 favorite
26:33that's a great example I mean I hadn't
26:35thought about that before but that is
26:36you know having this a pre-existing
26:39stream of information that exists that
26:40you you can dip your toe into whenever
26:43you like you can have a glimpse into and
26:45that's actually maybe the second part of
26:47the value of slack which is transparency
26:49across the organization and transparency
26:52seems like a loaded word because it has
26:53all these political connotations I don't
26:55mean Edward Snowden or any kind of
26:57political kind of transparency I just
26:58mean literally the opposite of a pasady
27:00so that for example the Technical
27:03Operations team can see what's going on
27:04with customer support so they're having
27:07a lot of people reporting this issue
27:08over and over again and somewhere on the
27:10tech ops team says oh crap we just
27:12changed a load balancer I bet that's why
27:14rather than that'd be an issue that's
27:15escalated to the customer support
27:16manager and then the next day's stand-up
27:18meeting the customers more manager
27:20relays it to the manager tech ops and
27:22goes down you know or any of the
27:23engineers can see what the designers are
27:25coming up with next the marking team can
27:28see where the sales team is having
27:29issues and they need better support
27:31documentation all that kind of stuff
27:32cool so um you know what okay so then as
27:37the usage goes on so there's lots of
27:39different ways to start what's the thing
27:41that you think people stop using like
27:44what kind of products have select
27:45displaced I don't mean that to be in an
27:48aggressive sales kind of way but just
27:50you know wow maybe they stopped use do
27:52they stop using the front-end to some
27:53things or today I think it's a good
27:56question I don't think it's sales at all
27:57because one of the things that we
27:58realized early on is if we don't either
28:01replace or consolidate some other form
28:03of communication or interaction between
28:05people then this is going to be another
28:07thing that they have to do and if it's
28:09another thing they have to do then it
28:10will inevitably fail and I don't want to
28:12sound like I'm picking on them but that
28:14was my experience of Yammer and many
28:16other people's experience of EMR and
28:17products like that like in that same
28:18category chatter and convo social cast
28:22where it wasn't quite sufficient
28:25by the way what I'm thinking this
28:27probably gonna happen with Facebook it
28:28worked - it wasn't quite sufficient to
28:30replace email usage internally or wasn't
28:32quite sufficient to replace use of
28:34messaging internally or whatever it is
28:36right and so it becomes this other place
28:38and what we found is that successful
28:40slack is very binary if we get to 80
28:42percent of the of the team using slack
28:44it might sound like I'm going to say
28:47something good but no zero percent will
28:49be using it soon it has to be a hundred
28:50percent it has to be you have to be able
28:52to trust that if you send the message in
28:54in this way then people will see it in
28:56the same way that if you send someone an
28:57email you know that they will see it
28:58they might hate you they might have too
29:01much to do it might get lost they may
29:02never respond to it because they have
29:04too much other stuff going on but you
29:05know that they will see it whereas if
29:06you can't trust that a message sent to
29:08this meeting will reach the group or the
29:10individual that you're trying to get to
29:12then eventually you'll stop using it so
29:14in most cases the thing that it will
29:16replace is email usage and I want to
29:19just stress because we get this in
29:21headlines quite a bit then we want to be
29:23an email killer emails not going away
29:25we've got you know at least three or
29:27four decades left of heavy email uses
29:29because it is lowest-common-denominator
29:30and I mean that in a good way yeah it'll
29:33cross organizational boundaries very
29:34easily it's how we coordinated this
29:36meeting today but I think if you use
29:39email as your primary means of a
29:41communication internally for the kinds
29:43of knowledge worker for lack of a better
29:45term teams that use slack that's your
29:48foolish I mean and it's such a big
29:49disadvantage that you will eventually be
29:52made to switch by losing out to your
29:55all right well I mean I think that that
29:58in a sense it's the practical answer
30:00because showing up the saying good
30:01replace email is a very challenging
30:03value proposition given that companies
30:06need it to run but you know certainly if
