See pricing
Thriving in chaos: top leaders’ strategies from Masters of Scale Summit
Masters of ScaleLeadership

Thriving in chaos: top leaders’ strategies from Masters of Scale Summit

A live Masters of Scale panel turns chaos into a practical operating model: plan to the length of your forecast horizon, use AI to widen judgment instead of replacing it, and distinguish real speed from sloppy haste. It is especially useful for PMs leading through fast-moving markets, noisy roadmaps, and AI overload.

November 3, 2026 · 31 min listen · 13 min read · DJ Patil, Raina Kumra, Hayden Brown, Baratunde Thurston, John Mills, Dara Treseder, William
0:00
–:––

Context

This live Masters of Scale Summit session is about what it takes to operate when the environment changes faster than the plan. Bob Safian brings together DJ Patil, Raina Kumra, Hayden Brown, Baratunde Thurston, John Mills, Dara Treseder, and William to talk about chaos theory, AI agents, public-safety decision making, and personal prioritization. For PMs, the episode matters because it reframes roadmap planning, AI adoption, risk management, and focus: the job is less about predicting the full path and more about building systems that stay useful when the path changes.

The Big Idea

When the world is chaotic, the winning move is not to predict every step. It is to build calm, flexible systems that keep judgment intact, shorten the planning horizon when needed, and let you move quickly without getting sloppy.

Across the panel, the strongest ideas point to the same operating model. Use AI and data to widen your options, but keep humans responsible for meaning, precision, and priorities.

Key Insights

Forecast horizon matters

DJ Patil's point is that chaos is not one uniform condition. With the double pendulum and the weather example, he argues that the real question is how far ahead you can forecast, not whether the system is complex.

  • A business environment can be mathematically understandable and still become hard to predict after a short time.
  • In some periods, you can plan confidently weeks or months ahead. In other periods, the useful horizon may collapse to days or even hours.
  • For a PM, that changes the right operating cadence. If the forecast horizon is short, detailed long-range commitments create false confidence. If the forecast horizon is long, constant thrashing is wasteful.

Calm beats keeping up

Raina Kumra rejects the idea that anyone can fully keep up with the current pace of AI change. Her line was simple: there is no keeping up, only keeping calm.

  • New capabilities appear so quickly that feeling behind is often misleading. The thing you feel late to may have appeared only a week ago.
  • Raina now manages a team of 12 AI agents. Her point is not that everyone needs exactly that setup. Her point is that modern knowledge work increasingly includes managing systems, not just doing tasks yourself.
  • The part you cannot outsource is judgment. Some agents are smarter than she is and better at math, she says, but some have the judgment of a toddler. PMs can hand off research, drafting, or analysis, but they still own what reaches customers, executives, or teammates.

Roadmaps need pioneer logic

Hayden Brown says leaders need to think less like navigators following a fixed map and more like pioneers moving toward a horizon.

She is not arguing against goals or metrics. She is arguing against pretending you know the exact route in advance.

  • Upwork has been pulling a public company back toward startup behavior. Hayden describes that as reintroducing agility and flexibility in both mindset and mechanics.
  • The operating pattern is: pick a direction, measure progress, learn continuously, and keep reevaluating what the next horizon should be.
  • That is especially relevant for PMs building in fast-moving categories. A roadmap can still express intent, trade-offs, and expected outcomes. What it should not claim is false certainty about the path.

AI should widen judgment

The most useful AI examples in the episode are not replace-the-human stories. They are let-the-human-see-more stories.

  • Upwork built an internal tool called crystal ball that sits on top of structured, governed business data. Hayden says questions that once took analysts weeks can now be answered by one person in hours.
  • John Mills says Watch Duty uses AI because it is a signals-intelligence problem. AI helps humans find the needle in the haystack faster and with less fatigue.
  • Both examples depend on the same condition: the human still interprets and validates the result. AI is valuable when it expands range and speed without removing accountability.

Speed is not haste

John Mills draws a hard line between moving quickly and moving carelessly. In public safety, that difference can literally be life or death.

  • His analogy is a surgeon. Nobody says, I had the fastest surgeon in the world. People want the most precise and accurate surgeon.
  • Watch Duty aims to be fast within minutes, because that is fast enough for the problem it is solving. Chasing seconds would not improve the user outcome if it reduced precision.
  • This matters for PMs because many product teams still celebrate shipping speed in the abstract. The better question is whether faster delivery improves the real user outcome, or just increases the chance of an avoidable mistake.

Attention is not priority

Dara Treseder offers one of the strongest prioritization rules in the episode: because something demands your attention does not mean it deserves your energy.

She and William use a few mechanisms to make that practical:

  • Glass balls are commitments they do not want to break, such as marriage, children's well-being, health, and financial stability.
  • Rubber balls are important but droppable commitments. Missing one hurts less and does not permanently damage the system.
  • They also talk about a family NPS made up of marriage NPS and kids NPS. The underlying idea is that important relationships need active measurement and renewal, not passive assumption.

