---
title: "Workday test evidence and agents — Kiweely"
description: "Workday has not published an AI evidence pack. Keep Tenant Compare, post-migration reports, and the events you ran. Agent chat expires in 14 days."
url: https://kiweely.com/guides/workday-uat-evidence-pack-ai-agents
---

# What counts as Workday test evidence when an agent helped

Implementation tenants are for unit and end-to-end testing. Evidence Workday retains includes post-migration summaries for two years, Tenant Compare, and the business process and integration results you run. An Ask transcript is not evidence after 14 days, and it is not a sign-off.

-   **Checked** against Workday on November 16
-   Applies to **Workday Extend**
-   6 min read
-   Published November 16
-   By Kiweely

Workday has **not** published a user-acceptance evidence pack, a test-case object, or an agent that signs a test. Implementation tenants exist for unit and end-to-end user testing, and also for integrations, legacy loads, prototyping, and training. The evidence is whatever Workday keeps after you run the test, plus the record you file outside the chat.

Deployment Agent with Ask can help you see why a step did not run, in a non-Production tenant. Workday deletes that chat after **14 days**, isolates it by tenant, and tells you the answer may be incomplete. A transcript is a lead. It is not the evidence.

## Evidence that survives

| Test you ran | What to keep | How long Workday says it lasts |
| --- | --- | --- |
| Configuration migration | Post-migration summary. Optional Tenant Compare, which you can save. | Post-migration summaries remain available for two years. |
| “Does Sandbox match the Implementation tenant?” | Tenant Compare: unmatched Reference IDs (only one tenant has the instance), matched instances with attribute differences, instances with no Reference ID, suspended domains and subdomains when you compare domain security policies. | The report you save. A progress bar under 100 percent is not a result. Refresh until it finishes. A zero count can mean “nothing to compare,” a type that does not use Reference IDs, or a missing ID. |
| Business process behavior | The event, the definition that inheritance selected, the effective date, and **Business Process > Test Rule** on an event that uses a condition. You need the Business Process Administration domain. | The event in the tenant. The course does not call the agent’s paraphrase a test. |
| A step someone completed | The business process audit of who acted. The setup topic says the framework supplies an audit trail after completion. The payroll course says you can see who performed or approved a pay cycle step and when. | The process history, not the chat. |
| Inbound or outbound data | The integration launch and its result. The EIB course’s pattern is launch and verify. Generating a template is not a result. | The integration event in the tenant. |
| Security access | The domain or business process policy after **Activate Pending Security Policy Changes**, and a user in that group actually opening the task. | The activated policy. An Ask explanation of the policy is not the activation. |
| Deployment Agent’s own answer | Nothing you cannot open. Workday’s own check of Ask is five calculation fields confirmed in **View Calculated Field**. | Chat: 14 days, then deleted. |

Customer Central is the place you run migration and Tenant Compare. Starting a migration in the source tenant only redirects you there.

## What an agent may add, and the label it needs

If someone uses Ask during a test, file a separate note the same day. Label it **agent lead**, not **test result**.

Include:

-   Tenant name, and whether the prompt said “in my tenant” or named the connected tenant. Without that, you may have used documentation Q&A, which cannot see instances.
-   The prompt and the objects it named.
-   The task where a person confirmed or refuted each object.
-   The Workday record that is the result: event, integration event, post-migration report, or saved Tenant Compare.

Do not file an assistant’s wording as “Workday passed.”

## A pack a reviewer can open

One folder per test, four items:

1.  **Scope.** Tenant, date, the worker or security group, the business process type, the definition name, the effective date. If several definitions exist, say which one inheritance should select: the organization’s, a superior’s, or the default. A definition with errors is skipped.
2.  **Action.** What was submitted, by whom. For a migration, the package name and the target tenant. For a load, the integration system and whether you launched or only generated a template.
3.  **Workday result.** A pointer to the event, the post-migration summary, or the saved compare. Not a screenshot of a chat in place of those.
4.  **Exception.** Any condition that did not fire, with the Test Rule outcome, or any prerequisite Object Transporter still listed.

Agent leads sit in an appendix. If the appendix is the only copy of a field name, the test is not finished.

### Prompts that produce a lead worth confirming

> I am a Workday administrator. In my tenant, the \[business process type\] for \[organization\] did not enter the \[step name\] step. Which entry condition is on that step, and which definition ran? I will confirm with Test Rule.

> I am a Workday security administrator. In my tenant, users in \[security group\] cannot view \[field\]. Which domain security policy controls the field, and which groups have View?

Both follow Workday’s published examples. After the answer, run the test or open the policy. Activate changes if you edit security. Then write the result from the task, not from the reply. Rate the reply only after that.

## Failure modes

| What gets filed | Why a reviewer should reject it |
| --- | --- |
| A chat export and no event | Deleted as a system record after 14 days, and never a completion audit. |
| “Agent said the rule is correct” | Workday says to validate against documentation and the tenant. Test Rule exists for events that use a condition. |
| A Tenant Compare at 40 percent | Not finished. |
| A zero on Tenant Compare read as “no differences” | Zero also means missing Reference IDs or a type that does not use them. |
| A Sandbox test after refresh, cited as last month’s scenario | Sandbox is replaced by a copy of Production. The scenario is gone unless you reloaded it. |
| Q&A used as proof of tenant configuration | Q&A cannot see security policies or instances. |
| Production chat as proof | Deployment Agent is not a conversational assistant in Production. |

## Frequently asked questions

### Does Workday require a particular evidence template?

Not on the pages checked. It requires you to use the tenants and tools it does provide: implementation tenants for end-to-end testing, migration reports, Tenant Compare, business process history, and integration results.

