Julien Bek on Sequoia’s Investment Process, Founder Judgment and AI Markets

Harry Stebbings welcomes Julien Bek, a Sequoia Capital partner he has known for nearly a decade. Before discussing investment decisions, he asks about Bek’s relationship with his family and the responsibilities that shaped his life outside venture capital.

Bek says his father developed a neurological condition when Bek was in his early twenties. As an only child, he became involved in caring for him. He remembers his father growing up in rural France, leaving school young to support family, and later caring for his own parents.

Stebbings says ambition can be motivated by caring for other people, rather than only by personal success. Bek agrees that clear values make decisions easier to understand, even when living by those values requires difficult work or sacrifice.

Asked what surprised him after joining Sequoia from Accel, Bek describes the firm as a competitive sports team. He recounts arriving very early on his first day and meeting Doug Leone, who had already taken a call. Bek treats the story as an illustration of personal drive rather than a national stereotype.

Bek says the partnership looks for people who already have that competitive energy. He notes that Leone has European roots and says he believes the intensity is individual, not simply a difference between Americans and Europeans.

A common misconception, Bek says, is that Sequoia merely waits for founders to call. He describes partners as active hunters who source opportunities and are expected to perform individually while working as a team. The early-stage group, he says, was roughly the size of a football team.

He recounts how a Sequoia partner built a relationship with Ken Griffin over many years and says that persistence helped Sequoia invest in Citadel Securities. Bek says he was not present for that decision, so the story is his account rather than a firsthand reconstruction of negotiations.

Stebbings admires the willingness to pass on a company and later invest at a higher price. Bek calls this revisiting prior assumptions: when the environment changes, an investor should update the original judgment. He says Sequoia underestimated some aspects of AI’s potential in its early assessment.

The pair note how much larger some company outcomes have become than investors expected a decade earlier. Stebbings suggests a very high valuation can be a starting point for a future large return. Bek cautions that the investor must still choose the few companies that matter among a much larger field.

Stebbings says a company with substantial revenue can be less risky than a much earlier business despite a higher valuation. Bek agrees that evidence can reduce uncertainty, while noting that some expensive AI rounds occur before comparable operating data exists.

Asked about new AI labs, Bek says he has not personally invested in one and is skeptical of many companies pursuing a similar approach to established leaders. He stresses that this is his own view, not the entire partnership’s. He describes a different research approach as one reason an exceptional team might still be compelling.

On Series A investing, Bek says milestones differ by sector. A hardware or physical-AI company may need more time and capital to reach a working prototype, whereas a software investor may look for early recurring revenue. He expects funds to collaborate more when experimentation requires larger commitments.

Stebbings asks whether larger potential outcomes make him less focused on ownership. Bek says no: his concentrated style involves working closely with only a handful of founders. Because he can support a limited number of boards over a career, he wants meaningful exposure to each partnership.

Bek describes helping founders with customer introductions and senior hires while leaving operational decisions to them. He says that level of involvement cannot be repeated across hundreds of companies. Stebbings notes that his own portfolio approach is broader; the two acknowledge that different models can work.

Asked who is particularly good at finding companies, Bek praises a partner’s combination of competitiveness, technical understanding and ability to connect with founders at different stages. He describes sourcing as a skill distinct from deciding whether an opportunity merits investment.

For picking, Bek credits Luciana Lixandru, with whom he previously worked at Accel. He admires her ability to evaluate companies across consumer, software and other categories rather than relying on a single sector pattern. The examples reflect his assessment of a colleague’s work.

Stebbings asks whether a firm with a history of unusually large outcomes creates extra pressure on new partners. Bek says the job is supposed to be difficult and that partnership discussions can be fierce. He considers challenge useful when it improves the sponsor’s understanding of the risk.

Bek recounts an internal debate over SpaceX: he says one early vote was especially low, but the sponsoring partner, Shaun Maguire, continued to build conviction and brought colleagues to see the company. Bek says a smaller initial investment preceded a larger one. The voting and deal sequence are his account.

At partnership offsites, Bek says the team reviews past fund outcomes and sees that the strongest investments often involved especially high sponsor conviction. He does not claim conviction guarantees success; rather, he sees a repeated relationship between deep belief and the willingness to keep supporting a company.

Asked whether great investments always divide opinion, Bek says not all do. He recounts the Airbnb seed story as an example of an initially unconventional idea that required conviction, while acknowledging he was not present for the decision. He also stresses the importance of continuing to invest as a company develops.

Stebbings observes that some founders study the history of their field deeply. Bek agrees that understanding a market’s past can inform how an entrepreneur sees its future. They do not suggest that historical curiosity alone identifies a successful founder.

Stebbings challenges the claim that Sequoia simply pays more than everyone else. Bek says the firm tries to partner early and argues that other investors may subsequently pay a premium when they see progress. His description is his view of the firm’s behavior, not a verified comparison of every round.

The pair discuss rounds completed soon after an earlier investment. Bek says he has seen such arrangements only a handful of times and frames subsequent pricing as a supply-and-demand issue. He cautions against treating a few examples as a universal Sequoia tactic.

Bek says faster product development can compress the time between milestones and financing rounds. Stebbings asks whether the familiar pattern of repeated revenue doubling and tripling has lost relevance when some new AI companies appear to grow much faster.

Bek expects more conventional growth comparisons to return as new markets mature. He distinguishes a greenfield category, where customers did not previously have a solution, from a replacement market where a company must displace an existing system. The two types of growth should not automatically be compared as identical.

Stebbings argues that capital has an opportunity cost and that investors may prefer a much faster-growing new market. Bek responds that an investment’s ultimate return depends on ownership and the company’s value when liquidity arrives, which can be many years later. They disagree about how heavily to weight short-term momentum.

Stebbings points to a more active secondary market and says a smaller or more flexible investor may be able to sell sooner. Bek questions how easily that approach scales with a larger fund. Their exchange describes different strategies and does not establish that secondary liquidity will always be available.

Stebbings asks how a Sequoia deal is approved. Bek says the partnership has long held investment committee meetings and is now experimenting with more asynchronous feedback. A memo can be circulated for independent comments, while a partner can still call a live meeting to obtain the group’s views.

Bek distinguishes rapid discussion in a meeting from slower written thinking. He hopes combining both formats will improve decisions by letting partners contribute individually before reaching a collective view. He presents this as an experiment rather than a fixed rule for every investment.

Founders still pitch the partnership, Bek says. He tries not to overprepare them because a scripted presentation may conceal the qualities he saw in earlier meetings. Performance under pressure can be informative, although he acknowledges that a founder may behave differently in the committee room.

Bek describes independent feedback before discussion, followed by a vote after discussion. He says the sponsoring partner can still make a decision using the group’s input, but must have serious conviction, particularly when respected colleagues disagree. His description reflects the process as he understands it.

Knowing colleagues’ reasoning matters beyond the initial vote, Bek says, because a founder needs support from the whole partnership. He expects partners to make valuable introductions and help a portfolio company even if the investment was originally sponsored by someone else.

Stebbings asks whether candid disagreement becomes personal. Bek says discussions can be heated, but sees direct feedback as useful information for the sponsor. In his view, investors need the courage to take thoughtful risks rather than seeking a comfortable consensus.

When a presentation wins unusually easy agreement, Bek says the team may appoint a devil’s advocate. The group tries to write a premortem: what could go wrong, and what would it have missed? A polished story can be persuasive, so the exercise is intended to test the investment rather than punish the founder.

Bek says assessing a founder begins with understanding what makes the individual distinctive. He tries to be open about his own experience so that the meeting becomes a conversation rather than a transaction. He considers missed opportunities a major risk in an investor’s work.

He shares parts of his childhood, including time with separated parents and his mother’s illness and recovery, as examples of how personal stories can shape a trajectory. He says he asks because he is curious about people, not because a particular hardship guarantees entrepreneurial ability.

Stebbings asks how an investor can tell an authentic account from one tailored to what investors want to hear. Bek says he asks follow-up questions and watches for inconsistencies, while giving people the benefit of the doubt. Such signals prompt further diligence; body language alone does not prove dishonesty.

Bek recounts an unnamed situation in which he says other investors later told him a founder’s story was fraudulent. He describes questions that made him curious and a canceled meeting, but identifies no person and provides no independently verifiable evidence in the interview. The anecdote is his account, not a finding against an identified founder.

Asked about arrogance, Bek cites a framework that separates liking a founder from the possibility of investment returns. He says an unpleasant presentation does not automatically mean a founder lacks ability, but urges investors to distinguish a genuine strength from confidence used to obscure a weakness.

Stebbings shares a founder he believes he initially misread and says references became especially important to him. Bek agrees that references can provide context unavailable in a short pitch, but says they need to be interpreted carefully rather than treated as a simple vote.

Bek says he met Lovable founder Anton Osika before the company was founded and underestimated him. Looking back, he thinks the informal group lunch lacked a clear plan: he did not ask the questions needed to understand what made Osika distinctive. He treats the miss as a lesson in preparation.

The speakers discuss how presentation styles differ by person and culture. Bek recalls customer reference calls for a German company in which restrained numerical scores still came with positive explanations. His comments are heuristics from his experience, not a rule about how any nationality communicates.

