This wsek I’ve pulled the sharpest, most practical material from four recent conversations and stitched it into one read around a single question: what exactly is the human edge in the AI era, and how do you build it? The throughline is striking - four leaders from wildly different worlds (a hypergrowth AI startup, a design-led SaaS, a public think-tank, and a coaching practice) describing the same shift in almost the same words.

💡 Top tip - Read the TL;DR below, then jump to whichever facet is your current weak spot. Each section stands alone, so treat this as a menu rather than a march.

Featured in this round-up:

Jessica Fain (product leader, Webflow; ex-Box, Slack, Brightwheel; former chief of staff to two Slack CPOs), "The art of influence: the single most important skill left that AI can't replace" - Lenny's Podcast
🎥 https://www.youtube.com/watch?v=RP4vJeIb7WU 📆 22-03-2026

Katie Burke (COO, Harvey; former CPO then COO, HubSpot), "A people leader's job isn't to make you happy" - First Round
🎥 https://www.youtube.com/watch?v=bcLZE2c6d90 📆 10-04-2026

Dave Martin (fractional CPO; author, The Product Momentum Gap), "How to lead when you don't fit in" - The Product Experience
🎥 https://www.youtube.com/watch?v=jcccW-wfL9Y 📆 03-06-2026

Ayushi Roy (Chief Program Officer, New America; lecturer, Harvard Kennedy School), "Why the best public-sector PMs delete systems instead of building them" - The Product Experience
🎥 https://www.youtube.com/watch?v=64DNky8_A7g 📆 13-05-2026

🕒 Estimated Reading Time: 14 mins. Time saved: ~4 hours of listening! 🔥

Jessica Fain has spent her career as a product leader at Box, Slack, Brightwheel and now Webflow, with a formative stint as chief of staff to two Slack CPOs (April Underwood and Tamar Yehoshua). Her thesis on Lenny's Podcast is unambiguous: "Influence might be the single highest leveraged skill for product leaders outside of AI."

She knows the resistance it triggers. Plenty of PMs would rather just do brilliant work and let it speak for itself - and then hit a wall when less talented colleagues get promoted. Fain has lived it. Early at Slack, some of her ideas won funding and excitement, and some "died on the vine," and she couldn't understand why. The answer came when she moved into the chief of staff role and watched decisions get made up close: "People completely misunderstand how executives make decisions, what is going on in the heads, in the calendars, in the incentive structures of executives."

The objection she hears most is that influence feels manipulative - politicking, ego, games. She rejects the framing entirely. "When people frame this as politics, they're completely missing the point. Politics is manipulating outcomes and people for your own gain. Influence is about increasing the odds that your good ideas survive." Done well, it isn't a performance; it's learning. Treat stakeholder conversations as discovery interviews that strengthen your idea, and you both build a better product and earn the exec's respect.

She is equally firm on ownership. Quoting her first PM manager, Annie Pearl: "It's not my fault, but it is my problem." If a leader doesn't buy your idea, that is an influence failure to be solved, not a them problem to be lamented. And if you genuinely don't respect the person you're trying to convince? "If you don't believe they know something you don't, you need to quit. You need to go work somewhere else." Respect the expertise in the room, or remove yourself from it.

Key Takeaways:

Fain's most useful reframe is that the very skills that make you a great PM evaporate the moment you walk into a leadership meeting. "As product managers, one of our best sets of skills is curiosity and empathy. But the moment we're talking to an executive, we forget those skills" - and start centring our idea, our approval, our next milestone.

Her mental model for an exec's day is vivid: a calendar that's "a strobe light going off." You wake at 8am with a huge list of urgent fires, then move from a finance budget meeting to an executive interview to a people problem to a legal issue to your product review. The PM has been prepping for three weeks; the exec "hasn't thought about you since" and "may not have gone to the bathroom today." The practical fix is to spend 30 to 60 seconds at the top of the meeting setting context: why are we here, where did we leave it last time, what do we need today - then stop talking, because "the moment you go off over 60 seconds, you've lost them." Her bonus line: "Was there anything else you were hoping to cover today?"

Her single best tactical phrase for handling a strong or wrong-seeming opinion: "That's so interesting. What led you to believe that?" It's genuinely curious, it surfaces the board pressure or past meeting driving the view, and it co-creates rather than confronts. Execs, she notes, "are so good at seeming certain" because their job demands fast decisions on thin information - so your job is to help them unpack the why, then bring your own domain expertise to bear.

She's also sharp on the difference between an idea and a passing comment. When an exec floats "I wonder if we could do this study," don't drop everything - ask "do you think that's more important than the three projects we're working on now?" Half the time it's a mandate; half the time it's "just a random idea, put it on the backlog." Execs move too fast to flag urgency themselves, so you have to ask for it.

