Climbing the Next Hill: How Slack's Product Chief Keeps a Mature Product Innovating

Host Lenny Rachitsky recalls building a startup, Localmind, on Foursquare's API before it was acquired by Airbnb. Weiss, who was head of product at Foursquare, says he has more scar tissue from that 2010 to 2015 period than from anywhere else. He argues you often learn more from things that do not fully work out, because the negative signal creates a feedback loop.

His postmortem is that Foursquare was caught in the "death star" of Instagram's ascent. The team had been close to the Instagram founders, who built on Foursquare's API early on. In hindsight, Weiss says, Foursquare was mistaken to treat the atomic unit as a person talking about a place they are at, rather than a person sharing a moment, which might have a place attached.

On local discovery, Google eventually replicated much of the functionality with its billion-plus Maps users. Weiss adds that a lot of this was luck and timing. Asked about consumer companies pivoting to B2B, he says paying customers give much faster feedback than a consumer business that hopes to reach enough scale to add ads, so B2B makes it easier to build an incremental, successful business.

Weiss has worked with strongly product-minded founders including Dennis Crowley and Stewart Butterfield. He says that if the CEO is an expert in your function, you will learn the most but also feel the least agency. His first lesson is to align on principles for what great product means at that company, so that the founder's intuition becomes shared language. Without it, he says, reviews feel Goldilocks-like: "not quite right" with no vocabulary to explain why.

His second lesson concerns timing. Weiss describes a U-curve of founder involvement over time: heavy early engagement to agree on strategy, goals and anti-goals, space for the team to explore, and strong involvement again at the end to check the quality bar. If the CEO skips that last step, he says, launch is much more painful. At Slack, Butterfield, engineers, designers, PMs and researchers would hold a joint bug bash on real code rather than static mocks.

Slack's four principles unpack its mission of making working life simpler, more pleasant and more productive. They include "be a great host", about saving people steps, and "don't make me think", a nod to the usability book, as a counter to the owner's delusion that users care as much as you do. Another is that "more clicks can often be okay", since in non-transactional software extra steps can build confidence and trust.

Weiss traces his AI experience to work on what became Google's knowledge graph about 15 years earlier, extracting facts from the web with natural language processing. What has changed in the last 6 to 12 months, he says, is that computers can now generate natural language that feels human, which opens massive creative applications.

He names a few durable principles. The promise of the UI must match the quality of the underlying data, and he sees overconfident, hallucinating LLMs as a failing that erodes trust, so products should acknowledge uncertainty and be transparent about sources. Product designs should also create virtuous cycles where natural use generates training data, as Netflix's ratings did.

On Slack's own plans, Weiss says there are many internal prototypes for asking questions of a company's accumulated conversations. He notes that a 2014 product-vision deck jokingly had "do magic AI stuff on top" as its fourth step. Organizationally, Slack has a central machine learning and search team plus several prototyping teams working on different customer problems on shared infrastructure. He expects AI eventually to be part of every product area's roadmap, as mobile is.

Complaint storms began around late 2019 or early 2020. Weiss explains that Slack employees are power users who live in the product all day, making it hard to empathize with new customers. A team, often including Butterfield, walks through a customer journey of another product on one shared screen, from landing on the website to first value, noting everything confusing or off-putting.

Starting with someone else's product generates inspiration and makes it easier to turn the same critical eye on your own. Rachitsky compares it to Stripe's friction logging. Weiss adds that it also calibrates the team on pace and quality, and he suggests it for teams that are stuck.

Roadmaps are treated as a diversified portfolio: new capabilities versus improvements, risky bets versus known ones. To make room for small improvements, teams run two-week customer love sprints, like a hackathon with a burndown list of low-effort, high-impact changes that all ship. Weiss says user-facing teams do this at least quarterly. He also credits Slack's roots, its founders came from the game Glitch, and the customer experience team, whose insights flow to product teams, and many staff did customer-support shifts.

On competition, Weiss says Slack's maxim is to be customer obsessed but competitor aware, in contrast to what he describes as Uber's competitor-obsessed approach. Slack did not set out to kill HipChat. It came with a mobile experience that synced across clients, search that worked, and consumer messaging features like emoji and reactions, and if you are ten times better on a couple of axes, behavior changes.

He sees Discord as more consumer and community focused, while Slack serves groups of people working together, a different audience. Microsoft Teams, he says, began as a defensive move to protect Office and has evolved toward video conferencing. Many customers use both, and Slack positions itself as connective tissue across tools with an open platform.

To keep zero-to-one innovation alive, Weiss says, culture matters, including the shared belief that Slack is still a small startup. One principle is literally "take bigger boulder bets", meant to counter constant incrementalism in which a team's life is measured by a KPI rising 1% a quarter. The metaphor is of being in a valley and seeing only the next hill, not the range behind it.

