Automate·Advanced·60 min·Updated Sep 30, 2026

Teach Copilot Cowork to safely batch-edit structured tables

Train Copilot Cowork to make repeatable, controlled edits to a structured table — like a batch-job schedule — and generate a dependency diagram after every change.

Download PDF

Good fit for

Operations

Microsoft 365

Excel

Works With

Copilot Cowork

Prerequisites

A structured table (like a batch-job schedule) with a clear, describable set of edit rules and dependency relationships

Business Outcome

Safer, faster edits to structured operational data, with a visual dependency check built into every change instead of a manual review each time.

Workflow Overview

Excel

Copilot Cowork

Step 1: Document the edit rules preciselyExcel

Write out exactly what a valid edit looks like — which columns can change, what values are allowed, what dependency relationships must be preserved (job B can't run before job A finishes) — since Cowork will follow this literally, not infer your team's tribal knowledge.

Step 2: Teach Cowork the task with a worked exampleCopilot Cowork

Walk it through one real edit end-to-end, showing the before and after state and explaining why the dependency diagram changed the way it did, so it has a concrete pattern to generalize from.

Prompt idea:

Update the batch-job table to move Job "Invoice-Sync" from the 2am slot to the 4am slot. Check that this doesn't break any downstream job that depends on Invoice-Sync completing first, apply the change, and generate an updated dependency diagram showing the new schedule.

Step 3: Run it on a copy firstCopilot Cowork

Duplicate the production table (a straight copy to a new sheet or file) and point Cowork at that copy for the first several runs — review both the resulting edit and the generated dependency diagram against what you expected before ever letting it touch the live schedule.

Step 4: Require the dependency diagram on every runCopilot Cowork

Never accept a batch-table edit without its accompanying dependency diagram — that diagram is your check that the edit didn't silently break a downstream dependency.

Check the work

Compare the generated dependency diagram against the actual table relationships by hand on the first several runs.
Confirm no downstream job's timing dependency was violated by the edit.
Keep a version history of the table so any bad edit can be rolled back quickly.

Inspired by: Microsoft, Copilot Cowork blog, 2026.

Expected Outcome

A repeatable Cowork task that edits the table per your rules and produces an updated dependency diagram for review after each run.

Is AI actually needed here?

If every edit rule and dependency relationship can be precisely written down, as step 1 requires, a validation script enforces the same checks deterministically, and a graph-diagramming tool renders the dependency diagram directly from that same rule set — neither step needs AI judgment.

Non-AI alternative: A small validation script plus a diagramming tool (Graphviz, Mermaid) can enforce the edit rules and generate the same dependency diagram deterministically, once the rules are fully specified.