I do love the point about algorithmic curation of Slack. It's one of my struggles with Mastodon... sometimes I want the reverse chrono feed, but quite often the algo on twitter shows me things I do like.
The challenge is training the algo on what the business needs and not what gets attention. The worst thing you could do is have tik-tok for Slack which just shows me all of the content that appeals to my monkey brain (whatever that may be).
I use reminders a lot in slack to get to inbox zero but then inevitably I have a list of messages to triage in my slackbot channel which isn't a great experience oftentimes.
I have to plug https://about.gitlab.com/handbook/communication/#slack as well - we had a 90 day retention policy at GitLab and I would've loved to see an even shorter one. There was nothing better for forcing a more async and documentation-first culture in my opinion.
Funny enough, we just had the retention policy debate internally at Mode, which was ultimately the thing that got me to write this post. I'm with you on that though - I think the moral hazard from "eh, this will be saved forever, I don't need to do anything else with it" creates a lot of bad habits in Slack. (But, if been treating Slack like a permanent archive for almost a decade, you also can't really just purge it either...)
On the algo feed thing, this is actually something that Yammer (which was basically FB for work) did really well. Conversations defaulted to threads (like FB), so you'd never have thread starters talking to one another. That meant the ordering of the threads didn't really matter, and if a thread got a new reply, it would bubble to the top of that channel's feed. It sort of operated like a manual retweet of the thread starter.
That meant that 1) important stuff would stay at the top of the feed for as long as it was active, and 2) you could browse the top 20 thread starters for the day and get a really good summary of what went on. And if you wanted to get into the details, you could just open up the thread and read the conversation.
It didn't have a lot of the shiny edges that Slack has, but structurally, it was by far the best corporate communication tool I've ever used.
Yeah, the tough thing is, this is a really hard algorithmic problem. If Slack were structured differently (like a forum and not chat), there are simple ways to bubble important stuff up without a bunch of "recommended for you" complications: https://benn.substack.com/p/the-product-is-the-process/comment/11210909
- I'd start by using some of the existing AI chatbots to provide a summary of what's said in public channels
- this combined with some tuning for an individual user might get close (e.g. show me a sample of conversation threads, I get to upvote or downvote whether they are relevant to me, create a relative score to compare *interestingness* to all Slack users vs to me
- even if the only thing that ever happens is better automated summaries, this would be a net win
I think my question on the summary thing would be, do you think people would actually read it? I feel like people have been writing corporate newsletters since the beginning of time, and I get the impression that they rarely get read. (They feel more like a way for people to say "I shared this with you so you should've known" than anything.) A slack daily digest feels like that to me, where we think we'd read it until we start getting it every day. Though maybe that's ok. It's like getting the newspaper everyday; even just looking at the headlines haphazardly works, as long as you do it regularly.
When I was at Yammer, I think we might've tried to do something like the trending or popular thing, and it either didn't work at all, or I'm making up in my head that we ever even tried it. (But also, that was 10 years ago, and the tech for figuring that stuff out is probably a million times better now.)
Eh, I don't think I agree with that. Open-ended and very malleable tools *might* do that, but tools eventually overwhelm processes. If you're a long-form, email-heavy company and you adopt Slack, I don't think you end up writing really long Slack messages to one another; you end up shifting more towards synchronous and staccato chat. People will only fight what a tool wants them to do for so long.
That's kind of my point, that tooling (be it communication tooling or otherwise) eventually dominates, no matter what your existing processes are. But I don't see that as tooling amplifying processes; I see it as tooling rewriting process.
Not sure I agree with this comment: “Slack doesn’t have any mechanisms for triaging work”
The data team I’m on has an alerts channel from our data pipelines, and we have a clear process built into the channel topic to mark something as yellow if in process and green if done (or non-issue)
Also, for personal triaging, if I can’t get to something actionable now or that I need to read, I’ll have Slack “remind me about it X” where X is tomorrow or in N hours
I get what you’re getting at with the post, just wanted to bring alternative perspective to that statement :)
Yeah, that's kind of my point. I hear you, Brent, that you can hack various workflows on top of Slack, but needing to wire together a solution like this is a lot to ask from a product that ostensibly exists to make that exact problem easier. Plus, I'm with Nicole that snoozing work kind of just compounds problem of Slack being all push and no pull.
