Sherif Mansour: Product Managers Need to Give Teams Direction, Not Just Code

Atlassian’s AI-builder experiment exposed a shortage of shared direction. Mansour explains how PMs can fill it.

SMSherif Mansour게스트 · Atlassian
채널
Lenny's Podcast
게시일
2026년 11월 22일
영상
1:21:38
읽기
11분
원문
영어
YouTube에서 열기
스크롤하면 영상이 오른쪽 아래로 작아져 계속 재생됩니다. 자막은 읽기 언어를 따릅니다.
읽기 언어2개 언어로 읽을 수 있음
이 글은 아직 한국어 번역이 없어 영어 원문으로 표시합니다.
요약

30초 요약

Atlassian tried making everyone on some teams an AI builder. Within a month, people returned to their specialties because they needed direction more than additional output, says its head of AI and product craft, Sherif Mansour. He sees the product manager’s job as sharing customer, market and business context so the whole team can decide faster. He also hires juniors to bring new AI habits alongside experienced colleagues who can judge quality, and uses customers’ workarounds to discover features worth building.

  1. 1A shared understanding lets the team make decisionsMansour groups the necessary context into customers, product and market, and business. Engineers who join customer conversations can make decisions without waiting for the PM.
  2. 2Building faster exposed a shortage of directionAtlassian tried squads where all disciplines built with AI. After an initial speedup, participants returned to their specialties because aligning the work became the constraint.
  3. 3Junior and experienced workers bring different AI strengthsMansour says juniors often introduce new tools, while more experienced colleagues can spot errors and evaluate quality.
  4. 4Customer workarounds can reveal what to buildA general chat helped Atlassian see demand for grouping whiteboard notes; the team could then turn that use into a visual feature.
  5. 5New managers can step back too soonMansour says good managers give the team room but also join in to explore alternatives and help reach an outcome.
  6. 6Ask who decides when you disagree with a leaderMansour recommends stating the disagreement, clarifying conviction and decision ownership, and explaining both sides to the team.

핵심 정리

  1. Share customer, market and business context so colleagues can make decisions without waiting for the PM.
  2. Invite engineers to customer interviews instead of relying only on summaries.
  3. Pair junior colleagues’ willingness to try new AI tools with experienced colleagues’ ability to judge output.
  4. Watch customer workarounds on a general-purpose canvas to identify useful product features.
  5. Ask a senior stakeholder whether an opinion is a directive and who owns the decision.
전문

챕터별로 다듬은 전문 읽기

원본 자막만 보기
01

A shared brain for product decisions

Mansour argues that PMs help teams decide by sharing customer, market and business context. Direct exposure to customers can produce ideas that a handoff misses.

Atlassian experimented with teams where product managers and designers all became AI builders. At first they delivered faster; after a month, one product manager reported that people had returned to their specialties because direction, not output, was the constraint.

Sherif Mansour, Atlassian’s head of AI and product craft, says a product manager’s most important job in a medium or large organization is to fill gaps in context. Instead of making every decision, the PM helps the cross-functional team develop a shared understanding.

He groups the information a team needs into customer context, product and market context, and the business context in which it operates. As organizations scale, trust and shared understanding break down, making that common ground harder to maintain.

Mansour tells PMs to stay away from the backlog when they can. If the PM knows why one task matters more than another, the team needs the same context so engineers and designers can make the choice themselves. The particular gap changes from vision to customers to market over time.

Lenny Rachitsky observes that this makes the PM responsible for equipping others rather than being the boss who approves each move. Mansour remembers feeling buried in decisions and sprint planning before he learned that sharing context helped both him and the team scale.

He tries not to conduct customer interviews without engineers present. An AI-generated summary can pass along findings, but joining the conversation gives an engineer their own moment of understanding and a direct sense of the customer’s needs.

The same principle applies to strategy: the team can explore the market, try tools and develop a vision together rather than receiving instructions after one person does the research. Their involvement can change what they notice and propose.

To identify a gap, Mansour asks why he reached a priority and what information another person is missing. Good designers likewise involve colleagues in exploring concepts, then refine the result, instead of only handing over finished designs.

