Disclosure: As of 4/1, I’m long Slack, unless I bought the wrong symbol, like ZM / ZOOM. 2020 is shaping up to be a hostile information environment, bring your salt shaker.

The ‘stuck at home’ economy is gangbusters. VR headsets are selling out, video chat esp Zoom is under heavy use (though security / privacy news is hurting them), Peloton’s stock is up presumably on sales numbers, and remember Blue Apron? They’re back baby.

Microsoft (I’m long as of 4/1) announced an 8x boost in some usage metrics for Azure, then amended that to 8x boost in Teams usage in some towns in Italy, which is a very different statement and I hope the SEC puts the fear of god into whoever wrote this post. Screenshot, yuck:

msft redaction

Also you can get mixed drinks to go in NYC.

At Slack usage and onboarding is setting records and the ‘shift from inboxes to channels’ has been ‘accelerated by 18 months’.

(Matt Ball ex Amazon says this accelerates and exaggerates all underlying trends (in the video space, but cool statement nonetheless) and I’m looking out for the word ‘accelerate’ everywhere now. I think this line is a neat way to think about change.)

The Slack CEO describes their future as a ‘cone of possibilities’. I won’t blame you if this reminds you of an off-brand childcare facility in the parking lot of a Denny’s. But it’s also not a bad way to think when you’re a crisis leader. He claims the middle path is less likely than the edges, a claim which is if nothing else a good thought exercise for strategic planning.

It’s good they’re focused on the challenges ahead despite some positive press. This year will be light on the picnic and heavy on the pickle, for everyone.

Core product problems require AI-light to solve

Slack is doing great now, but they’re asking themselves the same questions as every company seeing a traffic boost from the lockdown. How do we:

  • Convert new free users to paid users (guessing Slack is ok here)
  • Beg new users to stay past lockdown
  • Hope the bump creates continuing post-lockdown growth from habit change
  • Pray that increased energy in my ecosystem doesn’t lead to the evolution of a predator that either takes my food or I am its food

The Slack CEO’s tweet thread claims they’re not a WFH tool, they’re an ‘organizational agility’ tool. I think that speaks to their desire to maintain value post-lockdown and market the product as a communications tool that helps teams work.

This is total BS.

It ‘enables’ teams to communicate in that it transmits their words, but it’s so disorganized that the output is often unusable and requires multiple passes to answer questions. Slack actively hinders all non-communication forms of work by distracting the shit out of people and frying their brains worse than Broxholm.

(Broxholm’s brain fryer was in fact beneficial, if parasitic. Slack is probably neither).

Teams that use Slack experience serious pain points (described below) and all of these can’t be solved by refactoring UX or collecting extra user input; products that try to get users to self-categorize generally suck. Fixing this will require a total pivot to topic modeling, an ‘AI light’ technology that isn’t trivial to implement and may not be in Slack’s DNA. Read on.

Pain points


Tick-tick-tick. Turn your volume down. Getting these alerts is so annoying and so is missing them.

Slack will always be a hybrid of dangerous distraction and unreliable channel for getting ahold of someone until they develop the ability to humanely target these alerts. Not everything is important right now. Not everything needs to alert everyone right now, but most things need to alert someone eventually (if not handled).

The system needs to identify the topic under discussion and loop in the people who can help and are affected by it. This isn’t an easy problem. It doesn’t have to be perfect – users can do some of the heavy lifting when it makes a mistake.

Die-hards will argue that a channel is a topic. Or it’s a team. Or I don’t know.

In reality these topics are fine-grained and change over time. Nobody is up to the task of updating memberships, and people talk about things in the wrong channel, and it needs to be automatic.

Various competitors (discourse, zulip, quill most recently) are presenting threading as the solution. They may be right, but the two wildcards are (1) why hasn’t it caught on? are threads too heavy / inorganic somehow? and (2) do users know who they should be tagging? Plus if every thread requires you to tag N people congratulations, you’ve invented google groups.

Rollup / summarization

If you want to get anything out of Slack’s chat history, you have two options: scroll and search. Neither is ideal.

