Atlassian CEO: Cheaper Software Will Produce More Apps and AI Agents

In an interview with Bloomberg (Insight with Haslinda Amin, October 13, 2026), Atlassian co-founder and CEO Mike Cannon-Brookes pushed back on a view popular with investors: that AI coding will devalue enterprise software. His thesis: when software gets cheaper and faster to write, organizations create more apps, workflows and agents. The need to connect and control all of it grows too. The company he founded in 2002 in Sydney sells exactly this kind of coordination tooling: Jira, Confluence, Trello.

The logic: cheaper production increases volume

The argument is familiar from economics: when the price per unit falls, consumption rises. Cannon-Brookes applies this to enterprise software. An internal tool that once took a quarter of development time can be built in days with AI assistants. Every team is tempted to build its own: procurement gets an app, support gets its own workflow, finance gets a separate agent.

As a result, a company ends up with not less software but noticeably more. Each new app requires access rights, data, and rules for handing off tasks between people and machines. Agents add another layer: they need context, limits on their authority, and a log of their actions.

Where Atlassian's interest comes in

The thesis has an obvious beneficiary. Since 2024, Atlassian has been developing its Rovo AI platform and positioning its products as the layer where teams and agents work on shared tasks. If the world fills up with thousands of small apps and agents, the value of the place where they meet will, in the company's view, only grow. This is a vendor's perspective, and it should be read with that interest in mind.

Still, the core idea holds up to a common-sense check. The spread of microservices did not eliminate the need for orchestration, monitoring and access management. It created an entire market of tools around them. Agents may follow the same trajectory.

What remains in question

The available announcement of the interview contains no figures: no revenue forecasts and no data on the number of agents at customers. The points below are therefore our assessments, not Cannon-Brookes's words.

• If companies learn to build internal tools themselves, some of them may be able to do without a vendor's off-the-shelf product.
• The per-seat pricing model will come under pressure if agents take over some of employees' tasks.
• Agent management will become a category of its own, and classic SaaS companies will not be the only ones competing for it: cloud providers and model developers will too.

Key points of the interview:

• Cheaper and faster development increases the number of apps, workflows and agents.
• Demand grows for integration, coordination and management of this ecosystem.
• Enterprise software is not disappearing but shifting toward a coordination layer.

What teams should do

If the thesis is correct, the problems will arrive before the benefits. To start, take an inventory: which internal apps and AI agents already exist, who is responsible for them, and what data they connect to. Then you need a single registry with an owner for each agent and a clear access policy. Budget a separate line for maintenance costs: writing an app has become cheap, but owning it is still expensive. Finally, when choosing vendors, it helps to ask how a product works with third-party agents and open interfaces. This is where the main friction of the coming years will land.