别做五年规划，去跟着最强的人：Tamar Yehoshua 谈 AI 时代的产品领导力

Asked for the habits behind her career, Yehoshua says the most overlooked one is doing a really good job at whatever role you hold right now. Product managers in particular are ambitious and eye the next level, but nobody gets the next job without excelling in the current one. She also lists table stakes in tech: being technical, knowing the latest technology, understanding your product deeply, and understanding metrics.

Beyond that, she says leaders must understand people and their motivations. Building products means asking why someone wants to click a button and what will delight or frustrate them, and building teams means asking why someone wants to join and work hard. She credits her father, a psychiatrist, who quizzed her after family events on what others were thinking, which taught her to see the whole room.

She cautions PMs against relying too heavily on metrics and encourages building intuition by asking lots of questions. Quoting Marc Benioff, she recommends a beginner's mind and listening to customers. A feature you built does not automatically deserve front-and-center placement, because you have to earn that right.

Lenny asks what "a great job" really looks like. Yehoshua says it is not about achieving what you were asked to do, but about whether you built something people used and did the right thing for the business. A related test is whether you made the whole organization more productive, not only your own team.

She tells of her first management offer, early in her career as an engineer. She came back from a vacation and told her manager she did not think the cross-program team should exist, explaining how to organize the work instead. Her manager was stunned, agreed, and found her something else. Lenny sums the idea up in one word: impact.

Yehoshua says she loves well-run teams, but has seen that broken IT, broken marketing, turnover and reorgs are not correlated with a company's success. She gives examples of companies she knows where executives churn and people are unhappy yet the numbers are amazing, and a well-run company with a great CEO that flatlined. What matters is a focus on product-market fit, distribution, a working sales team and enough money in the bank.

Chaos is normal during hyper-growth because infrastructure and communication strain when half the company has been there less than six months. Once a company reaches a large scale, such as 5,000 or 10,000 people, it needs professional management and a growth-engine mindset. She also says some chaos is not okay, for example constantly changing strategy and projects so people can't achieve anything, and that some chaos simply may not suit you.

Lenny asks if strong product-market fit solves many problems, and Yehoshua puts it more bluntly: no product-market fit is a death sentence. Lenny recalls seeing nonstop chaos at Airbnb. She adds that this is not an excuse to skip building a functioning, motivated organization.

Yehoshua says she never had a five-year plan, and recently relieved a worried twenty-something simply by saying so. Early forks in the road feel significant, like an MBA versus a job or PM versus engineer. Her approach is to follow people from whom you will learn most, not just people you like, and to pick the best product thinker, engineer or salesperson you can find.

She suggests looking at where great people are going, since a nexus of talent does great things even if the company underperforms. Think of the PayPal Mafia, or her own network from years at Google. Lenny adds a list heuristic: write down the people you most respect and see where they work.

A friend's advice for leaders is to take a job where hiring people will make their careers, which she says raised her bar for every offer. She warns against chasing financial return because it is the hardest thing to predict, and one acquaintance who said he took a job as a "mercenary" regretted it. You don't need a top brand either: she made mistakes and joined companies that failed, and the point is that "skills can't be taken away."

At Amazon she had quarterly meetings with Jeff Bezos before AWS launched. He spoke last, after hearing every leader at the table, and he was consistent: principles like customer focus, and pet peeves like icons without labels, made it easier to operate in his company. When she mentioned that competitors had ten times as many people on a product, he replied that this was her advantage, and talked about taking a long-term view.

She went to Slack to work with Stewart Butterfield, whom she calls the best product thinker in the Valley. His 2014 master plan was a four-box grid: build a product people love, build a network (Slack Connect), build a platform (Slack Platform), and do some "magic AI stuff." What she learned most from him was prototyping: he would say he couldn't tell if something would work until he felt it, once asking designers to put everything in the interface behind a single button as a step toward a redesign.

On prototyping time, she says it makes you faster if your engineering infrastructure supports it and engineers are willing to throw code away. Product, design and engineering have to share the mindset, and a Google front-end head once called prototypers his secret weapon. From Marc Benioff she took a lesson about marketing: she appeared in his Dreamforce keynote twice and says he approaches marketing presentations the way a product person approaches building a product.

