Continuous delivery environments often struggle with brittle end-to-end (E2E) tests. When tests rely on fixed scripts and rigid UI selectors, minor interface updates frequently trigger false negatives, forcing engineering teams to spend significant time on maintenance.
Slack engineering has addressed this by shifting toward agentic testing. Instead of executing static sequences, AI-driven agents interpret high-level test objectives and interact with the application dynamically, improving resilience against UI and service changes.
In short
- •
Agentic testing replaces brittle, script-based E2E workflows with intent-based AI agents that adapt to UI changes in real time.
- •
This architecture reduces maintenance overhead by allowing tests to survive minor interface modifications that would break traditional static selectors.
- •
The trade-off involves increased complexity in the agentic layer, which must now handle planning, state evaluation, and dynamic action selection instead of simple linear execution.
From Scripted Sequences to Intent-Based Execution
Traditional E2E testing relies on predictable flows, stable selectors, and fixed API interactions. In fast-moving systems, these assumptions are often invalid, leading to high failure rates that do not reflect actual functional regressions.
Slack's approach shifts the responsibility of test execution to an AI agent. Engineers define a test objective rather than a sequence of steps. The agent interprets this intent, plans the necessary actions, and interacts with the application through UI or API surfaces.
Dynamic Adaptation and Validation
At each step of the workflow, the agent evaluates the current application state. If the UI structure changes or an element is relocated, the agent attempts to find an alternate path to complete the objective instead of failing immediately.
This dynamic capability allows the test suite to remain functional despite minor UI refactors. Once the agent completes the workflow, the final state is validated against assertions defined by the engineering team, ensuring that the core functionality remains intact.
Source
Slack Introduces Agent Driven End-to-End Testing to Improve Resilience in UI Test Automation
https://infoq.com/news/2026/07/slack-agentic-e2e-testing-ui
