How Automattic Runs a Global Distributed Company With P2 Blogs

Snapshot

Automattic is the company behind WordPress.com, WooCommerce, Tumblr, and Jetpack. (Source) They are also a principal commercial steward of WordPress, the open-source CMS used by 40.3% of websites tracked by W3Techs as accessed September 12, 2026. That measures WordPress usage, not Automattic’s own hosted footprint. (Source) Automattic’s About page lists 1,422 Automatticians, and How We Work lists 83 countries, as accessed September 12, 2026. (About, How We Work) Automattic remains fully distributed, with optional coworking and event spaces in New York and San Francisco. (Source)

Inference: Automattic’s scale makes it an instructive case for distributed companies above 100 people. Its model combines written coordination, periodic gatherings and optional shared spaces without requiring employees to share a daily workplace.

Core Philosophy

The Automattic Creed is a published statement of operating values, not just an HR artifact. The line that matters most for how the company runs: “I will communicate as much as possible, because it’s the oxygen of a distributed company.” (Source)

Other operating principles that shape daily behavior:

• “I am in a marathon, not a sprint.” — the company actively resists urgency culture.
• “I am more motivated by impact than money.” — stated signal for whom the company hires.
• “I will never pass up an opportunity to help out a colleague.” — collaborative default.

Automattic presents the Creed as operating values and says it seeks people who strive to live by them. The public Creed does not establish that every employee formally signs it. (Creed, Hiring guidance)

Communication Model

P2 blogs support written project discussions. Automattic’s current published toolset includes P2-themed WordPress.com blogs, Linear, private chat rooms and Slack. P2 sits alongside these tools rather than holding every project artifact. (Source)

Explicit communication expectations are published publicly:

1. Communicate often and publicly. Not just occasionally. “It is impossible to over-communicate at a8c.” Be active on P2/GitHub/Slack wherever your project lives.
2. Check notifications at least twice daily. P2s, GitHub, Slack — you’re responsible for staying current. “Never miss a ping.”
3. Set timeframes for every commitment. “Communicate the moment you know you’ll miss a commitment. When you have to be asked twice, it signals you aren’t reliable.”
4. Provide high-level summaries. Clarity about what you’re working on and how it is going, not just status updates.

(Source)

Automattic’s published account emphasizes project communication through shared tools. P2 is part of a wider documented toolset. (Source)

Support handoffs carry the next action. An October 16 WordPress.com team account describes Zendesk transfer notes recording scope, gathered context, confirmed and open facts, attempted steps, and what to do next. Specialist guilds join through Slack pings or routing macros while preserving the customer correspondence. AI assists summaries, diagnosis and internal-note drafts; Happiness Engineers retain judgment and responsibility for the final reply and handover. (Source)

Planning and Cadence

Automattic’s published development account describes feature-usage metrics and a one-button WordPress.com deployment system. The stated 60–80 daily deployments are a company claim, not an independently measured current rate. (Source)

A dated organizational frame, not a verified OKR process. Matt Mullenweg’s January 2024 reorganization account describes three strategic orientations for divisions. It does not document division-level OKRs. (Source)

Prototypes can become the discussion artifact. In an October 15, 2026 reflection, five staff from design, product, and program management describe finding alignment easier with working prototypes. They also report seeing polished AI output that did not solve a real problem, leaving people to question both the output and whether the idea should be built. This is a group observation, not a replacement of Automattic’s company-wide planning cadence. (Source)

Decision-Making Model

Strategic changes have a named author. Matt Mullenweg described the January 2024 divisional reorganization himself. That account is evidence of a particular strategic intervention, not a complete current authority map. (Source)

Day-to-day initiative has published expectations. The expectations document encourages context gathering, timely commitments and timeboxed attempts to move past blockers. It does not establish a universal decision-approval sequence. (Source)

Inference: written project discussions can preserve decision context, but the sources reviewed do not establish that every strategic decision follows a public P2 deliberation process.

Org Structure

• The official About page lists 1,422 Automatticians as accessed September 12, 2026. (Source) In a separate historical event, Automattic announced an approximately 16% workforce reduction on April 2, 2025; that announcement did not publish an exact post-restructuring headcount. (Source)
• The official About page identifies a family of products. That is not, by itself, a current divisional reporting chart. (Source)
• Movement between teams is described publicly. How We Work includes an employee account of rotating across teams and products. (Source)

Tools and Stack

Automattic currently lists P2, Linear, private chat rooms and Slack for project work. The First Day guide also identifies MGS search, Learnomattic and role-dependent support tools. Beau Lebens’s March 2020 account documents GitHub and Zoom historically; their universal current use is not re-established here. (How We Work, First Day, 2020 account)

Rituals

Trial project before hire. Every candidate who passes interviews does a paid contract trial project lasting 2–6 weeks before getting a full-time offer. This is not an internship — it is a working engagement where candidate and company both evaluate fit. The trial is done remotely. (Source)

