Companion guide · Updated 24 October 2026. Video scheduled for 24 October 2026 at 5:00 PM Central (America/Chicago, UTC−05:00). The list import and report behavior were tested on 23 October 2026; results and remaining limitations are described below.
Start with the decision
A useful report starts with a decision. For a service team, a leader may need the size of the backlog, an operations coordinator may need the next action on each late job, and finance may need the final cost of completed work. Those needs can share the same source list while using different layouts.
The practice package contains 300 fictional Juniper Field Services work orders. Each record has a stable key, dates and costs, plus detailed notes. The accepted portable import stores calendar dates as ISO text and costs as numbers. It preserved every source value across all 22 columns. The reporting cutoff is fixed at 23 October 2026 so readers can reproduce the same checks later.
Prepare the demo environment
Use a demo SharePoint site, a list you can read and edit, and a supported library where you can create the report. Confirm that Copilot in SharePoint and the relevant HTML controls are actually available to your account. Tenant rollout, license/entitlement and administrative controls can differ. The lab did not establish exact author entitlement or actual credit consumption. Do not change tenant controls just to reproduce this exercise.
Start by importing the data with the supplied field mapping. Confirm that the long notes survive and that blank ratings remain blank. Then ask Copilot to summarize the source before requesting a design. If the record count is wrong, correct the scope or import first.
Define and check the numbers
The baseline is 300 work orders, 150 open, 58 overdue open and 31 blocked. Completed jobs have $272,300 estimated cost and $291,710 actual cost, a $19,410 overrun. The North outstanding-work preset should return 30 rows. These controls are calculated from the supplied synthetic dataset.
Define the measures explicitly. In this example, open excludes Completed and Cancelled. Overdue means an open item with a due date before the cutoff. Completed-job variance compares the actual and estimated costs of completed work only. Costs incurred on an unfinished job cannot establish its final saving.
Build, browse and revise with prompts
Ask for the report's audience, structure, interactions and design in the initial prompt. Follow up with concrete changes: make a view the default, replace a chart, improve the long-note detail or add a clearer filter. Reopen the saved result and repeat the checks after each material revision.
What the tests established
The lab tested a source change separately from a design change. Changing one blocked work order to In progress reduced the original report's blocked count from 31 to 30 after reopening, without regenerating the HTML. Total, open and overdue counts stayed unchanged. The source record was restored. This result applies to the tested original; it does not establish refresh in every revision or location.
The initial report's three audience views, core totals, search, filters, sorting and long notes worked. Prompt-based revisions had mixed results: an overdue filter worked, while a stale search count required another prompt. Requested detail controls and a North outstanding-work preset still had defects in that revision. A final complete rewrite fixed those controls: the North preset matched all 30 open North records in the expected order, and Enter expanded the long notes. Test the saved result instead of treating response text as proof of success.
Where the HTML is saved and what moving it changes
Copilot saved the original in Documents. A preserved test copy moved into Site Pages and became a draft needing publication. After a loading delay, the moved page rendered the correct baseline and its search, notes and reset worked. A separate source-change refresh test after the move remains outstanding. No public sharing or publishing occurred, and separate ordinary/restricted viewer tests remain outstanding.
What administrators should verify
Before sharing, inspect the actual file location and its contents. Audience tabs help readers find relevant information; they do not protect confidential fields. Check the output, source and any exported copies with the intended access model. Record what happens after a move instead of assuming the report is portable.
Use the administration checklist to record author and viewer rights, Site AI settings, Restricted Content Discovery, allowed HTML/file types and link domains, spending policy and credit observations, retention/labels/DLP/audit support, and ownership. These are controls to verify in your environment; this lab did not establish their enforcement. The import file saved in Site Assets and any embedded or exported data also belong in the access review.
Download and follow along
Download the complete practice pack. It includes the 300-row CSV, native-date Excel workbook, portable ISO-text Excel workbook, field mapping, prompt pack, expected totals and administration checklist.
Start with the import guide, then run the source-check prompt, compare the control totals, build the report, browse the controls and apply one revision at a time. Test the actual saved output after each change. Preserve earlier versions before trying a move.
Import guide · Copyable Copilot prompts · Expected results · Admin checklist. No live tenant information is included. The files are a practice pack, not evidence of a completed tenant deployment.
Microsoft references
Copilot in SharePoint (opens in a new tab), HTML page workflow (opens in a new tab), list import (opens in a new tab). Feature rollout and the experience in your tenant can differ. Recheck current documentation before applying the procedure.
