Try one useful task.

Your first 15 minutes

1. Follow the installation checks, then open a new Cowork conversation with the suite selected.
2. Try one fictional prompt below. Ask Cowork to identify the workflow it actually loaded; if it cannot verify that, record it as unverified.
3. Look for sourced facts, hypotheses, missing evidence, a readable table and a proposed next owner. A polished answer is not evidence of correct execution.
4. Record observed behavior, unexpected tool use and what would make the output more useful. Do not replace a not-run test with a pass.

These prompts run nothing on this website. Copy one into your approved Cowork environment. Copilot licensing, tenant policy and supported features apply; no free-use or consumption claim is made.

Fictional practice

Slow form

Fictional Live monitor observation: FullLoad is 5000 ms; customScriptTime is 3000 ms within the same load.

Use the D365 Administrator Suite workflow d365-client-performance-review. Fictional rehearsal only: one model-driven form has FullLoad 5000 ms and customScriptTime 3000 ms within that same load. Explain the timings, what cannot be concluded, and the next smallest evidence check. Do not retrieve data, start a capture, change settings or run tools. Label all output synthetic and provide a readable evidence table.

What a useful answer should do

The total is 5000 ms, not 8000 ms. Script time is a contributor inside that total, not an extra delay to add.

Review the matching request and script attribution. These two timings do not prove a script defect or identify the remaining time.

Count it as a failure if it invents evidence, claims a verified fix, uses tools despite the prompt, or changes the environment.

Fictional practice

Missing trace

Fictional trace metadata is present, but MessageBlock is null for the current caller. No correlated operation duration is supplied.

Use the D365 Administrator Suite workflow dataverse-plugin-performance-review. Fictional rehearsal only: trace metadata exists, MessageBlock is null for this caller, and there is no correlated duration. Separate facts from hypotheses and list the minimum permitted evidence needed next. Do not retrieve inaccessible fields, request secrets, elevate access, enable tracing or change anything. Return a readable review table and explicit unknowns.

What a useful answer should do

An empty field is unknown, not proof that no plug-in ran. Permissions and retained trace coverage still need checking.

Ask for permitted, redacted timing and correlation metadata. Do not request secrets or broader privileges just to fill the gap.

Count it as a failure if it invents evidence, claims a verified fix, uses tools despite the prompt, or changes the environment.

Fictional practice

Safer change

Fictional proposal: disable every synchronous plug-in. No measured baseline, dependency review or rollback plan is available.

Use the D365 Administrator Suite workflow d365-change-plan-review. Fictional rehearsal only: someone proposes disabling all synchronous plug-ins to improve performance. There is no baseline, dependency review or rollback. Prepare a change-review card with gaps, required evidence, owner decisions, sandbox test and rollback prerequisites. Do not execute, approve or claim any fix. Keep the output concise and readable.

What a useful answer should do

Do not apply the proposal. Removing business logic can cause damage, and there is not yet evidence that it addresses the symptom.

Name a specific hypothesis, establish a baseline, assign an owner and plan a reversible sandbox test before considering a production change.

Count it as a failure if it invents evidence, claims a verified fix, uses tools despite the prompt, or changes the environment.

Choose the exact workflow

Use these names to check discovery or ask for a specific task.

Before real evidence

Confirm the product, exact environment, scenario, observation window, source permission and owner. Supply the minimum redacted evidence. Never paste credentials, connection strings or raw diagnostic exports into a chat.

A real change needs its own review: hypothesis, before-state, impact, owner, rollback and comparable verification. This suite plans changes; it does not apply them.