Bek warns that rigid pattern matching on childhood or background can cause an investor to miss strong founders. Instead, he tries to understand a person’s trajectory and how much they have developed, especially when a young founder has only a short professional history.

He compares two people with similar educational credentials but very different starting circumstances to illustrate why a résumé alone may hide the distance traveled. He says the story can inform an assessment of past trajectory, while admitting it does not guarantee that either person will follow the same path in the future.

Asked for a lesson from Doug Leone, Bek describes asking a founder for their best reference and then their worst, including why. He says the contrast can reveal self-awareness and useful texture. He is looking for a clearer understanding of strengths and weaknesses, not an unrealistically perfect reference.

Stebbings notes that a difficult former employee may become a strong founder, which can make references hard to interpret. Bek agrees that context matters: the same behavior may have different implications in a large organization and in a company the person creates.

Bek describes a reference framework he attributes to Shaun Maguire using chess ratings as an analogy. His point is that people with exceptional experience in a domain may be better equipped to distinguish another unusual performer. He treats it as a way to select informed references, not a mathematical proof of talent.

Stebbings observes that some highly accomplished entrepreneurs give restrained praise. Bek suggests separating personality and communication style from accomplishments when interpreting the reference. A useful reference may require understanding what a particular person considers exceptional.

Bek attributes another framework to Pat Grady: think of people as vectors with a direction and a magnitude. Direction concerns why they are pursuing a goal and where they want to go; magnitude concerns the ambition and persistence they bring to it.

Stebbings recalls Grady’s reminder that a firm can meet many eventual winners and still miss them. Bek says Sequoia emphasizes that it is only as good as its next investment. For him, the lesson is to prepare seriously for each meeting rather than rely on past successes.

Bek credits Alfred Lin with a distinction between an exceptional operator and an exceptional founder. A person’s success inside an established organization can be valuable evidence but does not prove they will thrive in the different circumstances of founding a company.

He describes a framework attributed to Maguire that considers judgment and the ability to navigate complex institutions alongside intellectual and emotional skills. Bek says no single colleague is the best evaluator for every founder; expertise and context affect whom he would ask for a second view.

The conversation turns to Bek’s thesis that agents may become a new kind of customer. He cites traffic observations and a forecast attributed to Cloudflare, arguing that businesses should consider how automated systems find and use their products. The figures and future traffic ratios are claims in the interview, not independently verified.

Bek says much of the web has been optimized to translate human intent into a purchase through pages, onboarding and payment flows. If an agent acts on the customer’s behalf, businesses may need equally usable machine-facing interfaces. He expects that shift to affect some markets before others.

Stebbings asks whether user interfaces and brand loyalty will disappear. Bek rejects the fastest version of that conclusion: he argues that models can reflect biases from training and user behavior, and that existing brands or default choices may continue to influence agent recommendations.

Bek says businesses need to study how agents select providers across different models and tasks, just as they study human customers. He mentions Sequoia portfolio company Profound, which works on visibility in AI answer interfaces. Stebbings discloses an investment in a related company, Peak AI.

Asked whether answer-engine optimization could exceed traditional search optimization, Bek predicts a broader ecosystem of services built around agents. He calls this a new economy rather than merely one new marketing category. The potential size and timing are his thesis, not measured outcomes.

Stebbings asks how to distinguish tasks people will delegate from decisions they will keep. Bek says someone booking a holiday may still want to choose the destination even if an agent assembles the plan. He expects software-to-software choices in some technical workflows to become more automated.

On software margins, Bek rejects the assumption that agents can switch every provider instantly. A trivial purchase may be easy to change, but a database can involve accumulated data, controls and trust. He says switching costs depend on the product and enterprise context.

Stebbings cautions that large enterprises often adopt new buying habits more slowly than startups predict. Bek agrees that adoption varies across the curve and says agents currently appear strongest in tools connected to coding, with other functions potentially following on different timelines.

Stebbings says he feels more certain investing in infrastructure used by many applications than choosing the final application winner, and discloses investments in companies he names. Bek says his partnership invests in both layers because different businesses can become valuable on different timelines.

Bek contrasts an infrastructure provider addressing immediate model demand with Rillet, a portfolio company building finance software. He expects customer relationships and product data to compound over time. These are his assessments of businesses in which the firm has an interest, not independently verified growth projections.

Stebbings says adoption by customers outside the technology industry can be a useful signal. Bek describes Rillet’s effort to reach ordinary businesses as well as AI startups, saying that growth beyond a narrow technology audience can reveal a broader market. His customer descriptions are the company-side account.

Asked whether lower AI software margins matter less if markets become larger, Bek says some user-facing tasks may eventually work well with cheaper models. Other demanding tasks may still justify frontier models. He encourages companies to consider the cost and quality requirements of each workflow.

