Monday, 7:10 a.m.
Maya Smith’s first Monday as Vice President of Product at Brindlepath started before she reached the office. On the train, she opened the overnight digest her team’s research agent had assembled: forty-one support tickets summarized into six themes, two competitor release notes, a spike in failed logins at one hospital customer, and a draft agenda for the quarterly business review with Ambermoor Health, Brindlepath’s largest account.
It was a good digest. It was also, she noticed, wrong in one important way. The agent had grouped eleven tickets under “notification fatigue” and recommended adding a setting to mute alerts. Maya read three of the tickets in full. The nurses were not tired of alerts. They were getting alerts for patients who had already been discharged to another team, because the handoff between the hospital’s care managers and the outpatient navigators did not update ownership in Brindlepath. Muting the alerts would hide the symptom and keep the disease. The real problem was that nobody owned the patient for the forty-eight hours after discharge.
That distinction, between the request and the problem underneath it, is the job. The agent did in twenty minutes what used to take a product manager most of a Monday. It could not do the part that mattered: notice that the pattern meant something different from what it looked like, and decide what to do about it.
This chapter is about that part.
What a product leader is actually for
Ask ten companies what a product manager does and you will get eleven answers. In some, the product manager writes tickets and runs standups. In others, the product manager sets strategy for a business line with its own profit and loss. The title covers both.
Underneath the variety, the role has one constant: the product leader owns the outcome. Engineering owns how the product is built. Design owns how it works for the people using it. Sales owns the deal. Product owns whether the whole thing creates value for customers and for the business, and whether the team is working on the right problem in the first place.
Owning the outcome means answering four questions continuously, not once:
- Is it valuable? Will the people it is for choose to use it, or buy it, over what they do today?
- Is it usable? Can they figure out how to get that value without a training manual?
- Is it feasible? Can we build and run it with the time, skills and technology we have?
- Is it viable? Does it work for the business: can we sell it, support it, price it profitably and stay within the law?
These four risks will be familiar to anyone who has read Marty Cagan’s work, and they remain the best short description of the job.1 This edition adds a fifth question that used to sit quietly inside viability and now deserves its own line:
- Is it trustworthy? Will customers, regulators and the public trust it with their data, their decisions and, in healthcare, their health? When a product includes AI, does it behave reliably, explain itself where it needs to, and fail safely?
Trust used to be a compliance checkbox at the end of a project. In a world where software makes recommendations about patients and AI writes answers that look authoritative whether or not they are right, it belongs at the start.
How the role got here
It helps to know where the job came from, because many organizations are still running an older version of it.
The feature coordinator (roughly the 1990s and early 2000s). In the era of boxed software and long release cycles, product management grew out of marketing and project management. The product manager gathered requirements from customers and sales, wrote them into a large specification, and handed it to engineering. Success meant shipping the spec on time.
The product owner (the 2000s and 2010s). Agile development broke big specifications into small increments and gave each team a backlog. The product manager became the person who prioritized that backlog and accepted the work. Delivery got faster, but in many companies the role shrank to managing tickets. Teams shipped steadily and still built things nobody used.
The outcome owner (the late 2010s onward). SaaS changed the economics. When customers pay every month and can leave any month, shipping is not success; adoption, retention and expansion are. Product managers began to be measured on outcomes: activation rates, net revenue retention, time saved for the user. Discovery, the work of finding out what to build before building it, became a core part of the job rather than a phase at the start.
The orchestrator (now). The current shift is less about what the product manager is accountable for and more about how the work gets done. A product manager now directs a team that includes AI tools and agents alongside engineers, designers and analysts. Research synthesis, first drafts of requirements, competitive monitoring, data pulls and test-case generation can all be delegated. The product manager’s leverage has grown enormously. So has the cost of pointing that leverage in the wrong direction.
Each stage absorbed the one before it. A great product leader today still needs to write a clear requirement, still needs to run a tight backlog, and still needs to be measured on outcomes. The new layer sits on top.
The four pillars of product leadership
The first edition described three pillars of SaaS product leadership. Experience since then has added a fourth.
