Google Drive is enough for an AI-readable workspace

Google Drive Is Enough For An AI-Readable Workspace

An AI workspace does not have to start as a custom app.

For a lot of operators, consultants, creators, and small teams, the better first version is much simpler:

a Google Drive folder
a start-here doc
a source map
a workbook
a few playbooks
a review gate
a place to capture what changed

That is enough structure for a human to understand the work, and it is often enough structure for an AI assistant to collaborate without needing every rule packed into one long prompt.

I made a general-purpose version here: General Purpose ICM Workspace Template.

The Drive-native template folder is here: General Purpose ICM Workspace Template.

The Tool Is Not The Point

The point is not that Google Drive is magical.

The point is that the workspace needs to be interpretable.

An assistant needs to know:

what the workspace is for
what sources are approved
what sources are off limits
what decisions have already been made
what should be drafted, routed, or rejected
what requires human approval
what changed after the last run

If those rules live only in chat, they are easy to lose. If they live in normal files and rows, they are easier to inspect, update, and hand to a tool like Claude Code, Codex, or another approved assistant.

Specific connector behavior varies by account, permissions, and tool setup. But the underlying pattern is practical: give the assistant a workspace it can read before asking it to help.

What The Template Includes

The template is built as Google-native files:

00 Start Here - General Purpose ICM Workspace
01 Agent Instructions - Use This With Claude Code Or Codex
02 Source Map And Boundaries
03 General Purpose ICM Workbook
04 Route Planner Playbook
05 Retro Capture Playbook
06 Review Gate
07 Example - Brain Dump To Workspace
08 Prompt Library
09 Third-Party Notices

The workbook has tabs for setup, source mapping, workspace mapping, brain dumps, route decisions, workflows, skills and playbooks, automations and heartbeats, memories and lessons, review gates, and agent runs.

That sounds like a lot, but the point is simple: separate the things that should not be mixed together.

A raw brain dump is not a decision.

A source is not a claim.

A draft is not approval.

A useful run should leave a trace.

The Two Playbooks That Matter First

The template includes two adapted public playbooks.

The first is a route planner.

Use it before starting non-trivial work. It turns a messy request into a route: desired outcome, workspace, approved sources, workflow stages, missing context, approval gates, and a starting prompt.

The second is retro capture.

Use it after a run. It compares the initial request with what the user actually needed, captures blockers and lessons, and proposes where the workspace should be improved.

That matters because most AI workflow improvement comes from the gap between what you thought you wanted and what the work revealed.

How To Use It With An Assistant

Start by giving your assistant access to the Drive files you want it to use.

Then use a prompt like this:

Use this Google Drive folder as an ICM workspace. Start by reading the Start Here doc, Agent Instructions, Source Map, Workbook, Route Planner Playbook, Retro Capture Playbook, and Review Gate. Then help me route one messy idea, goal, or project into the right workspace, workflow, and next action. Do not publish, send, schedule, or change external systems without my explicit approval.

The assistant should not start by doing everything.

It should start by routing the work.

Why This Is Foundational

Before you build automations, agents, or custom workflows, you need a place for context to live.

That place should make it clear:

what is known
what is assumed
what is approved
what is risky
what needs a human
what should improve next time

That is the broader idea behind ICM.

If you want to learn ICM from the community around Jake Van Clief's work, the Clief Notes Skool Community is a useful place to explore the method and see how other builders think about AI workspace structure.

Start Here

Open the resource page: General Purpose ICM Workspace Template.

Then open the Drive-native template folder: General Purpose ICM Workspace Template.

Run one messy request through the route planner.

After the run, use the retro capture playbook.

That is the loop: route the work, do the smallest useful next action, review what changed, and improve the workspace.