Key Takeaways:

A recurring mistake Fain sees is PMs trying to prove how hard they worked. She did it herself once - "we talked to 16 participants from these 15 geos and this is statistically significant" - and watched the exec she most needed glaze over and reach for their phone. "Put it in the appendix," she says. The baseline expectation is that you did the work; you don't need to perform it.

The counter-move is showing your options. Bringing a single recommendation often reads as if you missed something. She describes a product review with her CPO that went badly precisely because they'd brought only one approach; the feedback was "I don't understand how you're thinking about the permutations." They turned around a new doc in two days laying out all the options they'd considered and why each would or wouldn't serve the outcome - and the same recommendation suddenly landed. At Slack they formalised this as "Stuart plus two more": do exactly what the leader asked for, plus two other versions worth debating.

She also stresses fitting the format to the human. Some execs want a doc and ten minutes of silent reading; some want only the data; some hate a PowerPoint. "Execs, if you're listening to this, it's really helpful if you tell people." And she's a fan of leading with the recommendation first, then the evidence - the Barbara Minto pyramid - while adapting to each leader's style.

Key Takeaways:

Dave Martin, a fractional CPO who writes openly about being dyslexic, ADHD and most likely autistic, frames the same problem through neurodivergent leadership - but insists the tools help everyone, since "around 45 to 50% of people in tech are neurodivergent." His honesty is disarming: for much of a two-decade, multiple-exit career, he says, he was "faking thinking normal" - or, more precisely, masking.

He names the trap that kills influence: "going into detail, leading with detail, not translating into the exec language, not defining the signal you're trying to get across." You walk into the boardroom, get asked a question, and respond with edge cases to prove you're a safe pair of hands - and instead "you just lose the person in the exec room." You get labelled "not strategic" with shocking speed, and it takes enormous work to shed that label. The other half of the trap is emotional: you read the room's clarifying questions as challenges to your authority, get defensive, and spiral.

His fix is "signal prep" - three questions before any meeting that matters: What do I need from this room? What's my one-line recommendation? What will they repeat when I'm not in the room? He calls that third one "the retail" - the phrase that gets echoed in rooms you're not in. One leader he coached compressed an entire funding-and-restructure ask into "minutes not months for our customers," said it twice near the end, and that's the line that got repeated and approved.

The wrapper is CALM: Clarity, Authority, Learning, Momentum. Authority, he argues, comes from "asking for support and guidance rather than approval." Learning means visible, strong-opinions-loosely-held vulnerability, and creating a safe environment for your team to fail. Momentum is really about well-being and sustainable pace - he's blunt that burnout is "a leadership fault, because we set the expectation." He coached a head of product, "Mia," from being told she "wasn't strategic enough" to leading international growth in two to three months - not by teaching her strategy (which she had) but by fixing the signal drift, the imposter spiral, and the prep discipline. And his take on AI is the round-up's tidiest line: "It's like spell-check, but for influence" - a tool for the communication, never the judgement.

Key Takeaways:

Katie Burke went from CPO to COO at HubSpot and is now COO at Harvey. Her opening shot is a rejection of HR passivity: "Don't ask for a seat at the table - build the table and set the menu." Coming from a marketing background, she treats internal comms and employer brand as strategic differentiators ("people have a choice with how they spend their time"), and insists people programming "has to ladder up to the business strategy first and foremost."

The hardest part of the job, she says, is being the executive team's HR partner - able to say "let me help with your org" one moment and "the way you showed up in that meeting isn't what we expect of each other" the next. The currency is trust, and the failure mode is the "political monster" playing chess behind the scenes. Her explicit rule: "I will not share what you personally shared with me. I will share thematic feedback." Lose that discretion once and "you're never going to come to me again."

She is unsentimental about hard calls and allergic to false harmony: "I hate when people say we're a family." She prefers the high-performance sports team - and reaches for Michael Jordan, "a better teammate than he was a friend," difficult in practice but trusted because he worked harder than anyone and gave up the ball when the stakes were highest. On exits, she argues real transparency means the person hears the hard truth directly from you first, ideally one-to-one, while protecting information that isn't yours to broadcast: "Not every exit has to be a fiery exit... part of your job as a leader is to create as many graceful exits as possible."

She's also candid about the toll. At any moment "roughly 20% of people are unhappy with you." Layoffs carry a "hangover" of two to two-and-a-half years - shock, then survivor guilt, then lagging fear of the next one - so they're "a cultural decision," not a one-time business decision. And on headcount, her advice is to define the work before the heads, keep incentive structures simple ("simplicity wins"), and remember most managers aren't out for themselves - they're lobbying for their teams, which is its own bias to manage.