Structurally, Slack has incubated new teams from scratch in areas before they mature. Huddles and clips came from this in the pandemic, as customers who lived in Slack felt disconnected from teammates. For AI, the approach is to ask what customers wish Slack could do with new superpowers, then prototype with a couple of teams.

Such teams get a "get out of jail free card" on the normal planning and quarterly reviews, so that the pace of learning is what matters. The formula is to prototype quickly, learn from users, pilot publicly and then launch something that blows people away. Among rituals, Weiss mentions reading real tweets about the product at all-hands, and live Slack threads where PMs, engineers and designers react to usability sessions, which later become the research report.

From 2014 to 2017 Slack was almost entirely self-service and growing fast, and then the company focused on enterprise. Around 2019 Weiss saw that the fundamentals of self-service were weaker than they had been, as the next wave of customers had different needs and levels of sophistication. His team did user research and studied cohort curves, then threw the roadmap away.

They came up with hypotheses about new levers and told leadership they would likely drive no impact for six months, focusing only on learning. The result, Weiss says, was a doubling of the rate of new paying-customer growth in the year and a couple of years after. Key levers were comprehension and desirability for ordinary workers, and a trial of the paid product, as Slack had only a free tier and a paid tier.

The other lever was a new north-star metric, "successful teams": five people using Slack to communicate during the work week, which made a team 400% more likely to upgrade within six months. Rachitsky highlights the importance of an activation metric that predicts retention. Weiss describes product-market fit as stacked S-curves, each new audience needing its own fit. He puts the work at about 60% journey and positioning and 40% product features such as WYSIWYG message composition and offline mobile.

Asked what Slack did right in product-led growth, Weiss says the term did not exist when he joined. The core was a product customers loved enough to put their social capital on the line to bring in colleagues, with teams of 5 to 50. Teams inside large companies adopted it independently, and when the enterprise sales team was built, its first exercise was to sort customer domains by active users and call the most active ones.

Rachitsky ties this to Seth Godin's idea that winning products are the ones you tell your friends about. Weiss agrees that with marketing channels saturated, it is hard to scale a large business without some word of mouth and customer love.

Slack also built a pilot program with thousands of customers who have signed agreements to receive new functionality progressively, which is necessary because Slack is multiplayer. By launch, a feature has been proven with customers across many countries and industries rather than only internally. He recommends picking pilot participants with diversity and a higher tolerance for pain who will accept a feature being killed before it ships.

Weiss describes his blog post on the 10 traits of great product managers as a positive counterpart to an earlier piece on what PMs wrongly expect from the job. The first trait is living in the future and working backwards: carving out time for the six-month to two-year horizon. Rachitsky connects it to Amazon's working backwards, Brian Chesky's push at Airbnb and Paul Graham's advice to live in the future.

The second is amplifying the team and facilitating energy and momentum, not creating everything yourself. Weiss calls the "PMs as mini-CEOs" advice the most dangerous in the history of product management because it produces dictators instead of facilitators. A related trait is facilitating the pace and quality of decisions without playing tiebreaker all the time.

The next traits are impeccable execution and a focus on impact. He says he has never seen a successful PM who was disorganized, and that "impact solves all PM issues". He recalls Ben Horowitz at Foursquare saying leaders should do what they said they would do, or follow up and explain why. Rachitsky adds that being the shock absorber with an "I've got this" aura earns responsibility.

Weiss says that, especially at senior levels, writing is the only scalable way to influence a larger product organization. He recommends Stephen King's On Writing, while Rachitsky adds On Writing Well and recommends cutting the first paragraph. Both emphasize that writing scales influence when your days are full of meetings.

Another trait is optimizing for the pace of learning, taking bolder bets and accepting short-term wrongness for long-term levers. Weiss also describes data fluency, which he says is not about statistics alone: quantitative data, surveys or a hundred customer conversations all count as inputs. He pairs it with product taste, which he defines as an intuition for what people will love before testing, and recommends the book Creative Selection about early iPhone teams.

For early-career PMs, he would prioritize execution, a bias for impact even if local, and data and research fluency. For directors, the priorities shift toward pace and quality of decisions across teams, living in the future with long-term strategy, and writing as the highest-leverage use of time.

In the lightning round, Weiss recommends The Innovator's Dilemma for product strategy, Radical Candor for leadership and soft influence, and Doris Kearns Goodwin's Leadership in Turbulent Times. He names Succession and The Bear as recent favorites. His favorite interview question asks candidates what unfair secrets they have learned to improve the velocity and energy of a product team.

The biggest process change he has made at Slack was spending fewer cycles on static mocks and walkthroughs and prototyping paths in real software, even messy and disposable ones, because you have to "live and touch and smell" the software. He says this avoided months of design debate by quickly showing how it feels.

His two Slack pro tips are to customize the sidebar into sections with settings such as unread-only or sort order, and to use the quick switcher with Command-K to jump between channels, people and files. He closes by inviting listeners to send Slack feature requests, especially about getting the rest of their company to join.