Managers in particular have a hard task of keeping abreast of what happened today and stepping in where needed. My vote here is to empower people to bring you problems rather than participating in low-level conversations, but I’ve had the ‘slack is hard to keep track of’ conversation with enough leaders to know that this is at least an ask.

Ideally you’d be able to ‘zoom out’ and see the N things that happened today, and their resolutions, rather than every 👏 single 👏 message 👏.

Stack Overflow’s innovation over vanilla forums is to mark the answer and push it to the top. It suppresses discussion, possibly a bad thing for in-house organizational problems, but great for troll-ridden online spaces.

(It’s also not totally their innovation. Remember expertsexchange, the awkwardly misreadable domain for ‘become a member to see the answer’? They’re one of many victims of the law of ‘you can’t charge for UGC’.)

To do rollup, Slack needs to ‘mark the outcome’ and also ‘find the question’ – i.e. they need to set boundaries in the stream-of-consciousness channel that mark, more or less, a topic. Not every conversation in a channel will have an outcome, but this feature adds more value for disjoint unsolved problems that happen over many weeks.

Once again, this doesn’t have to be perfect; it needs to be editable and learn from its mistakes.

But bracketing / summarization makes the information reusable to time travelers from the future, as well as managers doing EOD doomscrolling to see what their team got up to.


Stuff gets forgotten. Examples:

  • ‘Someone else is already working on this’ (or has a reason they dropped it)
  • ‘This didn’t get done’
  • And my favorite, ‘this question has already been answered’

(The latter, despite sounding simple, is the most like a group mind. I imagine a class full of kids all trying to solve the same problem in parallel and posting their questions and answers, and whenever a kid posts a question that another kid has answered, they get access to the answer).

If Slack ever wants to be a series of tubes for information rather than a dumptruck, automatically calling out past successes and failures is part of the recipe. Yes, sometimes forgetting is good, but not when the information in question is summarized, labeled with an outcome, and immediately relevant to what I’m trying to do.

And this requires (say it with me).

Having AI in your DNA isn’t just a buzzword anymore

And Slack doesn’t. Time to get some new DNA?

Facebook did this strategically with their Instagram purchase, is engaging in a long process of remodeling Instagram to serve as the vessel of their consciousness, successfully survived the first challenge of Snapchat Stories, and alienated and / or pushed out the insta founders.

It’s not surprising they had to go – the skull cavity must be emptied and cleaned before MZ’s brain can be transfered inside. But it’s not clear yet whether insta’s blood will invigorate its aging host or whether FB’s toxic bile will infect and ruin Instagram, driving its hip users to the new hotness.

Repeating myself from above – all of Slack’s pain points need topic modeling to improve them. Or heavy, non-delightful UX, I suppose, but that would put Slack’s userbase in the same position as doctors, who despise filling out forms to the point that it’s burning them out of their careers.

I’m sure Slack has a bunch of text-AI companies on their list to acquire. But as with all product companies, change is made harder by the fact that for existing users, your change is probably not for the better. The hardest thing in product is adding features that impose short-term costs on your users but are critical to your long-term survival / keeping up with the joneses.

Slack, like most of SaaS, both is and isn’t a tech company. Another way to say that: they’re currently more like Netflix than Blockbuster, but if they don’t adapt the tables will be turned and they’ll be looking at how to shut down their last 2 stores in Alaska.

For every semi-successfull SaaS company, there’s an AI-enabled version of that company that solves some core problem slight better or cheaper. It might be onboarding, it might be employee footprint, it might be fraud / content moderation, it might be curation. These are labor-intensive information tasks that scale linearly with customer base, and in some cases cause a reduction in quality as your userbase increases.

The consensus path for Slack is to turn their integrations ecosystem into an app store, and to continue adding features that attract enterprise clients, without altering the basic platform or UX.

This approach is more in line with their DNA, and not wrong – one answer for Blockbuster Video facing the rise of Netflix is to not compete with Netflix, eschew innovation, maximize margins, and shut down locations as they become unprofitable. Maybe to buy Netflix in the short term and shutter it, preserving their high-margin business, but in the long term accepting the demise of your model and investing in what’s next.

Change is risky and hard for everyone. I shall be watching with interest.