Adding a universal chat interface to existing products exposes how users naturally want to manipulate data. Atlassian tracked user chat prompts in whiteboards to identify recurring requests, converting them into dedicated buttons and automated workflows.
Features developed for end users should double as actionable skills for AI agents. Exposing these capabilities via protocols like MCP lets internal and external agent systems invoke the exact same core actions that human users execute across tools.
Rather than discarding legacy software, incumbent SaaS platforms should insert AI steps into established user flows. Atlassian found that embedding agents into Jira automations immediately drove high adoption because customers already relied on those structures.
Software should handle agents using identical mechanisms built for team members, such as assigning tickets, mentioning them, and providing shared context. Designing unique, isolated agent-specific workflows often creates unnecessary friction compared to unified team primitives.
Atlassian discovered that having cross-functional teams exclusively write code with AI initially accelerated output, but quickly stalled progress. Without dedicated product and design leaders steering strategy and unblocking decisions, development velocity degenerates into aimless spinning.
Avoiding junior hiring under the premise of AI efficiency is flawed because juniors are significantly more adept at experimenting with emerging tools. Pairing them with senior engineers creates an effective dynamic where juniors drive tool adoption while seniors enforce architectural quality control.