One PM asked every engineer to try different AI products and report what worked. Atlassian also brings engineers to customer event booths to hear feedback directly. If teammates propose an unexpected roadmap item, Mansour asks whether they lack a strategic belief he has not shared—or whether they are right.

A PM in a squad normally has no direct authority over its members. Mansour prefers a model in which engineers understand and help shape the vision over one in which the PM hands over specifications and expects compliance.

In 2010, Mansour and an engineer watched Confluence customers mark unfinished meeting tasks with red stars and completed ones with green. The engineer proposed putting a checkbox in the document itself. Seeing the workaround firsthand helped her recognize the opportunity.

02

Match PM strengths to the work

A triangle of business, scientific and artistic approaches helps teams choose complementary skills. Mansour discusses career opportunities that build on strengths.

Mansour uses a product craft triangle from Joff Redfern: the general manager emphasizes business opportunities and revenue, the scientist favors hypotheses and experiments, and the artist begins with vision and experience. The framework helps a team choose a PM for its actual work.

Asked how to improve a favorite product, a general-manager type might size a business opportunity, a scientist might propose experiments, and an artist might describe a desired experience. Mansour says few people excel equally at all three, although generalists can be especially useful in a startup or a new project.

Rachitsky says the framework can explain why someone struggles on the wrong team. Mansour has seen PMs regain energy when moved to work better suited to their strengths. Teams should consider how a person complements others, not simply add another PM.

Mansour describes himself as a strong artist and a weak general-manager type. He talks openly about weaknesses so colleagues can acknowledge theirs, then pairs with people whose strengths complement his. Early in a career, he recommends trying different kinds of work before deciding where to deepen expertise.

Rachitsky adds that people still need a basic level of competence in other areas. Mansour agrees: exceptionally broad leaders exist, but most people benefit from developing the areas in which they are naturally strongest.

For career growth, Mansour tries to offer someone a task just above their current level that matches their strengths. He recalls stepping aside from a growing product so a PM with a stronger experimental bent could lead the next phase.

He suggests looking back at work for which colleagues praised you, examining how you approached it, and asking your manager for similar opportunities. A more senior PM can continue deepening a strength or deliberately try something different; advancement need not mean only moving up a level.

03

Faster building makes direction more important

Atlassian’s AI-builder experiment led participants back to their specialties. Mansour explains how PMs can use AI to close gaps in team context.

Rachitsky says his forthcoming hiring report shows more open PM roles than other tracked roles, while engineering openings have slowed and design is flat. He asks why PM demand appears strong after predictions that AI would make the role obsolete.

Mansour describes Atlassian’s experiment with squads in which engineers, designers and PMs all built with AI. An early burst of speed gave way to a recognition that the group needed people to set direction, not just produce more software.

His internal shorthand is “direct the ship, don’t row it.” A startup or brand-new project may need everyone to do whatever the moment requires; in an established company with many engineers, he believes alignment can matter more than adding another person writing code.

Rachitsky suggests easy prototyping can create many ideas while leaving teams uncertain about goals and quality. Mansour compares it with the earlier spread of mockup tools: many people could make a mockup, but access to a tool did not make every PM a good designer.

AI can amplify both strengths and weaknesses, Mansour says. Product, engineering and design may each cover more ground, but he expects their expertise to remain important to the quality of the outcome.

Asked whether PMs should ship production code, he returns to decision speed. First identify what information or understanding is missing from the team; then decide whether AI can help close that gap, perhaps through a prototype or market research.

04

Why juniors belong on AI teams

Junior workers can bring new AI habits while experienced colleagues assess quality. Mansour also explains why he values experience-oriented product managers.

For established organizations, he increasingly values PMs with a strong feel for experience, as well as people who bring new ways of working with AI. In interviews he asks for more detail than simply whether someone uses AI to summarize documents.

Mansour pushes back on the idea that AI eliminates the need to hire juniors. New entrants may not have established work habits to unlearn and can bring unfamiliar tools and practices to an existing team.

That does not make experience disposable. Juniors may struggle to recognize poor AI output without domain knowledge; experienced colleagues can evaluate quality and identify errors. He wants a mix in which each group contributes what the other lacks.

During Atlassian’s AI builder weeks, people from engineering, design and other crafts stop normal work and experiment at the same time. Mansour says younger participants often demonstrate new tools, while senior colleagues tend to focus on evaluation and quality.

