Lore

startup scaling

How to be a CEO when AI breaks all the old playbooks | Sequoia CEO Coach Brian Halligan

Brian Halligan — HubSpot co-founder, former CEO, and now Sequoia's in-house CEO coach — argues that starting a company has never been easier but scaling one into a durable, high-impact organization has never been harder, and that surviving as a founder-CEO requires learnable but unnatural skills (rigorous hiring and firing, blunt crisis management, explicit incentive alignment, and letting go of control) that most CEOs underinvest in.

Lenny's Podcast · 2026-02-15 · English

Key ideas

  1. Company formation is exploding while competition and noise make scaling harder than ever, so CEOs face growing pressure to be faster, better decision-makers; Halligan frames the founder-CEO role itself as uniquely unrescuable — 'no one is coming to save you' — and far more demanding than a 996 schedule.

  2. Most people, including experienced CEOs, systematically overrate their own gut-feel interviewing ability and underrate rigorous blind reference-checking, which Halligan says explains why C-level hires fail at a high rate (his estimate: at least 50% gone within 18 months).

  3. Companies systematically overvalue impressive outside/pedigreed résumés (e.g., Salesforce, Google, Microsoft, McKinsey hires) relative to homegrown talent, producing preventable attrition, when a blended 'core plus a few veterans' approach performs better and homegrown candidates are chronically underrated in close calls.

  4. A rarer class of 'five-tool' founder-CEOs (who can code, sell, set vision, and inspire) is emerging, and Halligan notes that even iconic past CEOs like Jobs and Bezos wouldn't clear that bar, raising an open question of whether AI will let more 'mere mortals' build like five-tool CEOs or instead further reward the few who already combine all the skills.

  5. Enterprise sales/go-to-market will be among the last white-collar domains AI transforms because it depends on human trust, even as AI is already reshaping software development, support, and (starting to) legal — but Halligan predicts the funnel itself will invert, with AI avatars and agents replacing search-driven discovery while the underlying sales roles and processes stay largely unchanged.

  6. Faster technology cycles create a 'tax on optionality': CEOs who used to let irreversible ('one-way door') decisions sit unresolved for up to a year now must resolve them quickly, and planning cycles have compressed from roughly a year to roughly three months.

  7. Because most company failures are self-inflicted ('overeating,' not 'indigestion'), Halligan argues CEOs should treat crises as forcing functions — delivering hard truths all at once and deliberately overcorrecting rather than nibbling at problems incrementally.

  8. Diffused ownership fails as companies scale past roughly 100-150 employees, because departments silo and individuals naturally start optimizing for their own team's metrics over company- and customer-wide value, requiring leaders to explicitly re-align incentives through comp design, public reviews, and recognition rituals.

  9. A company's culture gravitates toward being customer-, employee-, or investor-centric, and Halligan says HubSpot's early employee-centric bias (e.g., #1 on Glassdoor) was a mistake he later corrected by deliberately over-rotating structure, comp, and meetings toward customers.

  10. As a company scales, the CEO's offhand remarks get misread as mandates and the founder can no longer personally vet or own everything, requiring explicit internal signaling systems and a shift from execution-heavy ('perspiration') to vision-heavy ('inspiration') work, while confronting a personal limit on how many people one is willing to fully trust with ownership.

  11. Halligan ties his own near-fatal accident directly to a structural leadership decision — reorienting around 'life's short' and stepping down as CEO — illustrating his broader claim that crisis, more than foresight, tends to catalyze needed personal and organizational change.

  12. Category-defining teams and products often come from deliberately combining mismatched, highly specialized ('spiky') people rather than uniformly credentialed hires, a lesson Halligan illustrates through the origin of the Grateful Dead and says startups generally underrate.

  13. LOCK algorithm — Halligan's four-factor mental checklist for assessing CEO potential: Lovable (inspires intense followership), Obsession (deep, longstanding fixation on the problem), Chip on the shoulder, and Knowledgeable (deep domain expertise). Apply: When meeting a founder, test each factor concretely — e.g., ask whether a 28-year-old you would 'crawl across broken glass' to work for this person (Lovable), and be skeptical of founders who only discovered the problem ~6 months before founding (Obsession).

  14. LOCKS (Student add-on) — An extended LOCK framework adding 'S' for Student — a founder who is a constant, deep learner who studies history and context rather than skimming broadly — used when evaluating companies for Sequoia investment. Apply: Favor founders who go deep and historical in how they learn about their domain (cited examples: Winston Weinberg of Harvey, James of Profound, Gabe of Rogo) over those who are only surface-level 'learn-it-alls.'