Dec 16, 2022·edited Dec 16, 2022Liked by Benn Stancil
I've tried many "hacks". I turn off notifications for everyone other than my manager and the Chief Data Science Officer, but when I come back from building a viz for several hours, I've got to find where in each channel someone has mentioned me, then figure out the topic by going back to the beginning of the conversation, then figure out what action I am supposed to take when someone has posted "@Nicole???" It's awful. If it's important, it should be an email. If it's urgent, text me.
My only "hack" is to mute nearly every channel, and rely on mentions. Which is a bummer, because that ends up siloing everything in the same way that email does. But I'd rather that anxiety of having a thousand unread inboxes all the time.
Am I about to... set my slack status to “email me”? 🤔 And set up email slack notifications for important channels?
The thing that gets me about slack is they push you towards threads for organizing conversations and then the “unreads” view for threads is HORRIFIC. There is no good reason that threads can’t have some version of “all unreads” where you can skim messages without marking them as “read”. (Failure to build this into their fundamental architecture is still not a good reason.)
Plus the “oh people can search it and see their questions already answered” doesn’t really scale. And enforcing it makes you sound like an asshole--“hey can you help me with this?” “I know the answer but I won’t tell you because I want to teach you a lesson so search the channel first and try to piece it together from chaotic, fragmented half-discussions and then get back to me--even though I could have already answered your question in the time it took me to type this.”
This somehow hadn't occurred to me until writing this post, but Slack trying to put threads in a chat app seems impossible to really work. Threads work in forums and on places like FB because the feed doesn't need to be ordered, and because you'd never respond to a thread with another thread. But if there's a sense of it being like chat, thread starters are often replies to one another, so you have to keep them ordered. And you can't mix the two formats like Slack tried to do ("use it however you want!") because they're fundamentally at odds.
And I agree with you on the "look it up" thing, which is another place where the chat-based format seems to break down. If it were more forum or long-form communication, "look it up" seems a lot less rude than over chat. It's like if someone asks you a question in person and you know the answer and you're like, "this is a teaching moment, here's a library card."
Dec 16, 2022·edited Dec 16, 2022Liked by Benn Stancil
I think the problem here is that an army of poorly trained managers has been given permission to manage the largest number of individuals in all of human history. Never have we had so many 20 year olds managing so many people. The consequence is that a lot of this management responsibility is passed on to products. Rather than create the right culture, or best practices, managers rely on, for example, Slack to help all individual contributors figure out their role.
This is not necessarily a bad thing. But this does mean that hiring & onboarding processes have to be slower than with active management. And in these processes we need to give guidance to new candidates on the "right" use of tools in company X. For example, here is what I use for Asana:
Eeeh, maybe. Part of me things that that's still a tooling problem, where, to use a tool well, you have to use it in ways that feel unnatural. But I guess we learn how to use most things. Still, though, it seems weird to me that we need more of an instruction manual for how to use a to-do list right than we do to drive a car.
"The problem is Slack." ~ Slack has been around, in public, since 2013. I could be wrong but I don't think that alone explains something that was first observed 35 years ago.
So, I can buy the various harms that Slack perpetrates, and that better tools can solve them. But at the risk of having the point of the article whoosh over my head, I have *process concerns* (though they're interrelated to tooling).
Specifically, imagine a workplace that is remote-first, without Slack (or equivalent substitutes: Teams, Discord, a good ol' IRC server, etc). All of our notions of how you stay up to date with what's happening, build relationships with colleagues through little banal interactions and side chats... all of them are presuming an in-person environment where important async messages can be sent over email or deferred. In that world, we can concentrate for as long as we want to, and then swing by the water cooler when we need a break or some interaction. Relationship building can happen as you run out to grab lunch, or before / after a meeting. The latter happens partially in zoom calls, but most of the other formation of a team, and the synergies which makes it more than a sum of its parts, require a chat-like dynamic. And if you're not in-person, I'm struggling to even imagine what that looks like for a remote team.
In my pre-Slack-universe days, simple semi-async conversations happened over previous iterations of Microsoft messenger, message boards sometimes, or stuff on blackberries. And limited as that was, it never had to replace or compete with live, serendipitous human interaction.
So what are the working hypotheses? How do you have a remote-first company that makes people still feel like they're part of a team and know their colleagues to some degree, without the eternal distractions of Slack? If we can imagine a working paradigm for that, maybe we can then imagine tools to match. Maybe Gather.Town is on to something. But man, from here it just looks like a Catch-22.