30:09you ask a reporter I um use Twitter as
30:10an example like Twitter has changed the
30:12workflow of being a reporter and and it
30:16changed their nature of work it changed
30:17where tips come from and leads and who
30:19meets their story and how they talk
30:21but it seems like the same thing could
30:23happen it said once you get to 100%
30:25slack that like the daily workflow of
30:28somebody is different and then you know
30:30mobile seems to fit into that as well so
30:33do you think that in the places that
30:35have really deeply engaged with slack
30:36outside of this building that
30:38their workflow is really different that
30:40their nature of work has changed yeah
30:41and I think you know it slack is not
30:45some passe it's not an it's not magic
30:48it's not going to transform people by
30:51itself it's a tool yeah and it's a good
30:52tool but people will especially in the
30:54transition period struggle to figure out
30:57how they're supposed to use it most
30:58effectively it can feel like it's
31:00noisier potentially it can feel like
31:04because you have established patterns of
31:07behavior both in terms of your own daily
31:10routine like you know what the points at
31:12which you stopped to check email and in
31:15how people are expecting to behave like
31:17so for example the meeting notes are
31:18circulated after the meeting by email or
31:20something like that it can be a very big
31:24change and that's that the you know we
31:25think of that as a request that were
31:29were making of our customers you know
31:33that we were asking them to change a
31:35very fundamental and in some cases very
31:36ingrained set of behaviors that all
31:38interact and we're asking them to change
31:40it at the group level which is even more
31:41challenging than the individual level so
31:43it's a very very big ask and this is
31:47maybe a little bit of a sigh but that's
31:48why we focus so much on new user
31:50experience on communicating the value
31:51upfront and all those kinds of things
31:53but it can often take a while before
31:56people find their legs and even in our
31:58own case you know we're the people who
31:59make slack when we first started it was
32:01eight people working on the team and so
32:03slack was absolutely perfectly designed
32:05for an eight-person team and if you were
32:06like a 12% team's a piece of crap we
32:10found that out pretty quick because when
32:12we first bagged our friends use it we
32:15got our do is one of the very first
32:16external customers and not long after
32:19they started using it they had 120
32:21people on it made all these complaints
32:22and we didn't ever anticipate you know
32:24things you didn't really understand but
32:26even in our own usage as we've gone from
32:278 people to know how 100 people
32:29ourselves we had a very very mature with
32:34the board is like like consensus-driven
32:36or collaborative approach to
32:37decision-making where everyone gets
32:38their two cents in which is fine at
32:40eight people and fine if 15 people I'll
32:41even find at 25 people and then is kind
32:43not so great at fifty people and it's a
32:45disaster with a hundred people because
32:47just it's the amount of time you're
32:49taking up to have everyone give their
32:51input and then sheer number of
32:53messengers doesn't work so we've had to
32:55adjust our own behavior in light of that
32:57it's you know it really depends on the
32:59nature of the organization but the kinds
33:01of changes that that we typically see or
33:04have reported back to us are of course a
33:06drop in internal email usage and in some
33:08cases completely you know not using
33:11email at all internally which it was a
33:12side effect that we hoped for because
33:14again having all the messages in one
33:16place from the big memo down to the I'm
33:18going to be five minutes late is value
33:20the other one is canceling a lot of a
33:26very specific type of meeting which is
33:27the daily standup or the status report
33:30kind of meetings because there's a
33:33steady flow of information emanated from
33:35each team so if you want to know what's
33:36going on with with technical operations
33:38you just open their channel and have a
33:40look if you want to see what's going on
33:41with the front end engineering you have
33:42a look if you want to see what's going
33:43on with the sales team or what accounts
33:45are being closed or any of those bits of
33:47information you can just look for
33:48yourself I mean it's a classic you kind
33:50of use case I always talk about which I
33:51think of my friend is in big companies
33:53who once a week or once a fortnight they
33:55have to go in into an internal system