For PMs, the transfer is clear. Not every loud request belongs in the same priority tier. Some commitments are foundational and breaking them damages trust. Others are recoverable.

Technology should preserve humanity

Baratunde Thurston does not argue for rejecting technology. He argues for using it in a way that does not turn you into a machine.

His test is whether a tool helps or harms your relationship with:

  • yourself
  • other human beings
  • the rest of the family of life

This is useful for PMs because product work often rewards constant reaction. Baratunde's point is that pace changes perception and behavior. If your tools reduce reflection, empathy, or contact with reality, they may improve throughput while degrading judgment.

Mental Models & Frameworks

Double pendulum thinking

Use DJ Patil's double-pendulum example as a planning model.

  • A system can follow understandable rules and still become hard to predict after a short time.
  • The key question is: what is the forecast horizon for this decision?
  • Use it when setting roadmap confidence. A platform migration may support longer-range planning. An AI feature in a rapidly changing market may need weekly reassessment.

Pioneer roadmap model

Hayden Brown's pioneer idea is a roadmap model for volatile environments.

  • Direction: define the horizon you are moving toward.
  • Learning loop: expect to adjust the route as you learn.
  • Progress check: keep measuring whether you are moving meaningfully toward the destination.
  • Use it when the market is changing too fast for step-by-step certainty, but not so fast that you can abandon strategy entirely.

Glass balls, rubber balls

Dara and William's prioritization model separates commitments by failure cost.

  • Glass balls: if dropped, they break trust or cause lasting damage.
  • Rubber balls: if dropped, they bounce back.
  • The model is useful because it forces explicit trade-offs. Teams often say everything is critical until a conflict appears. This framework makes the real hierarchy visible before the conflict.

Relationship test for technology

Baratunde's adoption filter is simple and human-centered.

Before adding a tool, channel, or AI behavior, ask whether it helps or hurts your relationship with:

  • your own judgment
  • other people you work with
  • the broader world you are meant to serve

Use it when a tool clearly increases speed but may reduce reflection, empathy, or attention quality.

Family NPS as feedback

Dara's family NPS is a reminder that important systems need feedback loops, even personal ones.

  • The concept is not literally about sending surveys to your children.
  • The useful idea is that trust and satisfaction degrade unless you notice them on purpose.
  • For PMs, the translation is that team health, stakeholder trust, and cross-functional relationships should be treated as measurable operating inputs, not vague background conditions.

Decision Principles

Principle: Plan to forecast limits

  • When: a team is building in a market or technical area where the environment changes faster than the roadmap.
  • Why: DJ Patil's point is that the useful question is how long prediction remains reliable. Once the forecast horizon is short, detailed long-range plans become theater rather than guidance.

Principle: Govern data before delegation

  • When: a team wants AI to answer business questions or generate insights for decisions.
  • Why: Hayden Brown says Upwork's crystal ball works because the data is structured and governed. Fast answers are only useful when the underlying information is trusted.

Principle: Match oversight to consequences

  • When: a product or workflow can cause real harm, panic, or trust loss if it is wrong.
  • Why: John Mills's public-safety example shows that human review should rise with the cost of error. Autopilot is useful, but the pilot still has to watch it.

Trade-offs & Nuance

Agility versus predictability

Hayden Brown argues that large companies, including public companies, need more startup-style agility again. The nuance is that predictability does not disappear.

  • Investors, teams, and customers still need commitments.
  • The shift is away from false precision, not away from accountability.
  • A PM still needs KPIs, roadmaps, and plans. The change is that those tools should express direction and learning speed, not pretend the route will stay fixed.

Automation versus accountability

Raina Kumra and John Mills both embrace AI, but neither treats it as self-justifying.

  • More automation can increase range, speed, and coverage.
  • More automation can also multiply bad judgment at scale.
  • That is why Raina warns against putting low-quality AI output into the world and why John insists on a human watching the autopilot. Delegation scales output. It does not remove ownership.

Speed versus precision

John Mills's point is not anti-speed. It is pro-fit-for-purpose speed.

  • In some products, seconds matter.
  • In other products, minutes are fast enough, and extra haste only raises error rates.
  • The trade-off for PMs is between response time and confidence. The right balance depends on what failure costs the user, not what looks impressive inside the company.

Disconnection versus participation

Baratunde Thurston does not take a purity stance on unplugging.

  • Fully opting out of technology is unrealistic for most people doing modern work.
  • Fully submitting to machine pace is also damaging.
  • The useful middle ground is deliberate participation: use the tools, enjoy the tools, but keep enough distance to notice when the tools are reshaping your values or attention.

