The Last Manual Handoff: Redesigning End-to-End Testing in the Age of AI
summary
Argues that every testing box is automated but the arrows between them remain manual, shows Main Tester United — a context-aware chain of AI agents that carries requirement traceability left-to-right, auto-creates Jira tickets, test cases and executions, and files defects with logs, screenshots and repro steps. Cites a measured example: 36 minutes of execution occurs inside 65 days elapsed because of coordination delays, quantifies the handoff tax, and demos the approach working end-to-end.
AI-written summary · about
chapters (62)
- 0:00Start
- 1:50Session Opens - The Last Manual Handoff
- 2:20We Automated Every Task and End-to-End Testing Still Feels Manual
- 2:53The Gaps Between Tasks Were Never Touched
- 3:25Speaker Intro - Sachin Sharma, Senior Manager Quality Engineering at Adidas
- 5:58What We Genuinely Got Right: Test Execution Is Solved
- 7:32Why Local and Unit Testing Is the Easy Part
- 8:04Day One of E2E: "Is the Environment Up? Did SAP Receive the Data?"
- 9:07Every Pipeline Green and the Week Still Starts With That Thread
- 12:43The Mission Failed - and Both Components Were Correct
- 14:15Mapping It to the Testing Lifecycle: Requirements to Reporting
- 15:46The Boxes Are Automated; The Arrows Are Not
- 16:16What Engineers Actually Still Do Between Every Phase
- 16:46Deciding Whether a Failure Is Environment, Script or Real Defect
- 17:17Explaining That "Complete" Doesn't Mean "Pass"
- 20:22We're Doing the Opposite: Giving Testers AI, Changing Nothing Else
- 20:53AI-Enabled but Not AI-Native
- 21:55The Demonstration Effect: Wanting It Because Someone Showed It
- 22:25The Neighbour With the Bigger House and the Car
- 23:27See, Compare, Desire - How Most Organisations Adopt AI
- 23:58They Have Five Agents, So I Need Seven
- 24:29Capability Spreads Faster Than the Need For It
- 25:00The Aircraft With a Thousand Necessary Components
- 25:30The Mantra: Stop Adding, Start Removing
- 26:33The Handoff Tax: Time, Context, Energy, Ownership, Accuracy
- 27:04Before You Automate Anything New, Ask If It Adds a Handoff
- 27:34We Measure Execution Time and Never Measure Coordination Time
- 28:0536 Minutes of Execution Inside 65 Days of Elapsed Time
- 28:38The F1 Pit Stop: What If Each Crew Member Raised a Jira Ticket?
- 29:10They Engineered the Flow, Not Just the Tasks
- 29:41What If the Context Travelled With the Work?
- 30:44Context Flowing Left to Right Through Every Agent
- 31:14The Agent That Declines to Add a Test Case That Already Exists
- 31:44Requirements Traceability Reborn Inside the Agent Chain
- 32:15Reports With Context, Not Just Pie Charts
- 32:45Demo: Watch the Handoffs, Not the Agents
- 33:16A Front End So Non-Testers Can Run It Too
- 34:17Creating Jira Tickets, Test Executions and Cases in Sequence
- 34:47The Deliberately Failed Test and Its Root-Cause Report
- 35:17The Defect Filed With Everything Needed to Reproduce
- 35:48The Real Question: What If the First Context Was Wrong?
- 36:48The Bullwhip Effect Explained
- 37:19Seven Packets of Toilet Paper Becoming Seven Thousand
- 38:52The Same Amplification in Software Requirements
- 39:23A Message to Anyone Building Testing Tools
- 39:53One Wrong Interpretation and Every Downstream Agent Is Wrong
- 40:25Build for What Testers Do Between the Automated Steps
- 40:56Why So Many Test Case Generators Never Sell
- 41:57The Two Extremes: Full Human and Full Autonomy Both Fail
- 42:28The Bliss Point Borrowed From Food Product Development
- 42:58Finding Your Own Autonomy Sweet Spot
- 44:03Three Questions Every AI Testing Product Must Answer
- 44:34Who Owns the Context? Who Owns the Decisions?
- 45:05Who Owns the Consequences When the Machine Is Wrong?
- 45:36Two Kinds of Handoff: Intentional and Accidental
- 46:07Keep the Intentional Handoffs; Engineer Out the Accidental Ones
- 47:13Focus on the Arrows - SDLC Is No Longer Blocks
- 48:16The Next Revolution Is Between the Boxes, Not Inside Them
- 48:46Stop Asking What to Automate Next; Ask How to Make It Continuous
- 50:42Q&A: Best Practices for Enterprise AI Testing - Start With Tokenisation
- 52:14Q&A: The One Mistake QA Engineers Make With AI
- 54:16Q&A: The Hardest Part of E2E to Fully Automate
related talks
46mWhy Is Test Automation Still Manual and How Can We Make It Really Automated?
44mHow AI is Exposing the Existential Crisis of Software Testing
41mHow AI is transforming software testing
42mThe Power of Human and AI Collaboration in Software Testing
38mNext Generation Testing in the Age of Machines
39mRevolutionizing Test Automation: Generative AI To Empower Prod Mgmt, Accelerate SDLC
47mFull-Stack Test Automation
42mIt's Inevitable, Mr. Anderson: Machines, Humans, and the Future of Testing