33:56and put on a bunch operating data and
33:58then drop that into Excel and make
34:00charts and then put the charts into
34:01PowerPoint and then what bullet points
34:02in the PowerPoint and then email that to
34:04a team yeah and you can't do that on a
34:06smart phone you can't do that on a
34:07tablet but actually that should be a SAS
34:09dashboard and it should be a
34:11conversation in something like slack or
34:13it could just be you know Kate for
34:14Charles change you paste it into the
34:15slack Channel and you write why and then
34:17you're done and then now that's not two
34:19hours once a fortnight and it's not a
34:21paper copy of PowerPoint so let's super
34:27yeah I think we could keep going on this
34:29for a long time because we and maybe
34:30we'll end up doing a part two at some
34:31point to talk more about the depth of
34:34this but I you know kind of just want to
34:37ask in a sense like because this space
34:40has always had a bunch of products and
34:41they've always had a little trouble
34:43getting traction and yet slack is just
34:45taking off the growth you know Mark has
34:47tweeted the growth curves and we see it
34:49I mean every new company I see is using
34:51it and I've been up to Seattle and I've
34:53seen companies using
34:54I've seen people using it in DC I'm you
34:58must as a as an organization and you
35:00personally sort of have a belief or an
35:03insight about how people use tools that
35:04other people sort of either question or
35:07don't share what do you what do you
35:09think is that sort of secret magic means
35:11that slack holds that other people might
35:15not it's a good question and at the very
35:18beginning I talked about my own history
35:20and like most people with me itself are
35:22trying to build task management software
35:24at some point one of the you know when
35:27we were first started on this and we
35:29thought about that the core feature said
35:30and what we wanted to get in and what we
35:31didn't want to be do we decided we
35:35wanted to avoid anything that has too
35:37much structure or too much ideology
35:39behind it and so task management would
35:43be one of those things although there's
35:44a whole world and you know moving
35:46towards the the piling the metaphor that
35:49you used before versus the the finally
35:50metaphor having something that's as
35:54loose as a message which can include a
35:56link or it couldn't be a file that's
35:57uploaded it could be short it could be
35:59long they could have formatting it could
36:01not have formatting you know and and
36:03just having the basic model of people on
36:05the team or services you've integrator
36:07the team can emit messages into channels
36:09and you can join into these channels as
36:10you like having that very lucid
36:12perception it covers a wide variety of
36:14use cases so I think that's the key
36:16thing because the the more specific the
36:18tool the more constrained it is the less
36:19general its application will be and the
36:22less like Kadel work in any specific
36:23news case and just you know one more
36:25point on that when we thought about what
36:26we would what we would build as a task
36:29management some piece of bit of software
36:30and this is something that again that
36:31I've done before in the past you have to
36:34have some ideology about how work gets
36:37done you have to have some ideology
36:38about you know what priority means or
36:42what an assignment means or due dates or
36:44severity or any of those kinds of things
36:46how tasks should be described or defined
36:49and how they should be presented and
36:51ordered and all those kinds of things
36:52and the more specific you get and the
36:55more that ideology gets cooked into the
36:57model or the schema I mean in a literal
36:58sense the database model like the field
37:00names and the more columns yeah
37:03the the less likely that's going to work
37:05for any given individual because people
37:08are very idiosyncratic and groups of
37:09people are are like a multiplication of
37:12all the intricacies of the people making
37:15well thanks you know actually I have to
37:17say that it like it is an insight like I
37:20think that every tool that I've known in
37:2225 or more years has always started from
37:25like first we're gonna define our
37:26structure and then we're gonna define
37:28the methodology for how work is done and
37:31then we're gonna define the methodology
37:32for how it's stored and and catalogued
37:35as well and so that seems super unique
37:37because great type of chance to talk to
37:38Stuart Butterfield and with Benedict
37:40Evans here as well and so thanks so much
37:42this has been another a 16z podcast