  15. Parker Conrad's NDA/board-deck interview technique — Before a C-level interview, have the candidate sign an NDA, send them the company's last board deck or memo, then hold a 30-minute discussion about it. Apply: Watch whether the candidate is overly complimentary (red flag — a yes-person) versus genuinely challenging the material, since you want an executive who will push back rather than flatter you.

  16. Blind-reference scaled questions — A structured reference-checking method using pointed, scaled questions rather than generic strengths/weaknesses, since CEOs overrate their own gut-feel interviewing and underrate quality references. Apply: Ask references 'Would you enthusiastically rehire this person?', '1-10, how likely are you to try to rehire them?', and 'Were they top 1% or top 10% of your employees?' to separate genuine enthusiasm from lukewarm politeness.

  17. Hire slow, fire fast — A hiring/firing principle Halligan says most people invert (hiring fast and firing slow), which he ties to poor C-level hiring outcomes. Apply: Take the time needed to thoroughly vet senior hires, but once a hire clearly isn't working — especially once a team's trust in them has eroded — remove them quickly rather than waiting.

  18. Spiky-candidate panel scoring + smaller panels — An interview technique HubSpot adopted after finding that unevenly-rated ('spiky') candidates outperformed uniformly 'safe' candidates, and that shrinking panels from 8 to about 4 interviewers also improved hiring hit rate. Apply: When reviewing panel feedback, favor candidates with clear standout strengths and visible weaknesses over uniformly-good 'safe' candidates, and keep interview panels to roughly 4 people.

  19. 2004 Red Sox team-building model — A team-composition analogy: build a core of inexpensive, homegrown, high-quality talent blended with a few paid veteran 'free agents,' rather than an all-veteran or all-homegrown roster. Apply: When staffing an organization, deliberately mix long-tenured internal promotions with a small number of experienced outside hires instead of defaulting to prestige outside hiring.

  20. Kids table / adults table peer groups — Halligan's CEO peer-coaching structure at Sequoia: roughly 15 CEOs of sub-100-employee companies ('kids table') and roughly 15 CEOs of 100+-employee companies ('adults table') who work through problems together on a 'misery loves company' premise. Apply: Group founders with peers at a similar company-size stage so their problems 'rhyme,' letting peer discussion do most of the problem-solving rather than the coach alone.

  21. Five-tool CEO framework — Borrowed from baseball's 'five-tool player' scouting concept, this rates a CEO on five capabilities — can code, has taste, has vision, can sell the product, can convince employees — with true five-tool CEOs (e.g., Brett Taylor) being extremely rare. Apply: Use it to diagnose which 'tools' a given founder is missing and where they'll need to hire complementary strength, rather than assuming one person must have all five.

  22. AEO (Answer Engine Optimization) — The anticipated successor to SEO — optimizing for discovery via AI answer engines (e.g., Gemini, Claude, ChatGPT) rather than Google search rankings. Apply: Redesign company websites and content strategy around being surfaced accurately inside AI assistants' answers rather than just search-engine rankings.

  23. One-way door decision-making — A framework distinguishing irreversible ('one-way door') strategic decisions, which Halligan says he historically let sit unresolved for up to a year, slowing the whole team. Apply: Identify which pending decisions are truly one-way doors and resolve them much faster than felt natural in the past, since today's pace of change makes delay far costlier.

  24. Don't nibble the sandwich — A principle, credited to Google CFO Ruth Porat, that bad news like layoffs should be delivered all at once rather than in incremental waves. Apply: When facing necessary bad news, 'rip the band-aid off' and tell everyone at once instead of doing repeated smaller rounds, trusting employees as adults who can handle it.

  25. Next play — A mindset borrowed from Duke basketball coach Mike Krzyzewski, where a team deliberately moves past an unforced error instead of compounding it through overcorrection. Apply: After a mistake or crisis (e.g., HubSpot's March 2019 outage), publicly acknowledge it, then explicitly signal the org to refocus forward rather than dwelling on it.

  26. Never waste a good crisis — The claim that most good changes at HubSpot came out of crises, because crises created the license to take drastic corrective measures that otherwise wouldn't happen. Apply: When a crisis hits, use it as the forcing function to overhaul the underlying process rather than just patching the immediate symptom.

  27. Two people watering a plant — A metaphor for why diffused or shared ownership fails — with two caretakers, a plant tends to end up over- or under-watered rather than well cared for. Apply: Avoid assigning shared, ambiguous ownership of an initiative to multiple people or a committee; assign one clear owner instead.

