McKenzie Automation · A capability of McKenzie Robotics

From human task to autonomous robot.

McKenzie Automation is how the organization turns real work into robot capability — an eight-step framework for capturing what skilled people do, decomposing it into reusable instructions, testing it with humans before hardware, and automating it under natural-language control with guardrails.

8
Steps, capture → control
1
Skill library, reused across jobs
∞
Feedback · improve · retest
The method

How McKenzie automates work

We don't start from a spec — we start from a person doing the job. Each step converts tacit human skill into something a robot can execute and reason about, and every failure feeds an agile loop that improves the next pass.

Human work→ Data→ Instructions→ Automation→ Reasoning→ Natural-language control
From Human Task to Autonomous Robot — an 8-step framework for capturing, decomposing, testing, and automating work, shown as a circular pipeline from human work through data, instructions, automation, and reasoning to natural-language control.
© McKenzie Labs, Inc.
McKenzie Automation framework · © 2026 McKenzie Labs, Inc. — All rights reserved.
The 8 steps

Capture, decompose, test, automate

01 — CAPTURE

Capture real work

Bodycam and hand-level video of a skilled field worker doing the job naturally. Long shifts are broken into short, labelled task chunks.

02 — ANNOTATE

Annotate & script

Review footage, timestamp the key moments, and translate actions into instruction calls — surfacing the tacit moves that are hard to describe.

03 — PLAYBACK

Human playback test

Hand a second person only the instructions. If they can reproduce the task, the decomposition is working. If not, the instructions get sharper.

04 — TELEOPERATE

Teleoperate the robot

Run the same instructions while a human pilots the physical robot — changing one variable at a time to validate the hardware and the workflow together.

05 — SKILLS

Build the skill library

Automate reusable action verbs — grasp, place, press, open — and pair each with object metadata such as cup, handle, tray, or bag.

06 — PROGRAM

Program the task plan

Orchestrate the skill library with conditional logic and staged autonomy so the robot can run the full workflow: detect, pick, place, verify.

07 — REASON

Reason about goals

Convert the desired end state into formal, symbolic goals, then use inference and common-sense reasoning to plan the actions safely.

08 — CONTROL

Natural-language control

A person states the goal in plain language. The system parses it, builds a plan, checks safety constraints and consequences, then executes on the robot.

What we've learned

Key insights

Feedback → agile pipeline → improvements → retest. The loop is the product.

1

Most work compresses into a small set of repeatable instructions.

2

Instruction frequency follows a power law.

3

Object context matters as much as the action itself.

4

Failures feed an agile improvement loop.

5

Repeated execution creates a virtuous cycle of automation.

Built inside McKenzie Robotics

McKenzie Automation is the methodology behind McKenzie Robotics' humanoid and industrial platforms — machines built to take on work that is dull, dirty, or dangerous, operating alongside people in the real world.

Explore McKenzie Robotics →