Prompted by a former colleague, Lenny asks how to build better relationships with engineers. Yehoshua says it is the most important thing a product leader does, since great ideas go nowhere if they cannot be built. She advises evaluating your engineering partner before you join a company, and being clear about roles so that no one can play "mom against dad."

With Slack's CTO Cal Henderson she split ownership, never went around him, and held OKR reviews, weekly executive reviews and Monday red-yellow-green status meetings together with their chiefs of staff. She says it came down to respect and trust that each follows through, and to spending a lot of time together, documenting decisions and speaking up when you disagree.

As teams grew, OKR reviews became unwieldy, and at one point they added up to hundreds of hours. After Slack Clips launched, teams started posting a doc plus a short video, and she and Cal watched them together in a marathon, then posted follow-up questions in channel. They iterated on the process every quarter, as they would on a product.

Yehoshua says a common PM mistake is over-indexing on those who will be unhappy about a launch. This shows up most when you remove features: a small set of users who depend on them is vocal, but more people will be made happy. Usually more people will use your product tomorrow than today, so design for the larger future audience.

The key is to be respectful, transparent and authentic, avoiding marketing speak, and to give people choice and time to move, as Slack did when shifting from calls to Huddles. She recounts how Google's Rachel Whetstone, asked for an FAQ, instead phoned an upset IT engineer, listened and explained, and he changed his mind and told others.

Lenny links this to a parenting book about listening, and Yehoshua adds that her favorite parenting book is *How to Talk So Kids Will Listen*. She says the same applies to customers: people want to be heard and to understand your reasoning.

Yehoshua, who has a master's in AI, says she was skeptical until she saw GPT-3, and she thinks we are underestimating how much AI will change work. It will not happen overnight, but people who work out how to use it will pull far ahead. In five to ten years, she expects PMs, designers and engineers to blur as AI helps PMs prototype and designers generate working designs, though tools like Copilot and Cursor still need experienced engineers to catch errors.

Her advice for anxious people is to use the products, as good PMs always did with mobile. She cites a PM who fed a whole Discord channel into Gemini to ask about sentiment and requested features, and her own use of Gong transcripts in Glean, which required prompt tweaks so the summary distinguished customer requests from salespeople's suggestions. She also built a Glean prompt that checks a launch date against open tickets, Slack conversations and beta feedback to estimate confidence.

On whether AI threatens PMs, Lenny argues PMs are best placed because they know what to build. Yehoshua agrees with a caveat: AI is not good at creativity, so PMs who only do grunt work may lose out while strategic, creative ones thrive. She does not expect fewer engineers or PMs overall, since every company always wants more built and she thinks teams will simply get much more done.

Yehoshua explains that Glean started as enterprise search, indexing content from SaaS tools like Slack, Jira and Salesforce, using BERT models and vector embeddings from 2019. GPT-3 then added a natural-language chat interface. People understand search, but chat is harder: retention is low, and in her first week she learned that a big issue is users asking things the system cannot possibly know, like their next week's top priority.

To help, you need guardrails such as autocomplete, refinements and suggested prompts, and you have to educate users about non-determinism, since a CIO who enjoys ChatGPT on weekends expects deterministic software at work. She also warns that the industry changes so fast that work compensating for model weaknesses will vanish. Your differentiator must be something that gets better as LLMs get smarter.

Closing, she says she has never seen tech move so fast, that it is both exciting and scary, and that you have to change how you work to keep up. She keeps up with AI newsletters such as Ben's Bites and The Neuron and podcasts including No Priors and Cognitive Revolution.

Her book recommendations are *Switch* by Chip and Dan Heath, about setting a path and motivating people to follow it, and *Team of Rivals*, about Lincoln assembling his cabinet. She also mentions the Magpie Murders series and a small-batch Oakland dark chocolate called Bisou. Her favorite tech product, she admits, is Glean, which she uses 10 to 15 times a day.

Her life motto comes from her father, who told her when she was indecisive about college that "there are no right decisions, you make a decision right." You commit and avoid regret. For new parents she recommends *Healthy Sleep Habits, Happy Baby* and advises sharing your own day with your kids, so that they will share theirs. She can be found on LinkedIn and writes a Substack called Practical Intelligence.