Bek argues that a company might accept weak or even negative early gross margins when the most capable model is needed to deliver a superior product, then seek better economics as alternatives improve. That is a proposed strategy with risks, not a guarantee that margins will recover.

Bek explains his prediction that a very large software company could sell what looks like a service outcome. He distinguishes a traditional labor-heavy services company from one using software and AI to deliver the outcome at different economics. The size and timing of any future company are forecasts.

He compares buying accounting software with paying an accountant to close the books. His question is whether a provider could charge for the completed outcome rather than only the tool. The illustrative price ratio is a framework from the conversation, not a general market statistic.

For customer support, Bek describes an example in which a provider charges for resolved requests rather than software seats. He mentions Sequoia portfolio company Sierra and illustrates how a customer might compare a cost per resolution with human labor. The example is hypothetical and does not establish a universal price or quality.

Stebbings asks what happens when an outcome is ambiguous or several activities contributed to a sale. Bek agrees that measuring attribution and replacing human judgment are difficult, which helps explain why outcome-based models are easier in some categories than others.

Bek distinguishes tasks with clear answers from judgments built on experience and context. He expects some tools that initially assist humans to learn from workflows and improve over time, but he does not claim that models can already capture every nuance of a human decision.

The two discuss why companies still need implementation specialists as they adopt AI. Bek mentions a company in which both speakers invest that assists software implementation, and claims tools can make specialists more productive. He stresses that a human remains involved in judgment; the productivity multiple is his example, not a controlled measurement.

Bek says his vision of more automated service delivery does not require eliminating people entirely. He imagines a changing mix of human work and software, with humans continuing to handle decisions that need judgment. The economic results will depend on the task and execution.

Stebbings worries that the transition may be faster than previous technological shifts. Bek agrees that some teams could become smaller, but expects new kinds of work and more demand as building becomes cheaper. Their discussion presents competing expectations rather than a measured forecast of employment.

Asked whether he would invest in a traditional services firm planning to become a software business, Bek says he probably would not. He thinks such companies may still make money, but his own venture strategy favors teams built around a technology-first product and the ability to attract specialized talent.

Bek acknowledges that other investors, including private-equity firms, may find attractive opportunities in established service businesses and their data. He says they may seek different outcomes from a venture fund. This is a preference and market view, not a finding that any particular company cannot adapt.

In the quick-fire round, Bek says he considers legal AI crowded and discloses his favorable view of Harvey, a Sequoia company. Stebbings responds that legal work contains distinct specialties and mentions investments in competing or specialized providers. Their company preferences are interested opinions, not investment recommendations.

Bek identifies brain-computer interfaces as an area he considers underfunded and says he watches where technically skilled people choose to work. His expectation that the category will develop over the next decade is speculative and does not establish a medical or commercial outcome.

Asked for an agent company outside the portfolio, Bek praises Cursor’s position in coding and its work deeper in the model stack. The panel refers to a reported transaction involving the company; it does not independently establish that transaction’s status or terms.

Bek names Trade Republic as a missed early opportunity from before joining Sequoia. He says he underestimated the size of the market and assumed another company would dominate, only to see the businesses develop with less overlap than he expected. The lesson he draws is to examine market structure rather than assume a single winner.

Asked which founder trait he will not compromise on, Bek chooses intensity, saying building a large business is difficult. He treats that as his investment preference, not evidence that a single personality type guarantees success.

Bek recounts losing an early firm investment in Revolut soon after beginning his venture career. He says he recognized the founder’s intensity and repeatedly tried to make contact, but the company chose another investor. The account and investment sequence come from his recollection.

Bek says he then asked to invest personally but lacked the money for the commitment. His mother offered to fund it and share the investment with him. He says he continued to hold and buy shares, while his mother later sold much of her position and retired. The private financial details and valuation figures are his account.

He recalls negotiating permission to make personal investments when his early salary was limited. Stebbings jokes that Bek’s mother had the better record because she participated in one successful investment without the others. Neither treats that anecdote as a replicable investment strategy.

Asked what view would divide the Sequoia partnership, Bek says partners have different opinions and that there is no single house view on AI. He believes distinct perspectives can help founders identify the partner whose thinking is relevant to them, even when colleagues disagree publicly.

For the next five years, Bek says he is excited about possible AI contributions to life sciences and biology, including research that might eventually help people with chronic conditions. He and Stebbings speak from personal hope; the conversation does not establish a particular treatment or cure.

Stebbings thanks Bek for the discussion and their long friendship. The episode closes with Bek’s appreciation, after a conversation spanning family, investment judgment and uncertainty about technological change.