Atlassian’s junior PM cohort includes people from computer science, arts and law-related backgrounds. Mansour looks for curiosity, critical thinking and a willingness to try things. He says company and customer research suggests juniors adopt new tools more readily, while seniors get better quality and productivity from tools they use.

An artist-type PM often starts with a vision of an experience and works back toward the underlying problem. A scientist may lead with measurement and a general manager with a business case. He wants these approaches to create productive tension rather than treating one as universally superior.

As software becomes cheaper to produce and competitors gain similar features, Mansour expects the quality of the user experience to become a stronger differentiator. He does not expect every interaction to collapse into a chat with an agent.

He says an exercise can reveal an artist’s instinct even in a junior candidate. One applicant led with charts and market size; another walked through a complete user experience and discussed how it might create an emotional connection.

Rachitsky asks whether an artist’s instinct can be taught. Mansour says it can, though he thinks it is generally easier to accelerate in something that already comes naturally. Diverse teams remain useful as AI changes the work.

Rachitsky wonders whether flat hiring for designers means PMs are filling the gap. Mansour offers another hypothesis: companies may not yet have recognized how much experience will matter. He says design has an unusually strong role in shaping Atlassian’s roadmap, while acknowledging his broader hypothesis remains to be tested.

05

Start with a solution when it reveals a problem

General-purpose AI and existing customer workarounds can expose features to build. Adding AI within established workflows can also be a useful route to change.

“What problem are we solving?” is a useful question, but Mansour says starting with a solution can also expose valuable problems. In Atlassian’s Rovo assistant, a general chat and user-created agents allowed people to try many jobs before the company knew which should become dedicated features.

For example, people used chat to group notes on a Confluence whiteboard by theme. Because selecting and organizing the notes through chat was awkward, the team could make that observed use a visual feature. A general canvas became a way to discover a more focused product.

He says customers also put agents into Jira workflows for tasks such as triage and code review. Some uses should remain customizable; others can become native features once their frequency and value are clear. He compares the approach to paving a path that people have already made for themselves.

Another place to begin with existing solutions is company strategy: list assets and distinctive capabilities, then ask what other problems they can solve. Mansour uses Amazon’s large book inventory as an analogy for a capability that could support a broader store.

He cautions that observing workarounds still requires talking to customers. AI can make it easier to stay behind dashboards, but a team needs to understand what people are trying to accomplish.

“Do not bolt on AI” is another slogan Mansour wants to qualify. For an established product with millions of users, adding AI at points in existing workflows can be a practical way to evolve the experience rather than discarding the product and starting over.

Customers have already modeled many business workflows in Jira, he says. Atlassian can look at where work is handed between people and introduce agents at useful points, including formal steps in a process and less structured habits.

06

Advice depends on context

Mansour and Rachitsky discuss why opposing product advice can both be useful. Mansour describes product engineers as carriers of shared context.

Mansour argues that advice should carry its context. An AI builder may be a good fit for a new project, while a mature organization with plenty of developers may need clearer direction. Starting from a problem and starting from a solution can both be useful depending on the work.

He describes product engineers as colleagues who extend a PM’s reach by carrying and applying shared customer and strategy context. Rachitsky says he prefers presenting different perspectives to forcing all guests to agree, so listeners can decide which advice fits.

07

Managers, individual contributors and founder decisions

Senior individual contributors can shape strategy, while new managers may become hands-off too quickly. Mansour explains how he handles disagreement with a strong stakeholder.

Mansour rejects the assumption that people must manage others to participate in important strategy decisions. Atlassian has senior individual contributors working at executive level, and people can move between management and individual work as their strengths and circumstances change.

Rachitsky argues that management is itself demanding work. Mansour says Atlassian has both managers with larger teams who concentrate on coaching and managers with fewer reports who also own projects. The mix depends on people and the work.

The common mistake Mansour sees in new PM managers is becoming hands-off too quickly. Good managers create space for their teams, but also get involved when needed, explore alternatives together and help the team reach an outcome.

When working with visionary founders, Mansour imagines that the leader is on chapter ten of a movie while the rest of the team is much earlier. Passing an instruction along without understanding its reasoning produces compliance without conviction.