### Can we keep agent answers by configuring the agent in Production?

Configuring it in Production preserves the **setup** across a Sandbox refresh. It does not preserve chat, and you cannot use the agent as a conversational assistant in Production. Chat retention stays 14 days in the tenant where you talked.

### Is a proxy test the same as the worker’s security?

Only if the session used that person’s access. Deployment Agent does not document a proxy test. A desktop agent’s proxy session still has to be the person whose groups you mean to prove, and you should watch it. The domain policy, activated, is what grants the access.

### What do we attach to a defect?

The event or integration message, the definition and effective date, and the condition test. Attach the prompt only as the lead that told you where to look.

## Sources checked

Public Workday topics listed above, fetched 2026-11-16. Workday does not publish an AI evidence-pack product on those pages.

When the test has to be watched as a specific person, a desktop agent such as [Kiweely Vine](https://kiweely.com/vine) can run that session in a browser you watch. The evidence is still the Workday event, not the session notes.

## Sources checked

-   [Workday Tenants and Tools](https://doc.workday.com/workday-education/en-us/course-manuals/hcm-core-for-administrators/workday-tenants-and-tools.html)
-   [Migrate Packages with Object Transporter](https://doc.workday.com/admin-guide/en-us/manage-workday/tenant-configuration/implementation-tools/customer-central/object-transporter/dgv1633434385688.html)
-   [Concept: Tenant Compare Report](https://doc.workday.com/admin-guide/en-us/manage-workday/tenant-configuration/implementation-tools/customer-central/tenant-compare/concept--tenant-compare-report.html)
-   [Business Process Framework](https://doc.workday.com/workday-education/en-us/course-manuals/hcm-core-for-administrators/business-process-framework.html)
-   [Set Up Deployment Agent in Implementation Environments and Customer Central](https://doc.workday.com/admin-guide/en-us/workday-ai/agents/workday-built-agents/deployment-agent-in-implementation-environments-an/set-up-deployment-agent-on-implementation-environm.html)

## Related guides

[All guides](https://kiweely.com/guides)

-   [### The 2026 Workday AI map: Sana, ASOR, Workday-built, custom, partner, and external agents Workday’s public AI story has five layers: Sana orchestrates, ASOR governs, Workday-built agents ship in the product, partners and customers bring their own, and build tools such as Developer Agent create new ones. This guide places each layer with Workday’s own words.](https://kiweely.com/guides/workday-ai-ecosystem-map-2026)
-   [### What Workday’s Deployment Agent Can and Cannot Do in Your Tenant Workday’s ASOR-enabled Deployment Agent reads tenant configuration in Implementation, Sandbox, Preview and Customer Central. It doesn’t run in Production, and edits configuration only in partner early access. Workday publishes no price; the Rate Card meters 5 credits per Ask interaction.](https://kiweely.com/guides/workday-deployment-agent-tenant-inspection)

### More guides on Custom agents and Workday AI

[All 91 guides](https://kiweely.com/guides#custom-agents)

-   [Security workshops that follow Workday’s own access model](https://kiweely.com/guides/security-design-workshops-deployment-agent)
-   [Pick a Workday integration tool before you prompt an agent](https://kiweely.com/guides/ai-workday-integration-design-workshop)
-   [Is there an AI agent that can do Workday configuration, business processes and security changes in a tenant for me?](https://kiweely.com/guides/ai-agent-for-workday-configuration-business-process-security)
-   [Our Workday consulting firm wants an AI tool that does tenant work for our consultants: what are the options and costs?](https://kiweely.com/guides/ai-tools-for-workday-consulting-firms-options-costs)
-   [Discovery questions Workday’s Deployment Agent can answer](https://kiweely.com/guides/ai-workday-discovery-fit-gap-workshops)
-   [Use Deployment Agent beside Object Transporter, not instead of it](https://kiweely.com/guides/ai-assistant-workday-object-transporter)
-   [Planning Agent and Decision Intelligence](https://kiweely.com/guides/workday-planning-agent-decision-intelligence)
-   [Partner-built agents and the Workday marketplace](https://kiweely.com/guides/workday-partner-built-agents-marketplace)

### From the blog on Custom agents and Workday AI

[All posts](https://kiweely.com/blog)

-   [Connecting Claude or Copilot to Workday via MCP: Seven Security Questions to Answer First](https://kiweely.com/blog/workday-mcp-agent-ready-tools-security-checklist)
-   [Using Glean with Workday Extend: what Workday documents and what it doesn’t](https://kiweely.com/blog/workday-extend-glean)
-   [How to test Workday MCP tools: tool lists, security, write tools and non-deterministic agents](https://kiweely.com/blog/workday-mcp-tools-testing)

## Let Workday check the work.

Kiweely writes WQL and builds orchestrations in your own tenant, and Workday, not Kiweely, says whether they passed.

[Start for free](https://kiweely.com/download) [See how it works](https://kiweely.com/vine/how-it-works)

## Get the next guide in your inbox.

One email a month with new guides and one Extend tip. Nothing else.
