Big Technology Podcast · · 24 min

How Humans Will Work Alongside AI Agents — With Mike Cannon-Brookes

With Mike Cannon-Brookes

Summary

Atlassian CEO Mike Cannon-Brookes, on the Big Technology Podcast, argued that AI will reshape rather than mostly eliminate work, using the bank branch example to show how roles like developers, designers, marketers and finance staff will change while many jobs persist, with the caveat that five-year outcomes are unknowable. On the AI model landscape, he rejected the idea that frontier models are commoditising, saying model differences in strengths and cost matter, and that Atlassian's advantage lies in a harness that blends models to route each task to the best one. He dismissed fears that model labs such as Anthropic's Claude Code and Claude CoWork will crush SaaS vendors, arguing that focus and expertise limit labs, that vendors win through design, compliance and experience investment, and that frontier progress is available to all vendors. He also pushed back on the notion of 'headless' software, arguing that command-line and agent-only access cannot replace graphical interfaces, while Atlassian's teamwork graph, an enterprise context graph built over eight years, improves agent answers, lowers token costs and compounds through learning from agent sessions. He concluded that enterprise AI productivity gains remain smaller than hoped, mainly because design, product interfaces and people's ways of thinking about work have not yet caught up with the technology.

Key takeaways

  • Mike Cannon-Brookes on AI & jobs: AI will net create more jobs than it displaces; most change is job redesign, not elimination.
  • Mike Cannon-Brookes on Enterprise AI adoption & AI software: Model progress isn't reaching company output yet; design, experience and talent are the bottleneck.
  • Mike Cannon-Brookes on Frontier AI labs & model race: Models won't commoditise; they keep different strengths and cost profiles, and the harness matters more.
  • Mike Cannon-Brookes on Software & SaaS: Labs and vibe-coding won't easily erase software vendors; execution, compliance and design are hard to copy.
  • Mike Cannon-Brookes on Software & SaaS: Going headless is a mistake for software companies; they must keep investing in user interfaces.

What was said, by company and theme

5 views from 1 speaker. Every quote is checked against the episode's transcript (how).

Software & SaaS

Where everyone stands on Software & SaaS

  • Mike Cannon-Brookes · Atlassian

    Labs and vibe-coding won't easily erase software vendors; execution, compliance and design are hard to copy.

    Asked whether model makers moving into organisational software, such as Claude Code and Claude CoWork, could build competing products, he says he is not worried. He compares it to past fears that Microsoft, Google, Amazon and now model makers would crush everyone. Laws of physics apply to every company, including focus and expertise, so labs are unlikely to do all of this at large scale. Vendors compete by delivering for customers at scale with compliance and continual design and experience investment, which he calls an art and creativity problem rather than an algorithm problem. On vibe-coded copies of software, he says the cost of writing code has fallen but his own company's cost has also fallen. Progress from frontier and open-weight models is available to all vendors fairly evenly, so the playing field is level.

    “I don't know you can run your business with the expectation that they're just gonna go and do all these things because there are a lot of laws of physics that apply to every single company”at 18:32
  • Mike Cannon-Brookes · Atlassian

    Going headless is a mistake for software companies; they must keep investing in user interfaces.

    He says headless is brainless. If software companies believe most usage will go through a headless interface, so they can stop investing in design and experience, he calls that a path to death. Atlassian is a big investor in its CLI and MCP server, which is one of the most used MCP servers at the major labs, and agents interacting with software through them is effective. But he does not think people will sit at a command line and use all enterprise software, since it is inefficient. He cites a recent survey saying users disliked that way of working and found it unproductive, and points out that graphical interfaces replaced the terminal for a reason. He expects adaptive, dynamic UI driven by models, but says the idea that all software becomes a hidden database is naive.

    “I do think headless is brainless. ... I think that is a path to death. I think that is a very brainless move.”at 20:30

AI & jobs

Where everyone stands on AI & jobs

  • Mike Cannon-Brookes · Atlassian

    AI will net create more jobs than it displaces; most change is job redesign, not elimination.

    He thinks companies citing AI restructuring rather than AI replacement is probably more accurate. His view is that AI will net create far more jobs than it changes, though there will be a lot of job change. He uses bank branches as a parallel: in the 2000s people predicted branches would close with online and mobile banking, but branches have grown over the last decade in Australia and the US because they now do different, more customer-service-focused work. He expects software developers, designers, marketers and finance professionals to still exist but to work very differently. He also says the ratio between job families may shift, and that the useful question is how jobs change, not whether AI kills them. He adds a caveat that nobody knows what happens over five years.

    “my view is net net, it'll create far more jobs than it changes. There'll certainly be quite a lot of job change.”at 3:49

Enterprise AI adoption & AI software

Where everyone stands on Enterprise AI adoption & AI software

  • Mike Cannon-Brookes · Atlassian

    Model progress isn't reaching company output yet; design, experience and talent are the bottleneck.

    Atlassian is seeing efficiency gains from putting AI agents into business processes, but not as large as they would like in many workflows. Output quality is improving, and he sees the start of growth in creativity. He asks why, given all the model progress, those efficiencies are not yet flowing into company output and creativity. He attributes much of the answer to design and experience, meaning product interfaces must change as well as people's habits and their thinking about doing their jobs. He notes that changing how people think about their work and historical priors is hard, making it a talent challenge more than a technology challenge.

    “With all the model progress we've seen, why are we not seeing those efficiencies flow through into company outputs, company creativity? And I would say a lot of the answer is in the design and experience.”at 7:24

Frontier AI labs & model race

Where everyone stands on Frontier AI labs & model race

  • Mike Cannon-Brookes · Atlassian

    Models won't commoditise; they keep different strengths and cost profiles, and the harness matters more.

    Asked whether models are commoditising, he says he does not think they are, at least not in the sense that all models cost roughly the same and are interchangeable. He expects continued differences in flavour, strengths for different task types, and very different cost profiles. He says model progress is amazing, but model choice is less important than people make out. Atlassian's job is to harness progress, and jumping to the latest model is an expensive way to work that often gives slower, not better, results. Instead, Atlassian does 'model blending', deciding which model answers which question best, and he sees the harness running on top of the model as increasingly important for routing between models.

    “I don't think they're commoditizing maybe in that manner... I think they will continue to have different strengths, different types of tasks. I think they will continue to have vastly different cost profiles.”at 15:45

Topics

  • enterprise AI productivity gains
  • enterprise context graph
  • headless software
  • job displacement
  • model commoditization
  • model labs competing with software vendors

More from Big Technology Podcast

All Big Technology Podcast episodes

Ask your AI about this episode

Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.

Setup guide

A weekly email with the calls that mattered from the most important voices in investing. Unsubscribe any time. Privacy Policy

Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus