Gemba to Governed AI

00:00:00.000Every company has critical work that is more complicated than the process document says.

00:00:06.000Decisions live in people's heads. Exceptions live in email, screens, files, and experience.

00:00:13.000When that knowledge is missing, the business pays for it in waiting, rework, duplicate work, and quality risk.

00:00:21.000Capture Foundry captures how the work actually happens and turns it into evidence the organization can review.

00:00:30.000Let's follow one critical process from the expert who performs it all the way to approved, AI-ready standard work.

00:00:42.000Process Owner Owen Wright opens AAV Batch Review and Deviation Handling and selects Review Bioreactor Telemetry.

00:00:53.000He chooses Subject Matter Expert Avery Chen and requests recording, screenshots, notes, files, decisions, and exceptions.

00:01:03.000The email arrives with the exact process, exact step, and a button that opens the Capture Foundry desktop assignment.

00:01:14.000Avery does not draw a process map or write an SOP. She opens the assignment and simply does the work.

00:01:25.000Capture Foundry confirms consent, selected screen, requested evidence, and the exact map shape before recording begins.

00:01:36.000Avery opens the work screen, finds batch AAV-2026-0814, and checks pH against the approved range.

00:01:47.000The observed pH is low. She captures a screenshot, and the evidence count increases while the recording continues.

00:01:59.000She finds missing operator verification, adds the QA escalation note, marks sensitive information, pauses, resumes, and finishes the work.

00:02:12.000Capture Foundry organizes the recording, screenshots, notes, files, decisions, exceptions, and AI summary. Avery can edit what the AI suggests.

00:02:24.000She submits the capture back to the exact process shape. That knowledge is no longer trapped with one employee.

00:02:32.000QA Reviewer Maya Patel opens the same step, plays the recording, reviews screenshots, reads the note, and checks the decision explanation.

00:02:45.000Maya requests a change: add the required QA escalation rule when operator verification is missing.

00:02:57.000Owen receives the notification, updates the process map, adds the Operator Verification Complete decision, and creates version 1.1.

00:03:10.000The comparison shows the new decision, new exception path, new approval requirement, and the comment trail that caused the change.

00:03:22.000Approvals happen one at a time. QA approves. The plant manager approves. Compliance approves. The audit history updates.

00:03:35.000Recording the work captured the evidence. Review and approval turned it into trusted organizational knowledge.

00:03:47.000Now Capture Foundry creates the current-state value stream map from the approved process and the evidence already captured.

00:04:00.000The value stream builds left to right with process time, waiting, handoffs, rework, exceptions, systems, and evidence.

00:04:13.000This is where process knowledge becomes business value. The team can see where waiting and rework consume capacity.

00:04:25.000Gemba, Kaizen, structured problem solving, and optimization start from the actual current state instead of a blank whiteboard.

00:04:37.000The SOP is generated from approved version 1.1, with purpose, scope, responsibilities, procedure, exceptions, escalation, and evidence.

00:04:48.000Don't automate a bad process. First capture it, approve it, improve it, and only then decide what is safe to automate.

00:04:59.000Agent Readiness separates automatable work from human-assisted work and human-only decisions like final QA approval and batch release.

00:05:10.000The agent is tested against normal batches, low pH, missing signatures, conflicting data, and unauthorized release requests. Human-only actions are blocked.

00:05:21.000Before you cut it, change it, integrate it, or automate it, understand it.

00:05:28.000Capture the work. Approve the truth. Improve the process. Build the future.