Support onboarding has a VIP variation. The First Day guide describes support training through learning material, ticket observation and customer interactions. VIP hires wait for their lead’s instructions for a different rotation. (Source)

Annual support is a published expectation. How We Work states a one-week annual support rotation across positions. The page does not report completion rates. (Source)

Team meetups, typically 5–7 days. Division and team meetups combine bonding and strategy. The annual travel expectation is a total time budget, not a published count of meetups. (Source)

In-person gatherings. Current published guidance emphasizes division and team meetups; it does not confirm a 2026 Grand Meetup. (Source) Optional shared spaces also host events: Matt Mullenweg described a December 2025 event at the NYC space inherited through the Tumblr acquisition. (Source)

AI

Hiring expects AI literacy, and the bar varies by role. The current How We Work page lists AI literacy and competency among what Automattic looks for, with expectations that differ by role. The page does not say when that sentence was added. (Source)

A training cohort built tools outside engineering. On February 25, 2026, Automattic published an account of 54 employees brought to New York for two weeks of AI training earlier that year. The projects included content workflows, bug-report triage, and WordPress tools, and the account says marketers, editors, support specialists, and project managers built functional tools themselves. The author’s closing claim of a five-to-ten-times impact has no disclosed metric. It is not a measured productivity result, and a selected cohort is not the whole company. (Source)

Documentation changes are drafted by automation and approved by a writer. A July 15, 2026 WooCommerce account describes an internal plugin that watches releases and requests, drafts documentation changes, scores them, and sends previews to a writer. Automated checks flag substantial deletion, broken formatting, placeholders, and new image links. By late June, 39 of 117 reviewed documents fell below its standards. A terminology automation accidentally inserted writing guidance into live pages. The work shifts toward source maintenance and approval. The post does not publish net hours saved. (Source)

Support reviews assistant interactions every day. On August 31, 2026, the interim Chief Customer Experience Officer wrote that Automattic reviews assistant interactions daily, with the stated focus on whether the customer was helped. He says a question that reaches someone who cannot answer it is a routing failure, and that documentation maintenance is mostly automated. This is an operating statement, not a customer-outcome study, and it does not show that AI reduced staffing. (Source)

A support team owns assistant improvement. An October 1, 2026 account names an AI Squad of Happiness Engineers who review Odie interactions for incorrect answers, recurring problems, tone, and clarity. It also describes selected domain-management tasks delegated to Odie after testing, review, and guardrails, with Happiness Engineers available when needed. The post does not specify a complete permission model or show that every support action is autonomous. (Source)

A review-request workflow gained limited automatic sending. An October 19 account describes plain-rule checks, recorded facts, judgment gates and confidence checks before Odie asks for a product review. After reviewing two months of confident flags, Automattic enabled bot-labeled emails for qualifying WooPayments tickets handled by a Happiness Engineer. Other cases retained human sending; broader rollout was an intention, not an established result. (Source)

Benchmark scores are checked against real conversations. Automattic’s October 14 self-test used 71 scripted chats, an external rubric, Claude scoring, and a second-judge sample. Missing human handoff inflated logged-out Odie’s automation score. Manual WooCommerce-chat review caught problems the benchmark missed. Happiness Engineers and engineers planned fixes and retesting. Changed scripts, a small sample, and missing cases limit comparison with other vendors. (Source)

Design guidance is tested in the work. An October 27 account describes a design-brief skill, agent instructions that flag unresolved decisions, and human review. A two-designer pilot iterated through feedback. Paired agent builds were scored by separate agents; when both groups received the unresolved product decisions up front, the violation gap nearly disappeared. This is an early self-run pilot, not proof that instructions guarantee better products. (Source)

What They Do Well

Inference — operational lessons from the documented practices, not measured comparative outcomes.

• Written records support later reference. Preserve useful decision context where future teammates can find it.
• Support work creates direct user contact. Design an appropriate rotation for each product and role.
• Contract trials expose real collaboration. They provide evidence beyond an interview without proving a universally superior hiring method.
• Gatherings have an explicit time budget. Treat travel as an operating choice to evaluate, not proof that every distributed culture requires the same investment.

Tradeoffs and Weaknesses

A historical governance event affected staffing. Mullenweg reported that 159 people accepted the October 3, 2024 Alignment Offer amid the WP Engine dispute. This is his account of the offer and departures; it does not establish comparative risks of distributed versus colocated culture. (Primary account)

Inference: dependence on a founder’s strategic judgment is worth examining separately from communication tools. Written discussion does not eliminate governance risk.

Communication overhead at four-digit headcount. Inference: written project discussions across many products can create a substantial reading load. Filtering and curation become part of the coordination work.

Communication volume can obscure judgment. Inference: frequent public updates could become performative if evaluated by activity rather than usefulness. This is a risk to assess, not a measured Automattic outcome.

