Async Week Process Automation
Document work, let AI critique it, then automate the highest-value steps.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 98%
Async Week is a focused process-improvement cycle that turns everyday work into reusable automation. A person first records a recurring task in Loom rather than trying to reconstruct every nuance from memory. Loom's output becomes a standard operating procedure that both colleagues and language models can interpret. A reasoning model then examines the SOP, identifies inefficiencies, and proposes steps that could be improved or automated. The operator implements selected recommendations with a coding agent, conventional code, Zapier, or another no-code tool. Instead of expecting employees to optimize work while simultaneously delivering it, the organization reserves a week or recurring block for documentation and experimentation. The result is an increasingly codified operating system in which processes can be reviewed, taught, measured, and replicated at scale.
Origin
Steph Smith introduced Async Week on Marketing Against The Grain as a four-stage workflow combining Loom, generated SOPs, OpenAI's o1 reasoning model, and coding or no-code automation.
Core principles
- 01A process must be visible before it can be improved.
- 02Recorded work preserves nuances that written instructions often miss.
- 03Structured SOPs help both humans and language models understand a task.
- 04AI should identify improvements before tools automate them.
- 05Dedicated improvement time overcomes the pressure of routine work.
How to run it
- 1
Select a recurring process
Choose a task you perform frequently enough that reducing its time, errors, or coordination cost would matter. Define the task's starting inputs and expected output.
Pro tip Start with a process that is stable, repetitive, and currently dependent on one person's memory.
Watch out Do not begin with a process whose desired outcome is still unclear.
- 2
Record the work
Use Loom to capture the process while completing it normally. Explain decisions, exceptions, handoffs, and tools as they occur.
Pro tip Narrate why you make each decision, not only which buttons you press.
Watch out An unnarrated screen recording may omit the judgment needed to reproduce the task.
- 3
Generate and refine the SOP
Turn the recording into a structured standard operating procedure. Check that its inputs, actions, decision points, and outputs accurately represent the work.
Pro tip Edit the generated SOP while the process is still fresh in your mind.
Watch out Do not assume an automatically generated SOP captured every exception correctly.
- 4
Request an AI process review
Give the SOP to a reasoning model and ask it to identify waste, unclear decisions, failure points, and automation opportunities. Require concrete recommendations rather than a generic summary.
Pro tip Ask the model to rank recommendations by impact, effort, and risk.
Watch out Treat recommendations as hypotheses until they are tested against the real workflow.
- 5
Implement the best opportunities
Use a coding agent, custom code, Zapier, or another no-code platform to automate the selected steps. Preserve human review wherever judgment or significant risk remains.
Pro tip Automate one bounded step before attempting an end-to-end replacement.
Watch out Do not automate a broken process without first correcting its logic.
- 6
Measure and institutionalize
Compare the revised process with its previous time, quality, and error rate. Update the SOP and give the team recurring protected time to repeat the cycle.
Pro tip Reward documented improvements so process work is treated as part of the job.
Watch out Without scheduled improvement time, delivery pressure will crowd out the practice.
In the wild
A marketer records how they collect campaign metrics, clean a spreadsheet, identify anomalies, and write a weekly summary. The generated SOP is reviewed by AI, which recommends pulling metrics through APIs, validating missing fields automatically, and drafting the recurring narrative. A coding agent implements the data collection while the marketer retains final editorial review.
→ The recurring report requires less manual collection while preserving human judgment over interpretation.
A manager gives each employee a protected week to record one important process, generate an SOP, and obtain AI improvement suggestions. Employees choose feasible automations and demonstrate the revised workflows at the end of the week.
→ The organization gains documented processes and multiple targeted efficiency improvements without requiring every employee to code.
Common mistakes
Skipping process definition
Automation recommendations will be unreliable if the operator cannot identify the task's inputs, decisions, and desired output. Document and clarify the process first.
Expecting optimization during spare time
Employees usually prioritize immediate delivery over process improvement. Reserve explicit time, such as an Async Week or recurring weekly block.
Automating every recommendation
AI suggestions vary in feasibility and risk. Rank, test, and validate each change before expanding it.
Is it for you?
Best for
It is best for individuals and teams performing recurring knowledge-work processes that currently live in employees' heads.
Not ideal for
It is not ideal for rare, unpredictable work whose inputs and desired outputs cannot yet be defined.
From the transcript
“what is stopping you from going and recording your processes in loom loom then generates Sops for you”
“you feed that sop into open ai's new 01 model and you say reason through how this can be automated or improved”
“just Implement what I'm calling async week you go off on your own you go those four steps”
From the episode
How I Use Loom + Ai To Automate Everything I Do - Steph Smith