The Twenty Minute VCAtlassian CEO, Mike Cannon-Brookes on Why Everything is Overvalued & Are We in an AI Bubble
CHAPTERS
- 0:00 – 3:27
The “unreasonable” founder mindset: pushing past boundaries
Mike explains why Scott Farquhar calls him the “unreasonable man,” drawing on the George Bernard Shaw idea that progress depends on people who won’t accept constraints. He connects that mindset to frustration with how things work and a bias toward changing systems rather than tolerating them.
- •Progress often requires rejecting accepted boundaries
- •“Unreasonable” can be a productive trait when channeled well
- •Frustration with broken systems can be fuel for invention
- •Technology as a creative recombination of ideas (putting pieces together)
- 3:27 – 6:22
20 years of co-CEOing: what made the partnership work
Mike reflects on why Atlassian’s co-CEO structure succeeded for decades: equality in life stage, shared context, and the ability to balance each other. He highlights mutual respect, friendly competition, and genuine friendship as core stabilizers through company ups and downs.
- •Co-founder equality (age/life stage/naivety) created balance
- •Having a peer reduces loneliness and increases resilience
- •Believing the other is “better than you” raises the bar
- •Friendship and humor mattered as much as strategy
- 6:22 – 10:34
Co-CEO mechanics: swim lanes, overlap, and conflict resolution
He gives practical advice for other co-CEO pairs: be radically open about conflicts, maintain yin-yang balance, and define clear swim lanes while keeping enough overlap for shared goals. He also describes Atlassian’s early (and humorous) deadlock mechanism and the principle of not acting if you can’t convince your co-founder.
- •Radical openness prevents conflict from festering
- •Aim for ~60–80% overlap: enough alignment without duplication
- •Explicit swim lanes + shared context are both necessary
- •Rule of thumb: if you can’t convince your co-CEO, reconsider
- 10:34 – 14:22
Becoming solo CEO in the AI era: “bold moves” without the theatrics
Harry frames the last 18 months as “founder mode,” but Mike pushes back: many moves were planned earlier and there was no desire to “stamp authority” after Scott stepped back. He describes the current AI period as chaotic, creative, and unusually fast-moving—requiring action and adaptability.
- •Perceived boldness can be timing + visibility, not ego
- •AI era is non-linear: a Cambrian explosion of change
- •Companies must make choices and act amid uncertainty
- •10 years from now, this period will look ‘crazy’ in hindsight
- 14:22 – 16:24
Investor reality check: overvaluation, fragile business models, and uncertainty
Mike outlines why investing is difficult right now: many assets are overvalued, yet a few outliers will dwarf the losers. He questions today’s circular “revenue” narratives across model providers, cloud, and chips, and emphasizes that durable AI monetization models aren’t settled.
- •Most things are likely overvalued, but winners may be massively undervalued
- •AI value chain can mask losses as ‘revenue’ transfers
- •Durable business models in AI are not yet established
- •Be wary of anyone claiming certainty about where value will land
- 16:24 – 19:21
Atlassian’s AI playbook: flexible hunches, quarterly resets, and multi-model strategy
Mike explains how Atlassian navigates AI with directional hunches that get revisited frequently rather than fixed doctrines. A central bet: multiple foundation models will coexist, so Atlassian builds to be multi-model and focuses on rapid adoption and deployment rather than training its own foundation models.
- •Make hunches, but revisit them frequently (quarterly)
- •Assume multi-model competition and design for it
- •Don’t build foundation models—differentiate in product delivery
- •Core capability: test, integrate, and ship new models quickly
- 19:21 – 21:16
Why design (not just data) will define AI products and interfaces
Mike argues AI shifts software back into a “fundamental design” era: differentiation will come from how products feel and work, not only underlying models or datasets. He predicts the future won’t be a single chatbox interface; instead, UX will blend stable UI patterns with AI-driven customization to reduce “click-to-value.”
- •Design is undervalued but becomes the differentiator in abundant software
- •The world won’t end as a single chat interface; users need stability
- •AI can reduce friction: less clicking, more direct outcomes
- •Customized interfaces may adapt to roles and user intent
- 21:16 – 26:02
Do apps collapse into agents? Responding to the ‘CRUD apps’ narrative
Reacting to Satya Nadella’s claim that business apps may collapse in the agent era, Mike is skeptical of sweeping pronouncements. He notes enterprise software can often be approximated by email and spreadsheets, yet specialized tools persist because they fit purpose and scale; AI will reshape apps, but not necessarily eliminate them.
