Mike Cannon-Brooks, co-founder and CEO of Atlassian, joins Alex Rampell and the host to discuss the AI transformation of software and the so-called "SaaS apocalypse." The conversation explores how AI is fundamentally changing software from passive databases into active systems that can perform work.

The discussion covers three distinct categories of SaaS companies facing different levels of AI disruption, pricing psychology insights from Predictably Irrational, and why some fears about AI replacement may be overblown. Cannon-Brooks shares how Atlassian is adapting by building AI agents into existing workflows while maintaining customer trust.

Key topics include the evolution from filing cabinets to databases to AI-powered systems, the challenges of consumption-based pricing, and the critical design problems of human-AI collaboration in enterprise software.

From Filing Cabinets to AI: The Evolution of Software

"The whole history of software from 1960 until 2022 was you would take a filing cabinet and you turn it into a database," starting with IBM and American Airlines' Sabre Systems in 1960 - Mike

Traditional software digitized storage but didn't increase efficiency dramatically - you still needed CISOs, IT provisioning, and human retrieval of information

"The cool thing about everything that's happening in AI land is that the filing cabinet can do work" - AI enables software to accomplish tasks independently rather than just store data - Mike

Three Categories of SaaS Companies in the AI Transition

Category 1: Seats tied to work AI can replace - companies like Zendesk face existential risk as customer service agents become unnecessary

Category 2: Seats as pricing proxies unrelated to work output - Workday charges per employee but employees don't use it to produce outcomes, making it safer from AI disruption

Category 3: Hybrid models like Adobe where seat reduction impact varies - not as stark as the other categories

"Public markets couldn't tell the difference between the three" types, causing broad SaaS selloffs regardless of actual AI vulnerability - Alex

The Psychology of Software Pricing and Fairness

Predictably Irrational by Dan Ariely explains why per-seat pricing "feels fair" even when marginal costs are near zero - customers accept paying more for more seats intuitively

Alex gave Predictably Irrational to product managers to understand pricing psychology, citing the locksmith example where customers pay more for incompetent but time-consuming work

"Humans are kind of capable and willing to pay for incompetence" when it feels like more effort was expended - Alex referencing the book's core insight

Consumption-based pricing faces customer resistance due to unpredictability and lack of control, unlike controllable usage like AWS storage

Vibe Coding Reality vs Hype

"The idea I would vibe code my own workday and then run it is terrifying" - Mike dismisses replacement fears while seeing extensibility value

David Ricardo's 1817 theory of comparative advantage applies - businesses should focus on core competencies rather than rebuilding software tools

Software contains decades of embedded edge cases and business rules that aren't easily replicated through vibe coding

Real value comes from AI extensibility - building custom applications on top of existing platforms for specific use cases like "Miami team conference room booking"

Businesses as Process Collections, Not Filing Systems

"Businesses are a set of processes" rather than static systems of record - Mike challenges the database-centric view of enterprise software

Input-constrained processes have fixed work volumes (customer service tickets, legal contracts) where efficiency gains reduce costs

Output-constrained processes involve unlimited creative work (marketing, software development) where efficiency gains increase output rather than reduce input

"I have 10,000 people who walk into buildings every day and bring their brains and walk out and take their brains with them" - knowledge work is fundamentally about process coordination - Mike

Atlassian's AI Integration Strategy

Building AI features into existing workflows first - ticket summarization in JIRA helps the fourth person on a case understand context without changing workflows

"Customers rave about" simple AI features that improve existing processes, even though they're "very unexciting from a magic point of view" - Mike

Developing agent frameworks that can integrate both Atlassian's agents and third-party agents like AgentForce or Gemini into workflows

"Most businesses will have three to five large-scale agent platforms running internally" requiring integration capabilities - Mike

The Trust and Design Challenge of AI

"Customer trust is really hard in these areas" - users fear AI doing things without understanding what happened or why - Mike

"Give people a chat box that can do unlimited power and they're like, tell me a dad joke" - unlimited capability leads to user paralysis without proper design

Human-agent loops require careful balance - too many confirmation steps frustrate users, too few destroy trust

Document creation with Rovo demonstrates paradigm shift - 75% traditional editing interface with 25% chat for AI commands, but requires user education