A Playbook for More Human Workplaces
By Clay Parker Jones
Most "people problems" at work are really org design problems: tiny habits, incentives, and structures that quietly steer good people toward bad outcomes. Clay Parker Jones (who leads Organizational Design & Development at Airbnb, and previously worked at Undercurrent, R/GA, Black Glass, and co-founded August) published Hidden Patterns (Matt Holt Books / BenBella, March 10, 2026, 256 pages) as a field guide to those invisible systems. The book is not a cover-to-cover narrative. It is a cookbook of 75 remixable patterns, inspired by Christopher Alexander's A Pattern Language, that teams can mix and adapt. The working claim is blunt: work isn't broken, it is waiting to be better designed, and borrowing someone else's "best practice" usually falls apart because what works in one place rarely matches what another place actually needs.
Parker Jones wants you to give up on copying peer organizations. Best practices travel easily in a slide deck and then fail on contact with local incentives, rituals, and power. What travels better is a smaller unit: a pattern, a proven response to a recurring workplace problem that you adapt to local taste. The book treats each pattern like a chess opening or a mirepoix: humble, repeatable, and unfinished until your team cooks with it.
Parker Jones's blunt diagnosis of modern work: there aren't enough real teams. What most companies call a team is a co-acting group. People meet, then scatter back to work they each do alone. A real team keeps working on the same thing between meetings. It has a shared purpose, a decision it owns, or a number it watches together. Great work experiences, in his telling, almost always happen inside that kind of relationship, not inside a recurring status call.
The book's architecture comes from architecture. Alexander and colleagues distilled fieldwork into remixable design principles for the built world. Parker Jones does the same for organizations: 75 numbered patterns across five layers, from foundations (true purpose, rule of law, safety) through structuring, direction, practice, and learning, ending in space. He has said writing it felt more like composing software than telling a story, because the pieces are meant to compose into each other. A goal index and related-pattern suggestions turn the book into a choose-your-own-adventure rather than a sermon.
The origin story is a 2013 Undercurrent meeting with the CMO of a major hospitality company. Parker Jones pitched a mobile strategy; the CMO opened a binder of other consultants' decks and pointed to the same idea already gathering dust. The problem wasn't missing ideas. It was the systems underneath them. He started calling that the invisible architecture of work: how influence, information, and decisions actually move, which org charts almost never show. In the Network of Teams pattern he uses Deepwater Horizon as the hard case: hierarchy designed for stable conditions can sanitize, delay, or lose frontline warnings on the way up.
Most organizations still split strategy at the top from execution below. Agile rituals get layered on without moving power, so teams feel "empowered" while every real call still escalates. Expanded Available Power means explicitly vesting authority through charters, governance, and documented decision rights so people closest to the work can act. Declaring empowerment without transferring it lands as theater. Repeated escalation is the diagnostic: the organization has not actually expanded who can decide.
Default consensus asks everyone to actively support a decision before it moves. In practice that produces pre-meetings, political horse-trading, and a final meeting that is comfortable theater. Consent & Consensus, rooted in sociocratic practice, asks a different question: do you have a reasoned objection? "I can live with this" is enough to move, while still surfacing real harm. One lightweight move: write the decision where everyone can see it, then ask whether it will cause harm. That shifts the room from politicians to scientists.
Everything in the organization should carry an expiration date. Dissolvability treats teams, roles, and structures as temporary solutions to current problems, not monuments to defend. When the mission is done, the team disbands. When a structure stops creating value, it gets reconfigured. That is normal in the rest of life and strangely rare at work. Modular, loosely coupled systems make this possible: you can shut one temporary team down without collapsing everything around it.
Active Steering replaces annual bets with shorter, data-informed decision cycles. You wouldn't steer a ship once a year; Parker Jones argues organizations shouldn't either. Iterative Shipping is the companion: get smaller pieces of work into the world before they feel finished, so feedback arrives while you can still use it. For teams stuck waiting for a giant transformation, he offers a path: write the biggest possible change, then work backward to what you could try tomorrow with no change at all. Meeting Roles (facilitator, notetaker, timekeeper, rotated by skill rather than seniority) and a shared Logbook (who owns which decisions, which teams exist and why) keep the day-to-day architecture searchable instead of trapped in someone's head.
Pacific Glazing Hub only works if the people building it keep working on the same thing between meetings. Give the Hub crew (Steve, Travis, whoever owns the next module) a written team charter: mission, membership, decision rights, and what "done" looks like for this quarter. If the group only exists as a recurring status call, it is co-acting. Name the shared outcome (one estimating slice, one field workflow) and keep working it between the calls.
Handing Travis a Hub module isn't empowerment until decision rights move with the work. Write which calls he owns (scope tradeoffs inside the module, ship/no-ship for the slice, who to pull in) and which calls still sit with Steve. Expanded Available Power is the Deming move in org-design clothes: fix the system of authority, not the person's motivation. If every choice still escalates, the onboarding failed even if the code shipped.
Don't wait for every stakeholder to love the submittal tracker or the glass size calculator. Write the decision ("ship this checklist vocabulary to one real architectural spec") where the room can see it, then ask for reasoned objections, not universal enthusiasm. Consent gets a useful tool into the field. Consensus theater keeps the binder of good ideas gathering dust.
Treat each AI takeoff trial as temporary by design: a named experiment, a short window, a clear kill or keep criteria. When the learning is done, dissolve the squad or the ritual without mourning it. That matches how PGC already runs small software tests, and it keeps failed trials from hardening into permanent process nobody remembers requesting.
Hub knowledge trapped in Steve's head is the Logbook problem at company scale. Keep a short, searchable record of which modules exist, who owns which decisions, and what was tried. In reviews, rotate facilitator, notetaker, and timekeeper by skill, not title. The meeting becomes a service to the work instead of a performance for the most senior person in the room.
Stop copying other companies' best practices; spot the hidden patterns that shape how work actually moves, then adapt small, remixable fixes so teams, decisions, and power line up with the real work.
[Best practices] inevitably fall apart, because what makes something work in one place rarely matches exactly what's needed somewhere else.
A pattern, in these pages, is a proven response to a recurrent problem in modern work. It is not a silver bullet, a full operating system, or a promise of enlightenment. Think of it more like a chess opening or a mirepoix: humble, repeatable, waiting for you to adapt it to local taste.
When we empower one group, we need to take power away from where it already exists, and give it to those people that need it.
Personal note for Steve: This one sits next to Deming and Tiny Experiments without competing with them. Deming says the system is the problem. Tiny Experiments says run small tests. Hidden Patterns gives you the vocabulary for redesigning the system in bite-sized moves: real teams instead of co-acting groups, power that actually transfers when you onboard someone like Travis, consent instead of ship-by-committee, and experiments that dissolve when the learning is done. Hub is already a platform for that kind of design. The next useful move isn't another tool. It's a one-page charter for the next Hub slice, with decision rights written down, and a short logbook so the architecture isn't only in your head.