Ahh, so, I think there's a decent medium here. Slack as a chat app is good, and I'm in favor of that; I think the problem is when it becomes the way that everyone communicates and makes decisions and all of that. My ideal would actually be:
- A tool built around longer form communication (Yammer was quite good at this to be honest) for most conversation. It's the one that's archived, searchable, etc. It's essentially fills the role that email fills, though ideally, isn't quite email.
- Slack (or some other chat service) is the for real-time chat. But it only keeps a couple weeks of history; there aren't that many channels; it's all chat with no threads; etc. It's meant to mirror in person chat, and nothing else. That way, you don't rely on it for making big decisions or keeping historical records; it's just an ephemeral way to coordinate, talk through something in real time, or create somewhat of a social dynamic.
"Second, unlike email, Slack doesn’t have any mechanisms for triaging work ... With Slack, the only two options are read and unread"
you never used the Slack Bot? you can right-click on any message that might need review later, then ask Slack to remind you about it again. it saves these reminders and you can access them anytime. you mentioned how email allows managing workflows by taking additional actions. if you right-click any message in slack, there are additional actions available
Reminders create another thing for me to be notified about later though. For me, I want to file things in different folders that I can go to when I have time, not have an alarm clock that I have to constantly snooze until it happens to go off at a convenient time.
I want a pull model, basically, and Slack is nearly all push.
I do love the point about algorithmic curation of Slack. It's one of my struggles with Mastodon... sometimes I want the reverse chrono feed, but quite often the algo on twitter shows me things I do like.
The challenge is training the algo on what the business needs and not what gets attention. The worst thing you could do is have tik-tok for Slack which just shows me all of the content that appeals to my monkey brain (whatever that may be).
I use reminders a lot in slack to get to inbox zero but then inevitably I have a list of messages to triage in my slackbot channel which isn't a great experience oftentimes.
I have to plug https://about.gitlab.com/handbook/communication/#slack as well - we had a 90 day retention policy at GitLab and I would've loved to see an even shorter one. There was nothing better for forcing a more async and documentation-first culture in my opinion.
Funny enough, we just had the retention policy debate internally at Mode, which was ultimately the thing that got me to write this post. I'm with you on that though - I think the moral hazard from "eh, this will be saved forever, I don't need to do anything else with it" creates a lot of bad habits in Slack. (But, if been treating Slack like a permanent archive for almost a decade, you also can't really just purge it either...)
On the algo feed thing, this is actually something that Yammer (which was basically FB for work) did really well. Conversations defaulted to threads (like FB), so you'd never have thread starters talking to one another. That meant the ordering of the threads didn't really matter, and if a thread got a new reply, it would bubble to the top of that channel's feed. It sort of operated like a manual retweet of the thread starter.
That meant that 1) important stuff would stay at the top of the feed for as long as it was active, and 2) you could browse the top 20 thread starters for the day and get a really good summary of what went on. And if you wanted to get into the details, you could just open up the thread and read the conversation.
It didn't have a lot of the shiny edges that Slack has, but structurally, it was by far the best corporate communication tool I've ever used.
Still waiting for Slack to produce a digest feature like https://getlowdown.com/
What I'd like to see:
- show me the most important message in the last x hours, based on recency/frequency of interaction
- show me the conversations where I needed to engage
- show me the top 3 conversations where my input was not necessary
Yeah, the tough thing is, this is a really hard algorithmic problem. If Slack were structured differently (like a forum and not chat), there are simple ways to bubble important stuff up without a bunch of "recommended for you" complications: https://benn.substack.com/p/the-product-is-the-process/comment/11210909
Agreed that this is not easy.
- I'd start by using some of the existing AI chatbots to provide a summary of what's said in public channels
- this combined with some tuning for an individual user might get close (e.g. show me a sample of conversation threads, I get to upvote or downvote whether they are relevant to me, create a relative score to compare *interestingness* to all Slack users vs to me
- even if the only thing that ever happens is better automated summaries, this would be a net win
I think my question on the summary thing would be, do you think people would actually read it? I feel like people have been writing corporate newsletters since the beginning of time, and I get the impression that they rarely get read. (They feel more like a way for people to say "I shared this with you so you should've known" than anything.) A slack daily digest feels like that to me, where we think we'd read it until we start getting it every day. Though maybe that's ok. It's like getting the newspaper everyday; even just looking at the headlines haphazardly works, as long as you do it regularly.