Key Takeaways:

Burke's sharpest cultural point cuts against a decade of "employee happiness" thinking. Her VP of talent's line: "The resort has to match the brochure." There is no globally correct culture - only an honest one, where what you advertise matches what people experience.

She defends Amazon's culture precisely because it's clear about what it is: "I have a much bigger problem with people who advertise it's all unicorns and rainbows, and when you come in it's cutthroat." Harvey, she says, is "intense but reasonable" and makes no apology for it, because the interview process matches the day-to-day. She also thinks the late-2010s went too far toward coddling - "the Disneyland approach to employee experience is gone" - but warns against jerking the wheel the other way into perks-stripping performative toughness.

Then there's "Berrygate," now legendary at HubSpot. Free berries got replaced by a "smoothie bar" the exec team thought was an elegant win-win; a Slack channel erupted (complete with arguments about the fibre content of whole berries) over not being consulted. Burke's response: "I think we've completely lost the plot... if you have spent the last 45 minutes complaining about Berrygate, you clearly don't have enough work to do." Her framework, borrowed from a former HubSpot CEO, is "protein versus sugar": pay real attention to substantive complaints that block winning or clash with your values; let the sugar go. And she watches "frequent flyers" - the people complaining 75% of the time and fixing 25% - more than any single gripe, including the new hire who ends week one with "18 observations on things that could have been better." That, in her experience, is a leading indicator of someone who won't last.

Key Takeaways:

Ayushi Roy, Chief Program Officer at New America and a Harvard Kennedy School lecturer, brings the most counter-intuitive insight of the four. In private-sector product "we're often trained that building is the answer. In the public sector context, unbuilding is the answer." Her career began monitoring aid for trafficked women and children during the Syrian civil war, where she watched "decisions of enormous consequence being made with really poor data" - which turned her from a policy person into a delivery person.

The signature story: asked to build a new system for the Children's Health Insurance Program, her team's discovery found 14 existing systems doing the same job - duplicated vendor contracts, ballooning budgets, and quarterly reporting burden on every state. So she went back with a proposal that made executives look at her "like I had horns on my head": "What if I removed 10 systems instead of building you one?" That became a favourite interview question: "What's something you built that you wish you never had?"

She also draws a line between a design problem and an engineering problem - "technology can rarely solve an upstream policy problem. Implementation does not make up for poor policy design." A laptop hinge tested to open 5,000 times is an engineering problem; an outcome with no known approach is a design problem, and pretending the latter is the former forces teams into "compliance-driven buildout as opposed to customer-focused buildout" - a government feature factory. Her IRS Direct File example shows the alternative done right: thin-slice by tax-scenario complexity (not income), roll out internally to IRS staff first to build sponsorship, and make failure survivable even under congressional scrutiny.

The deeper lesson for any leader is that you're managing an entire service, not shipping a product. "The product part of your job is often the easiest part of your job." The hard part is the buy-in that allows the product, the legacy systems, the legislation, the budgeting, the media and the oversight. She illustrates with childcare vouchers in Illinois, where modernising the application without fixing the no-waitlist policy "just feeds applicants into the top of a broken funnel." Build less, understand more.

Key Takeaways:

Pull the four conversations together and a practical playbook emerges for the AI era:

Reframe influence as a craft, not a compromise. Bring customer-grade curiosity to leaders. Ask "what led you to believe that?", anchor to their incentives, lead with the recommendation, and own the outcome when ideas die. As Fain puts it, the job is increasing the odds your good ideas survive - that's not politics, that's product leadership.

Prep your signal. Before the meeting that matters, answer Martin's three questions: what you need, your one-line recommendation, and the line they'll repeat when you've left. Lead the room rather than reacting to it, and treat clarifying questions as interest, not attack.

Make the honest call. Burke's bar: tell people the hard truth directly and first, keep confidences religiously, separate protein from sugar, and make sure your culture's brochure matches the resort. The 20% who are unhappy at any moment are the price of leading at scale - listen for the signal in the noise without being run by it.

Have the courage to delete. Roy's question - "what would you remove?" - is the senior move. Restraint and judgement read as leadership; reflexive building reads as a feature factory. And remember the product is the easy part; the service, the buy-in and the constraints are the real work.

The common thread is the one every guest landed on independently. AI is a spell-checker for influence, a prototyping accelerant, a draft-generator. It dramatically lowers the cost of communicating and producing. It does not lower the cost of judgement, trust, restraint, or the courage to make an unpopular call. That gap is the human edge - and it's where the next decade of product and leadership careers will be won.

That’s a wrap.

As always, the journey doesn't end here!

Please share and let us know what you liked or want changing! 🚀👋

Alastair 🍽️.