How Atlassian builds real team connection across 13,000 distributed employees

Executive overview

Most companies assume culture is built in the office. Atlassian's research found the opposite: sporadic office attendance builds almost no lasting connection.

Their Team Anywhere model removes location as a variable entirely — employees work wherever they choose, and the company optimises every meeting for the person not in the room. Connection, productivity, and AI adoption are each treated as deliberate design problems, not side effects of co-location.

Real connection is built through intentional gatherings with a purpose, not proximity.

Team Anywhere: distributed-first, not remote-first

• Distributed-first means optimising for the person working remotely, even when others are in the office together
• Three in-office, one remote? The three go to separate rooms so everyone is on equal footing
• The model launched as a one-way-door decision — no coming back — which forced faster problem-solving
• Early unknowns (tax implications, internal mobility) were acknowledged openly rather than papered over with policy
• Collaboration zones mapped which time-zone overlaps (e.g. Sydney–California) were viable for joint work

Intentional togetherness (ITG)

• Intentional togetherness gatherings replace the assumption that office attendance builds bonds
• Teams convene once per quarter, cross-functionally, around a specific problem to solve
• Two to three days together on ideation, strategy, or team-building; then back to distributed work
• Connection from a purposeful sprint lasts two to three months — matching the quarterly cadence
• Format is flexible: each team chooses what works for the problem and people involved

Workday design for distributed teams

• Atlassian's internal research divides the day into three blocks:
  • Meeting blocks: 20–30% for mission-critical synchronous meetings
  • Deep work: 30–40% for solo focused work — writing, strategy, analysis
  • Collaboration windows: ~30% for time-zone-overlapping work with colleagues
• Without intentional structure, distributed workers default to back-to-back Zoom and accumulate fatigue
• Writing culture and async tools (Loom, Confluence pages) are foundational, not optional add-ons

No PowerPoint culture

• Presentations are banned — no slides, no keynote decks, regardless of seniority
• Instead: write thoughts on a page before the meeting; attendees read silently for 5–10 minutes at the start
• Comments accumulate in real time; the presenter reads the room before speaking
• Result: denser, more honest conversation — people critique ideas, not performances
• Writing enforces substance over style and creates a default-open single source of truth
• With AI, well-documented organisations will compound the advantage — richer data for AI to work with

Async communication replacing meetings

• Loom video messages have eliminated close to half a million hours of meetings across Atlassian
• Team meetings moved from weekly to needs-based (roughly fortnightly) with async decision-making between
• Decision workflow: record a Loom → drop in Slack → team watches independently → tick yes/no → leader decides within 24 hours
• Async removes posturing, body-language reading, and extrovert dominance — quieter contributors produce more articulate input
• Avani uses voice notes, Looms, and Confluence pages when travelling; synchronous meetings are the last resort

HR-built AI agents (no engineers needed)

• Atlassian's belief: AI adoption requires cultural transformation, not just tool rollout
• Three principles: create fast learning loops; build upskilling through tinkering; don't lead with cost-efficiency as the end goal
• Nora — onboarding agent built by the talent team in Atlassian's Drovo studio; 70% of new hires use it on day one; personalised by team and role
• Coco — helps managers navigate sensitive compensation conversations with context-aware guidance
• Angel (Avani's personal agent) — pulls from calendar, email, and Confluence to surface top three weekly priorities
• Non-technical HR staff built all agents using prompt iteration in a no-code studio — no engineering support required
• Bottom-up creation drives adoption; teams own what they built

Four stages of human–AI collaboration

• Stage 1: Using AI to summarise or rewrite a single document
• Stage 2: Enterprise search — pulling company context into a communication
• Stage 3: Deep research — asking AI to analyse data and surface recommendations
• Stage 4: Strategic sparring — running hypotheses, challenging AI output, building work products together
• Strategic users (stage 4) save ~105 minutes per week and produce 10x more creative output than stage 1–2 users
• Most people stall at stages 1–2; the goal is to accelerate the journey to stage 4

Predictions for the future of work

• The "where we work" debate will disappear — leaders will focus entirely on how teams work
• Async communication will become the foundation of modern teamwork, even in traditionally in-person businesses
• Human–AI collaboration will shift from occasional tool use to integral daily workflow — the organisations that get there first will compound the advantage