My instinct is that people would read it if:
1) limited in size (eg top 3 or 5 threads where you did not participate but are relevant to you
2) incorporated trending (this topic was among the top 3 discussed this week). Topics would need to be filtered so they weren’t too broad.
When I was at Yammer, I think we might've tried to do something like the trending or popular thing, and it either didn't work at all, or I'm making up in my head that we ever even tried it. (But also, that was 10 years ago, and the tech for figuring that stuff out is probably a million times better now.)
For better or worse, tools tend to amplify existing processes - or lack thereof.
Eh, I don't think I agree with that. Open-ended and very malleable tools *might* do that, but tools eventually overwhelm processes. If you're a long-form, email-heavy company and you adopt Slack, I don't think you end up writing really long Slack messages to one another; you end up shifting more towards synchronous and staccato chat. People will only fight what a tool wants them to do for so long.
We are actually talking about similar things. See Conway's Law.
That's kind of my point, that tooling (be it communication tooling or otherwise) eventually dominates, no matter what your existing processes are. But I don't see that as tooling amplifying processes; I see it as tooling rewriting process.
Not sure I agree with this comment: “Slack doesn’t have any mechanisms for triaging work”
The data team I’m on has an alerts channel from our data pipelines, and we have a clear process built into the channel topic to mark something as yellow if in process and green if done (or non-issue)
Also, for personal triaging, if I can’t get to something actionable now or that I need to read, I’ll have Slack “remind me about it X” where X is tomorrow or in N hours
I get what you’re getting at with the post, just wanted to bring alternative perspective to that statement :)
I get those reminders, too, then I snooze them because I am focusing again.
Yeah, that's kind of my point. I hear you, Brent, that you can hack various workflows on top of Slack, but needing to wire together a solution like this is a lot to ask from a product that ostensibly exists to make that exact problem easier. Plus, I'm with Nicole that snoozing work kind of just compounds problem of Slack being all push and no pull.
I've tried many "hacks". I turn off notifications for everyone other than my manager and the Chief Data Science Officer, but when I come back from building a viz for several hours, I've got to find where in each channel someone has mentioned me, then figure out the topic by going back to the beginning of the conversation, then figure out what action I am supposed to take when someone has posted "@Nicole???" It's awful. If it's important, it should be an email. If it's urgent, text me.
My only "hack" is to mute nearly every channel, and rely on mentions. Which is a bummer, because that ends up siloing everything in the same way that email does. But I'd rather that anxiety of having a thousand unread inboxes all the time.
Am I about to... set my slack status to “email me”? 🤔 And set up email slack notifications for important channels?
The thing that gets me about slack is they push you towards threads for organizing conversations and then the “unreads” view for threads is HORRIFIC. There is no good reason that threads can’t have some version of “all unreads” where you can skim messages without marking them as “read”. (Failure to build this into their fundamental architecture is still not a good reason.)
Plus the “oh people can search it and see their questions already answered” doesn’t really scale. And enforcing it makes you sound like an asshole--“hey can you help me with this?” “I know the answer but I won’t tell you because I want to teach you a lesson so search the channel first and try to piece it together from chaotic, fragmented half-discussions and then get back to me--even though I could have already answered your question in the time it took me to type this.”
This somehow hadn't occurred to me until writing this post, but Slack trying to put threads in a chat app seems impossible to really work. Threads work in forums and on places like FB because the feed doesn't need to be ordered, and because you'd never respond to a thread with another thread. But if there's a sense of it being like chat, thread starters are often replies to one another, so you have to keep them ordered. And you can't mix the two formats like Slack tried to do ("use it however you want!") because they're fundamentally at odds.
And I agree with you on the "look it up" thing, which is another place where the chat-based format seems to break down. If it were more forum or long-form communication, "look it up" seems a lot less rude than over chat. It's like if someone asks you a question in person and you know the answer and you're like, "this is a teaching moment, here's a library card."
Preach it! 🙌
A ha! Substack made you write this way didn’t it?
(wait, i don't get it?)
It's basically a Conway's Law joke. He's going for the niche multiplier on the humor.
oooh. Ok I should've been able to figure that out.
And, jokes aside, actually, kind of? When I first started this whole thing, I planned on writing a few short posts a week, like the blogs from the 2000s. But the aesthetic of the site + the email part make it feel more essay-ish, and so I think I stumbled into that format. If the site had been more like old blogs where full posts stack on top of each other (like this: https://archive.nytimes.com/krugman.blogs.nytimes.com/?module=BlogMain&action=Click®ion=Header&pgtype=Blogs&version=Blog%20Post&contentCollection=Opinion), I probably would've gone with the other format.