Pillar 1: Market understanding
You cannot build for a market you do not understand. Market understanding means knowing who has the problem, how badly, what they do about it today, who else is trying to solve it, and what constraints shape the solution. In healthcare those constraints include regulation, reimbursement and clinical workflow; in fintech they include licensing and fraud; in education, procurement cycles that follow the school year. Every market has its own physics. Chapters 2 and 3 are about learning them.
Pillar 2: Financial insight
This is the pillar most product managers neglect, and the one that most separates a product manager from a product leader. Every product decision is also a financial decision: what it costs to build, what it costs to sell and support, how it changes what customers will pay and how long they stay. A product leader who can walk into a budget meeting with a clear view of unit economics gets resources. One who cannot gets opinions. Chapters 6 and 7 cover the numbers in detail.
Pillar 3: Customer-centered innovation
Innovation is not novelty. It is a better way to get a job done for a real person. Customer-centered innovation means staying close enough to users to see the problems they have stopped noticing, and having the discipline to measure whether what you shipped actually helped. In healthcare, that includes the patient who never logs into your software but whose care depends on it.
Pillar 4: Trust and responsibility
The new pillar covers security, privacy, compliance, accessibility and, increasingly, responsible AI. It is not a department’s job that product hands off; it is a property of the product that product leadership owns. A breach, a biased model or an inaccessible workflow is a product failure, not just a technical one. Chapter 11 covers trust by design, and Chapter 17 covers the specific risks of AI features.
All four pillars rest on the same foundation: owning the outcome. A product leader who is strong in three pillars and weak in one will be found out by that one.
What great looks like
Over the years I have hired, managed and worked beside a few hundred product managers. The great ones share habits that have little to do with frameworks.
They can say what problem they are solving in one sentence, without jargon. If you cannot explain the problem to a smart outsider, the team does not understand it either.
They talk to customers every week, not every quarter. Not sales calls and not escalations, though those matter too. Conversations whose only purpose is to understand how someone works.
They separate the request from the need. A customer asking for a mute button for alerts is telling you something true about their pain and something unreliable about the cure. Maya’s Monday morning is the job in miniature.
They make decisions and write them down. A decision log, with the reasoning and the evidence, is worth more than any roadmap slide. It lets the team move without waiting for the product manager, and it lets everyone learn when a decision turns out wrong.
They say no, often and kindly. Every yes is a no to something else. Great product leaders protect the team’s focus and explain the trade-off rather than hiding it.
They know their numbers. Not every metric in the dashboard: the three or four that tell them whether the product is working, and the one or two that tell them whether the business is.
They share credit and own failure. A product leader has authority over very little and influence over almost everything. Influence comes from trust, and trust comes from how you behave when things go wrong.
Where the week goes
One of the fastest ways to understand how the job is changing is to look at a product manager’s calendar. Figure 1.2 compares a typical week for a product manager at a growing SaaS company before AI tools, and the same role with AI assistants and agents handling much of the preparation and synthesis. The numbers are illustrative, based on the pattern I see in teams I work with, not on a survey.
Two things stand out. First, the time saved is real and large, mostly in research synthesis, status reporting and first drafts. Second, and more important, the time does not automatically flow to the work that matters. In teams that adopt AI tools without changing their habits, the freed hours fill with more meetings and more documents. In teams that do it well, the hours go to customer conversations, sharper decisions and coaching. That reallocation is a leadership choice, not a side effect of the tools.
Brindlepath in practice: the decision log
One of the first habits Maya introduced was a decision log: a shared document where every significant product decision was recorded with five fields: the date, the decision, the options considered, the evidence, and what would make the team revisit it.
The first entry was the one from her Monday morning. Decision: do not build a mute setting for notifications. Options: mute setting; change notification rules; fix ownership after discharge. Evidence: 11 tickets, three read in full, all caused by patients already discharged to another team. Revisit if: notification complaints continue after ownership is fixed.
Within a few months the log had changed how the company argued. When the head of sales pushed again for the mute setting, Maya did not have to win the argument from memory; she pointed to the entry and asked whether the evidence had changed. It had not. When a decision did turn out wrong, such as an early choice to show discharges from all units by default, the log made it easy to see why it had seemed right at the time and what the team had missed. Nobody was blamed. The team simply learned faster.