Common Mistakes

Mistake: Confusing detail with certainty

Detailed plans can feel responsible while actually hiding uncertainty. In chaotic environments, a richly specified roadmap may only give the team false reassurance if the forecast horizon is short.

Mistake: Shipping unreviewed AI output

Raina Kumra's warning is blunt. AI can generate a lot of output, but some of it still has the judgment of a toddler. Sending raw AI analysis, copy, or recommendations onward without human review spreads errors faster, not insight faster.

Mistake: Letting urgency hijack energy

Dara Treseder's point is that urgency is a poor proxy for importance. Teams get trapped when every incoming request is handled as a top priority instead of being tested against longer-term values and real failure costs.

Practical Application

Score your forecast ranges

For each major initiative on the current roadmap, write down how long the underlying assumptions stay trustworthy: a quarter, a month, a week, or only days.

  • Review quarterly bets with longer horizons in a lighter cadence.
  • Review short-horizon bets much more often.
  • Use the shortest forecast range, not the loudest stakeholder, to decide planning cadence.

Build a trusted data layer

Create a governed source of truth that AI can query safely, similar to Upwork's crystal ball.

  • Define which metrics, tables, and customer signals are authoritative.
  • Set permissions and naming clearly enough that non-analysts can query them.
  • Only then add natural-language access. Fast AI answers are valuable only if the team trusts the data underneath.

Insert precision checkpoints

For workflows where wrong answers can harm users, reputation, or operations, add explicit human verification steps.

  • Mark which outputs AI can draft, which outputs humans must approve, and which outputs should never be automated.
  • Use stricter review for anything customer-facing, alert-like, or high-risk.
  • Borrow John Mills's test: if this output were wrong, what would it cost the user?

Classify glass and rubber

Run a prioritization exercise with the team using Dara and William's language.

  • List the commitments that are glass for the quarter, such as customer trust, a critical launch date, or a reliability target.
  • List the commitments that are rubber, such as nice-to-have experiments or lower-stakes requests.
  • When conflict hits, cut or delay rubber items first instead of debating priorities from scratch.

Review technology through relationships

When adopting a new tool or AI workflow, ask Baratunde Thurston's three-part question in writing.

  • Does this tool improve or weaken my team's relationship with its own judgment?
  • Does this tool improve or weaken collaboration with other people?
  • Does this tool improve or weaken connection to the real users and real world the product serves?

A tool that boosts throughput but weakens all three is probably a bad trade.

Questions to Consider

  • Which current product decisions truly have a quarter-long forecast horizon, and which decisions only justify week-by-week planning because the underlying market or technology is moving too fast?
  • Where is your team using AI without a structured and governed source of truth, which makes the output faster but not necessarily more reliable?
  • What on your roadmap is a genuine glass ball that would break trust if dropped, and what is only a rubber ball that can bounce back later?
  • In which workflow are you celebrating speed even though the real user need is precision, as in John Mills's example that minutes can matter more than seconds?
  • Which requests are consuming attention right now even though, in Dara Treseder's framing, they do not deserve your energy over a meaningful time horizon?

Bottom Line

This episode's core lesson is that chaos is not a reason to stop planning. It is a reason to plan differently: shorten the forecast horizon when prediction weakens, use AI to expand what humans can see, and protect judgment, precision, and trust as the non-delegable work. A PM who can do that turns volatility from a source of panic into a source of advantage.

Case Studies Mentioned

Upwork's crystal ball

Upwork built an internal AI tool called crystal ball that sits on top of the company's structured, governed data. Hayden Brown says it democratizes not just access to information but access to insight, letting one person answer questions in hours that previously took analysts or teams weeks. The product lesson is that AI becomes much more useful when it is paired with trusted internal data rather than asked to reason from scattered sources.

Watch Duty's alert workflow

John Mills describes Watch Duty as a signals-intelligence organization operating in public safety. AI helps the team sift large volumes of data, but humans still interpret and verify what gets surfaced. He gives a concrete failure mode from the broader category: if a system broadcasts every signal without interpretation, it can mistake training exercises for real fires and create dangerous false alarms.

Treseder family operating system

Dara Treseder and William run career, marriage, parenting, and service commitments with explicit operating rules: glass balls versus rubber balls, shared journaling, promises that are kept, and a family NPS that asks whether the marriage and kids still feel chosen and supported. The product lesson is not to treat home like a startup. It is that durable systems survive pressure when their priorities, feedback loops, and trust signals are explicit.

Notable Quotes

"There is no keeping up. There is only keeping calm." (Raina Kumra)

"Move fast and break things is a really dangerous proposition. You really have to be precise." (John Mills)

"Because it's demanding your attention does not mean it deserves your energy." (Dara Treseder)

"I want you to wake up still wanting to choose me." (William)

AI PM course

Everyone hears the same episodes.
Few can do what they describe.

Start for free