  28. DRI (Directly Responsible Individual) — A single-owner accountability model that every CEO 'at the adults table' becomes religious about once the company is siloed across departments. Apply: For any cross-functional initiative or metric, name one person with authority to direct other divisions and be held accountable for the outcome, rather than relying on a committee.

  29. Two steps forward, one step back (no silver bullet) — A corrective to the belief that any single hire, investor, event, or product release will be a 'silver bullet' — growth is really an uneven, crisis-punctuated pattern. Apply: Set expectations for yourself and the board that growth will be uneven rather than smooth, so setbacks aren't mistaken for failure.

  30. CV > EV > TV > ME — A prioritization hierarchy HubSpot put on the wall — Customer Value, then Enterprise Value, then Team Value, then individual value — to counter leaders who optimize for their own team's metrics at the company's expense. Apply: Build this ordering explicitly into performance reviews and public praise/criticism so leaders are rewarded for company- and customer-level outcomes, not just their own department's numbers.

  31. Segmented Employee/Customer NPS tracking — HubSpot's practice of running quarterly Net Promoter Score surveys for both customers and employees, broken down by department, to catch leaders quietly damaging morale or customer trust. Apply: Watch for a department's steady NPS suddenly dropping as an early-warning signal of a leader optimizing for their own team over the company, and act quickly since scores rarely recover once trust is lost.

  32. Public praise/criticism management style — An approach HubSpot used early on, loosely modeled on how Halligan describes Jensen Huang running Nvidia — no one-on-ones, with praise and criticism both given openly in large management-team meetings. Apply: Use shared management forums rather than private 1:1s to call out both good and bad examples of company-wide thinking so lessons propagate across the whole leadership team.

  33. Champagne Awards — A quarterly recognition ritual at HubSpot's company meetings — a signed bottle of Veuve Clicquot given to someone, usually for exemplifying 'solving for enterprise value.'. Apply: Create a recurring, visible ritual that publicly celebrates the specific behavior you want more of.

  34. Customer-centric / employee-centric / investor-centric orientation — A framework claiming every company has one dominant 'center of gravity,' and that HubSpot was, in its early years, more employee-centric than customer-centric. Apply: Diagnose which orientation your company actually has, and if it's misaligned with what the business needs, deliberately overcorrect culture, comp, and meeting structure toward the desired one.

  35. Customer panels in management/board meetings — A practice of bringing customers directly into HubSpot's management-team and board meetings so leadership and the board can ask them tricky questions directly. Apply: Regularly invite real customers into internal leadership meetings and ask pointed questions ('what do you love/hate about us?') to surface unfiltered feedback.

  36. Retention/NPS-based comp plan — A compensation redesign at HubSpot that paid the management team based on retention and NPS rather than revenue, used deliberately to pull the company's culture toward customer-centricity. Apply: When you want to shift organizational behavior, change what leadership is actually paid on, not just what's said in meetings, to match the desired priority.

  37. Message-tagging system (Dharmesh's intent tags) — A system Dharmesh Shah built to tag internal messages by intent level — do-this-week, let's-discuss, just-FYI, and a 'pleading, not ordering' tag — to stop offhand CEO comments from being misread as mandates. Apply: When communicating internally, explicitly label the urgency/authority level of a comment so casual remarks aren't escalated into full initiatives.

  38. Perspiration/inspiration ratio — A framework describing how the CEO's work mix should flip as a company scales — roughly 90% perspiration/10% inspiration at startup, inverting to roughly 90% inspiration/10% perspiration at scale-up. Apply: As your company grows, deliberately shift your own time from hands-on execution toward vision-setting and staying connected to customers, letting go of doing every job yourself.

  39. Trust surface — Halligan's term for the limited number of people a CEO is genuinely willing to fully delegate ('DRI') important initiatives to, which he calls a common personal scaling bottleneck. Apply: Deliberately work to expand the number of people you trust with real ownership, since a too-small trust surface frustrates capable people and limits how much the org can scale.

  40. Spiky teams (category creation) — A team-formation idea illustrated by the Grateful Dead's origin — combining specialists from wildly different backgrounds (bluegrass, country, avant-garde jazz, marching band) produced an entirely new musical category. Apply: When building a founding or product team, deliberately combine people with very different specialist backgrounds rather than hiring uniformly similar profiles.

  41. AI-clone / personal chatbot (Delphi-style) — A personal AI chatbot trained on all of a creator's public content plus private documents, used to answer audience questions and, in a newer feature, interview the user back. Apply: Build or use such a clone to scale one-to-many interactions (e.g., replacing office hours), differentiating it from generic chatbots by feeding it private/proprietary material.