The log also became the best onboarding tool the company had. New product managers read the last six months of entries in their first week and understood more about the product’s history and reasoning than any slide deck could tell them. Dev Raman’s engineers began reading it too, because it told them why they were building things, not just what. Two of them started adding their own entries for technical decisions that affected the product.
The decision log is not complicated, and it is not glamorous. It is one of the cheapest ways a product leader can make the team’s judgment visible, shareable and improvable, which is, in the end, what owning the outcome requires.
Common mistakes
Acting as the team’s boss. Product managers who issue instructions to engineers and designers get compliance, not ideas. The role carries accountability for the outcome, not authority over how others do their work. Lead through clarity, evidence and trust.
Becoming the team’s secretary. The opposite failure: taking notes, chasing tickets and scheduling meetings while nobody owns the hard questions. If most of your week is coordination, you are doing a project manager’s job, and nobody is doing yours.
Confusing busy with valuable. A full calendar and a long list of shipped features feel like progress. The question is whether anything changed for customers. Review your last month: which of your hours moved an outcome?
Ignoring the business. Product managers who talk only about users and never about revenue, margin or retention end up with roadmaps the leadership team does not trust. The four pillars need each other.
Treating trust as someone else’s job. Security, privacy and accessibility failures land on the product, and on its leader, regardless of which department was supposed to catch them.
Letting tools set the agenda. New AI tools arrive every month. Adopting them without changing how you work just produces more output. Decide what you want more time for, then choose tools that create it.
Try this
Take one week of your calendar and sort every hour into four groups: customer understanding, decisions and strategy, team leadership, and coordination or reporting. Then ask three questions. Which group is largest? Which recurring tasks in the coordination group could an AI assistant draft for you? And what would you do with five extra hours a week if you had them? Write the answer down, and protect those hours on next week’s calendar before anything else fills them.
How AI changes this
AI changes the product leader’s job in three ways, and it is worth being precise about each.
It removes toil. Synthesizing interviews, summarizing support tickets, drafting requirements, writing release notes, pulling data, monitoring competitors, preparing meeting briefs: all of these can now be done in minutes by a well-instructed assistant. A product manager who still does them entirely by hand is choosing to be slower than their peers.
It raises the bar on judgment. Because the first draft is now cheap, the value moves to knowing whether the draft is right. Maya’s research agent produced a plausible, well-written, wrong conclusion. The skill is not producing the summary; it is reading three tickets in full when the summary does not smell right. Product leaders need to become excellent editors of machine output: checking sources, spotting confident errors, and asking what the summary left out.
It changes what the team can attempt. A small team with good AI tooling can prototype, test and learn at a pace that used to require a much larger organization. That makes discovery cheaper and lowers the cost of being wrong early, which is good. It also makes it easier to drown in output, which is not. The constraint shifts from capacity to clarity: the team that knows exactly what problem it is solving will outrun the team with more tools.
What does not change: accountability, relationships and taste. An agent cannot be accountable for a product decision, cannot earn the trust of a skeptical chief medical officer, and cannot feel the difference between a product that is merely correct and one that people love. Those stay with you.
A practical rule I use with teams: delegate the first draft, never the final call. Let agents gather, sort, summarize and propose. Keep humans on decisions that affect customers, money, safety or people, and make sure a human has actually read the evidence behind each one.
Why SaaS, and why healthcare examples
Most of the principles in this book apply to any product. I focus on software sold as a service because the subscription model makes the product leader’s job uniquely continuous. There is no launch-and-move-on. Every month, customers decide again whether the product is worth paying for, and every release can raise or lower that answer.
The examples lean on healthcare because that is where I have spent most of my career, and because healthcare is a demanding teacher. It has every hard problem in one place: many stakeholders with different goals, a buyer who is rarely the user, heavy regulation, entrenched legacy systems, and real consequences when software fails. The lessons travel. If your product serves banks, schools, factories or retailers, swap in your own regulations and stakeholders and the frameworks hold.