He asks what led a leader to a conclusion and tries to catch up on the thinking. He says experienced leaders are often right in his experience, particularly in complex enterprise markets, but the goal is understanding the context and reaching the best customer outcome, not automatically proving them right.

If he still disagrees, Mansour states the case and asks who owns the decision and how strongly the other person feels. Once a decision is made, he presents the real advantages and disadvantages of both options to the team rather than attributing the choice to a particular executive.

Rachitsky describes Dharmesh Shah’s practice of labeling an executive comment as information, a suggestion, a recommendation or a plea. Mansour says PMs can ask for that clarification themselves when someone senior has not signaled whether an opinion is a directive.

08

Creating opportunities and sharing context

Mansour discusses advocacy, life outside work and Atlassian’s open documentation. He is investigating whether remote teams’ written context helps them use AI.

Mansour asks listeners to create opportunities for others, just as someone created opportunities for them. The practical step is to identify a person whose potential they believe in and advocate for them.

In the closing questions he mentions reading fantasy books with his children and returning to the Bible. He enjoyed the film Project Hail Mary and describes playing destination-themed games with Gemini inside Google Maps during a family drive.

The maps example illustrates his workflow argument: the family could talk to AI without leaving the navigation screen, while the destination supplied useful context. For Mansour, the experience was valuable because it fit an activity already under way.

He encourages people to separate their identity from their job and maintain relationships outside work. Rachitsky shares the reminder that children are the people most likely to remember a parent consistently working late.

Mansour explains that a display behind him is made from Atlassian shirts, each commemorating a product or company moment. His wife stitched favorites together as a birthday surprise, and he later used the collection to tell stories from the company’s history.

He describes Atlassian as open by default: executive reviews are often recorded and searchable internally. Shared documentation gives people and AI agents more context, though Rachitsky observes that easier access can also expose information to people who should not see it.

Mansour is testing a hypothesis that remote teams may obtain more value from AI because they write and share more of their work digitally; he says data scientists are still investigating. Both in-person collaboration and distributed work have benefits.

His final request is to ask why advice worked in its original setting before repeating a compelling headline. The same apparent contradiction can make sense under different conditions.

같은 채널의 다른 글

1:27:15EN
Lenny's Podcast· 2026년 11월 15일팟캐스트 · 2명Mona Sarantakos: How Meta built trust into Muse before launch게스트 Mona Sarantakos · 진행자 Lenny RachitskyMuse’s product lead explains its seven-month build, the surprise popularity of shopping, and why real user stories are its strongest activation tool.AI Product ManagementAI agentsProduct Strategy+2
58:33EN
Lenny's Podcast· 2026년 11월 8일팟캐스트 · 2명Gemini’s product chief on testing models and building ahead of them게스트 Tulsee Doshi · 진행자 Lenny RachitskyTulsee Doshi explains a small-eval starting point, her 60-20-20 product framework and why model behavior itself is a user experience.Product ManagementFoundation Models평가(Evals)+2
1:27:47EN
Lenny's Podcast· 2026년 11월 1일팟캐스트 · 2명Hector Serrano on Building a Paid Creator Program at Wispr Flow게스트 Hector Serrano · 진행자 Lenny RachitskyHow a creator-led growth team recruits, compensates and supports paid UGC creators—and uses founder storytelling alongside ads.Growth StrategyCreator EconomyContent Strategy+1
1:25:47EN
Lenny's Podcast· 2026년 10월 25일팟캐스트 · 2명Rob Snyder’s PULL Framework for Finding Early Demand게스트 Rob Snyder · 진행자 Lenny RachitskyA practical conversation about customer urgency, alternatives, early sales and retention.Company BuildingProduct Strategy
1:20EN
Lenny's Podcast· 2026년 9월 4일팟캐스트 · 2명Tara Seshan on OpenAI’s Vision for ChatGPT and Work게스트 Tara SeshanA product lead explains a goal of simplifying model and interface choices.OpenAI제품과 UXAI agents
0:46EN
Lenny's Podcast· 2025년 12월 31일팟캐스트A View on Effort and Team PerformanceA short argument that difficult moments can reveal how teams respond under pressure.LeadershipCompany Culture