I think the problem here is that an army of poorly trained managers has been given permission to manage the largest number of individuals in all of human history. Never have we had so many 20 year olds managing so many people. The consequence is that a lot of this management responsibility is passed on to products. Rather than create the right culture, or best practices, managers rely on, for example, Slack to help all individual contributors figure out their role.
This is not necessarily a bad thing. But this does mean that hiring & onboarding processes have to be slower than with active management. And in these processes we need to give guidance to new candidates on the "right" use of tools in company X. For example, here is what I use for Asana:
https://tinyurl.com/asanabestpractices
Eeeh, maybe. Part of me things that that's still a tooling problem, where, to use a tool well, you have to use it in ways that feel unnatural. But I guess we learn how to use most things. Still, though, it seems weird to me that we need more of an instruction manual for how to use a to-do list right than we do to drive a car.
"The problem is Slack." ~ Slack has been around, in public, since 2013. I could be wrong but I don't think that alone explains something that was first observed 35 years ago.
Yeah, I didn't mean Slack was literally the problem; I more meant that technology can look like a big advance but can make us less productive.
So, I can buy the various harms that Slack perpetrates, and that better tools can solve them. But at the risk of having the point of the article whoosh over my head, I have *process concerns* (though they're interrelated to tooling).
Specifically, imagine a workplace that is remote-first, without Slack (or equivalent substitutes: Teams, Discord, a good ol' IRC server, etc). All of our notions of how you stay up to date with what's happening, build relationships with colleagues through little banal interactions and side chats... all of them are presuming an in-person environment where important async messages can be sent over email or deferred. In that world, we can concentrate for as long as we want to, and then swing by the water cooler when we need a break or some interaction. Relationship building can happen as you run out to grab lunch, or before / after a meeting. The latter happens partially in zoom calls, but most of the other formation of a team, and the synergies which makes it more than a sum of its parts, require a chat-like dynamic. And if you're not in-person, I'm struggling to even imagine what that looks like for a remote team.
In my pre-Slack-universe days, simple semi-async conversations happened over previous iterations of Microsoft messenger, message boards sometimes, or stuff on blackberries. And limited as that was, it never had to replace or compete with live, serendipitous human interaction.
So what are the working hypotheses? How do you have a remote-first company that makes people still feel like they're part of a team and know their colleagues to some degree, without the eternal distractions of Slack? If we can imagine a working paradigm for that, maybe we can then imagine tools to match. Maybe Gather.Town is on to something. But man, from here it just looks like a Catch-22.
Ahh, so, I think there's a decent medium here. Slack as a chat app is good, and I'm in favor of that; I think the problem is when it becomes the way that everyone communicates and makes decisions and all of that. My ideal would actually be:
- A tool built around longer form communication (Yammer was quite good at this to be honest) for most conversation. It's the one that's archived, searchable, etc. It's essentially fills the role that email fills, though ideally, isn't quite email.
- Slack (or some other chat service) is the for real-time chat. But it only keeps a couple weeks of history; there aren't that many channels; it's all chat with no threads; etc. It's meant to mirror in person chat, and nothing else. That way, you don't rely on it for making big decisions or keeping historical records; it's just an ephemeral way to coordinate, talk through something in real time, or create somewhat of a social dynamic.
"Second, unlike email, Slack doesn’t have any mechanisms for triaging work ... With Slack, the only two options are read and unread"
you never used the Slack Bot? you can right-click on any message that might need review later, then ask Slack to remind you about it again. it saves these reminders and you can access them anytime. you mentioned how email allows managing workflows by taking additional actions. if you right-click any message in slack, there are additional actions available
Reminders create another thing for me to be notified about later though. For me, I want to file things in different folders that I can go to when I have time, not have an alarm clock that I have to constantly snooze until it happens to go off at a convenient time.
I want a pull model, basically, and Slack is nearly all push.
I also want a pull model and I hate how most of the marketing for such communications tool uses the word collaboration
Erm no it’s not necessarily collaboration
Figma can count even though I am not a designer
And in most work places they heavily rely on what I call ad hoc communications for coordination purposes
If it’s coordination for one off tasks, again fine
But if it’s repetitive task with stable data models, then it should be structured and automated as much as possible to free up thinking bandwidth
The culture/leadership matters
Leadership needs to have an opinion abt what tools and how to use them