  42. Barbell strategy media consumption — A media-diet habit Halligan attributes to Marc Andreessen — consuming only real-time content (X/Twitter) or books at least 10 years old, nothing in between. Apply: Deliberately skip most contemporary books/media and split consumption between current, fast-moving commentary and time-tested older works.

Insights

HubSpot's own hiring data reportedly ran counter to conventional wisdom: a candidate rated unevenly by an 8-person panel (some 4/4, some 2/4 — 'spiky') outperformed a candidate rated uniformly 3/4 by everyone, and shrinking panels from 8 interviewers down to about 4 also improved hiring outcomes.

Halligan claims McKinsey hires specifically tend to fail at founder-led startups not due to competence but worldview mismatch — most founders are skeptical of conventional wisdom, while people who go work at McKinsey tend toward conservatism.

By Halligan's own 'five-tool CEO' bar (code, taste, vision, selling, convincing employees), neither Steve Jobs nor Jeff Bezos would qualify, since neither personally wrote code — suggesting today's expectations for founder-CEOs have risen past history's most celebrated examples.

HubSpot's department-level NPS tracking reportedly showed that once a team's trust in a leader collapses (e.g., a steady 65-68 score dropping to 30), it tends to keep falling even after corrective feedback (30 down to -5) rather than recover — the empirical basis for Halligan's 'fire fast' stance.

Halligan says he made the literal decision to step down as HubSpot CEO 'at the bottom of the cliff' during a near-fatal snowmobile accident, directly tying a personal health crisis to a corporate succession event (handing the CEO role to Yamini Rangan).

Dharmesh Shah, Halligan's co-founder, reportedly never had a single direct report across HubSpot's entire history, having set that condition at founding and holding to it even as the company scaled to thousands of employees.

Halligan claims that inside HubSpot itself, AI is already working well in software development, customer support, and (starting to) legal — but has barely touched go-to-market, reinforcing his prediction that enterprise sales, built on human trust, will be one of the last functions automated.

Halligan claims Major League Baseball ownership (he is a Red Sox part-owner) is less profitable than commonly assumed because the league lacks a salary cap, citing roughly a 4x payroll gap between the Dodgers (~$400M) and Marlins (~$100M) as evidence of what he calls a 'broken model.'

«The thing about being a founder CEO is there's no one there to rescue you. Your parents aren't going to rescue you. Your VC is not going to rescue you.»

— 00:01

«Starting a company has never been easier. Scaling one into a durable, high impact organization has never been harder.»

— 00:08

«I look for four things. I call it my lock algorithm.»

— 00:49

«My other piece of advice, and no one listens to me on this, is hire slow and fire fast. People hire fast and fire slow.»

— 09:43

«Within 18 months after you hire a C-level exec, at least 50% of the time they're gone. There's this high mortality rate on them. It's harder than people think.»

— 09:58

«there are five tool CEOs like Brett Taylor's one. You can code, you have taste, you have vision, you can sell the product, you can convince employees like this kind of super CEO.»

— 22:23

«Honestly the only thing that's different today it's exactly the same as it's been for 100 years except they call their sees or their system consultants for deployed engineers.»

— 32:15

«there's a massive tax in optionality when you can move this fast and try a lot of things.»

— 36:38

«When you have to eat a sandwich, don't nibble.»

— 38:25

«Companies are far more likely to die of overeating than indigestion.»

— 41:47

«Never waste a good crisis.»

— 42:04

«If you want to kill a plant, have two people water it.»

— 43:52

«It's one person is responsible for a goal, a metric, some outcome you want.»

— 45:37

«EV is greater than TV is greater than me.»

— 50:43

«You don't work for your boss, you work for Apple.»

— 53:44

«I have trust issues. Like I only trusted a small number of people at HubSpot to be a DRI to really drive something important.»

— 62:11

«I don't really like being CEO of an 8,000 person company. Doesn't really suit me. Like my harmonic motion is off.»

— 70:59

Reception

Comments are overwhelmingly enthusiastic and grateful, with many praising specific insights and the guest, though a few skeptical or critical remarks appear.

This is an unusually technique-dense founder-coaching interview — Halligan packs in a large number of named, reusable frameworks (LOCK, DRI, CV>EV>TV>me, hire-slow-fire-fast, spiky teams) grounded in specific HubSpot data and anecdotes, making it a strong reference source, though its AI-adoption predictions (AEO, sales avatars, timelines for enterprise-sales automation) are more speculative and time-bound than the hiring and culture material.

74:37

↳ Lenny's Podcast · YouTube

Watch original