What Founders Can Copy

Inference — transfer guidance from this case, not validated headcount thresholds or outcome guarantees.

1. Use a paid work trial where appropriate, and evaluate the collaboration as well as the output.
2. Define product- and role-appropriate customer contact.
3. Publish behavioral expectations that explain commitments and communication.
4. Specify where durable project context belongs across the toolset.
5. Budget in-person time deliberately, then assess its value for your team.

Where This Model Breaks

Inference — risks to investigate when transferring this model, not observed failure thresholds.

• Governance relies too heavily on one person. Examine strategic authority and disagreement mechanisms independently of workplace location.
• Written volume exceeds attention. Define what people must follow and how important context is summarized.
• Activity substitutes for clarity. Review whether communication helps decisions rather than merely demonstrating participation.

Related Practices

• Async Communication — the foundational practice enabling P2-based work
• Remote Onboarding — paid trials and support onboarding
• Documentation Systems — P2s only work if people write clearly and consistently
• Retreats — division and team gatherings

Related Frameworks

• Sync vs Async Decision Matrix — when to use Slack vs P2 vs Zoom

Sources

1. How We Work — Automattic: https://automattic.com/how-we-work/
2. The Automattic Creed: https://automattic.com/creed
3. Expectations — Automattic: https://automattic.com/expectations/
4. First Day at Automattic: https://automattic.com/fieldguide/welcome-to-your-first-day/
5. Automattic’s Big Re-Org (Matt Mullenweg): https://ma.tt/2024/01/automattics-big-re-org/
6. 159 employees leaving Automattic (TechCrunch, Oct 2024): https://techcrunch.com/2024/10/04/159-employees-are-leaving-automattic-as-ceos-fight-with-wp-engine-escalates/
7. Automattic Alignment — Matt Mullenweg (October 3, 2024; primary reported offer/departures): https://ma.tt/2024/10/alignment/
8. W3Techs WordPress usage statistics (40.3% of websites, accessed September 12, 2026; WordPress-the-CMS, not Automattic-the-company): https://w3techs.com/technologies/details/cm-wordpress
9. Automattic restructuring announcement (April 2025): https://automattic.com/2025/04/02/restructuring-announcement/
10. About Automattic (current team count and product family, accessed September 12, 2026): https://automattic.com/about/
11. Automattic Spaces (optional coworking and event spaces): https://automattic.com/spaces/
12. Aldeas — Matt Mullenweg (December 11, 2025): https://ma.tt/2025/12/aldeas/
13. What Automattic’s AI Enablement Training Means for WordPress (February 25, 2026): https://automattic.com/2026/02/25/ai-enablement-wordpress/
14. WooCommerce, AI documentation project (July 15, 2026): https://woocommerce.com/posts/ai-documentation-project/
15. Deflection is Not a Support Strategy (August 31, 2026): https://automattic.com/2026/08/31/customer-support-ai-deflection/
16. Meet Odie — Happiness Everywhere (October 1, 2026): https://happinessengineer.blog/2026/10/01/meet-odie/
17. Benchmarking our own AI assistant — Happiness Everywhere (October 14, 2026): https://happinessengineer.blog/2026/10/14/benchmarking-our-own-ai-assistant/
18. When building is cheap, judgment is the job — Automattic Design (October 15, 2026): https://automattic.design/2026/10/15/when-building-is-cheap-judgment-is-the-job/
19. How AI helps us ask for reviews at the right time — Happiness Everywhere (October 19, 2026): https://happinessengineer.blog/2026/10/19/how-ai-helps-us-ask-for-reviews-at-the-right-time/
20. Working towards higher-quality agentic experiences — Automattic Design (October 27, 2026): https://automattic.design/2026/10/27/working-towards-higher-quality-agentic-experiences/
21. The internal collaboration behind a single reply — Happiness Everywhere (October 16, 2026): https://happinessengineer.blog/2026/10/16/the-internal-collaboration-behind-a-single-reply/

Inferences

• Sustaining written project discussions at four-digit headcount likely requires established norms and curation. Adopting the same software does not establish those norms; this is a transfer hypothesis, not a verified chronology of when Automattic introduced each practice.
• Support onboarding can connect functions to customer problems, with a different rotation for VIP hires. The sources do not compare its effect with analytics or other companies’ practices.
• The 2024 offer and departures make governance worth examining. They do not show whether distributed employees detect disagreement earlier or later than colocated employees.
• Automattic’s AI accounts move work toward review: a writer approves documentation drafts, and support reviews assistant interactions every day. The training cohort shows people outside engineering can build tools. None of these accounts measure a staffing reduction or a productivity multiple.

Work with Alex

If your company is struggling to maintain alignment and culture clarity as you scale beyond 50 people without a physical office, Alex helps leadership teams build the operating systems that prevent this kind of fragmentation.