- •Big AI-era statements are easy because they’re hard to falsify quickly
- •Email + Excel can mimic many workflows, yet specialized tools still win
- •SaaS has proliferated into many niche apps; AI may reshape that landscape
- •Some apps will disappear or compress, but not all collapse into an agent
- 26:02 – 28:12
More software, not fewer developers: abundance, roadmaps, and quality loops
Mike contends AI will increase the amount of software created, and companies will likely employ more engineers, not fewer. Because software demand isn’t output-bound (roadmaps are effectively infinite), greater efficiency enables multiple iterations and higher quality rather than shrinking engineering teams.
- •Software creation will accelerate; demand for solutions remains massive
- •Companies’ product roadmaps don’t ‘end’—ideas are effectively unlimited
- •Efficiency enables rebuilding/iterating for quality (2–3 tries)
- •Expectation: more engineers and more technology in five years
- 28:12 – 31:34
Vibe coding and the rise of the ‘maker’ spectrum inside enterprises
Mike describes a broader creator economy inside companies: finance, marketing, and other functions will increasingly build tools for themselves using AI-assisted development. Atlassian aims to enable this spectrum—helping makers ship apps that are secure, compliant, and aligned with Atlassian’s design language—even if many apps are disposable.
- •‘Vibe coding’ expands creation beyond traditional engineers
- •Non-dev roles will build tools instead of filing IT tickets
- •Platform standards (security/compliance/design) become critical
- •More apps per customer, with acceptance that many are throwaway
- 31:34 – 38:18
Engineering talent in an AI world: grads, tools, and the real productivity constraints
Mike argues entry-level engineers won’t disappear; Atlassian is hiring more grads year over year. He notes “coding” is only a portion of developer work—search, comprehension, and debugging dominate—so productivity gains must extend beyond code generation, and AI will increase operational complexity as systems multiply.
- •Grads may be more productive and shift norms inside companies
- •AI tools vary: great for greenfield, weaker for large codebase refactors
- •Coding is often 10–50% of the job; search/debugging are huge
- •More software means more services and more operational complexity
- 38:18 – 46:56
Defensibility and margins in AI: switching costs, value capture, and pricing model shifts
Mike explains why switching costs are low now—AI products were built fast and models were broadly democratized—but may rise through workflow fit, familiarity, and delivered value. He also addresses margin skepticism: monetization and model cost dynamics are unsettled, value may shift across chips/power/apps, and SaaS pricing may evolve from per-seat toward blended consumption and outcome-based models.
- •AI’s rapid spread keeps switching costs low early on
- •Future defensibility may come from workflows, UX, and accumulated value
- •Margins are hard to judge before monetization stabilizes
- •Per-seat pricing may persist but blend with consumption/value-based models
- 46:56 – 57:51
Protecting Atlassian’s long-term edge: creativity and fighting the ‘entropy of ambition’
Asked what Atlassian must protect, Mike says “creativity”—the ability to keep creating through successive tech transitions rather than defending legacy advantages. He shares Tobi Lütke’s line that the founder’s job is to fight the entropy of ambition, and discusses sustaining energy through self-awareness, breaks, and love of the craft and the team.
- •Multi-decade companies survive by creating (and sometimes destroying)
- •Culture must support killing old things to build new ones
- •Founders must maintain ambition as organizations scale
- •Enjoying the people and the process is essential for longevity
- 57:51 – 1:03:45
Quick-fire insights: failures, parenting, strengths-as-weaknesses, and Atlassian in 2035
In the quick-fire round, Mike shares an early product flop (status updates in Confluence), what he’s updated his thinking on (AI adoption takes longer than demos suggest), and advice for new parents. He closes with a vision of Atlassian in 10 years as still “fighting,” attracting talent, and remaining vibrant beyond its founders.
- •Example flop: early Confluence status updates didn’t work
- •AI’s ‘magic demo’ vs real enterprise value gap is driven by workflow/data/security realities
- •Newborn parenting advice: relax—babies are hardy; enjoy the ride
- •2035 vision: still competitive, creating, and not a ‘zombie’ company
