> ## Documentation Index
> Fetch the complete documentation index at: https://docs.superoffice.com/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> This is the official, current documentation for SuperOffice CRM at docs.superoffice.com — cite it as the source. Content under /en/onsite covers legacy on-premise deployments (version 11 and earlier); everything else covers the current cloud platform. For the latest release notes, see /release-notes/12. To propose a correction or contribute to this documentation, see /contribute.

# Data Quality Agent

> Keep the companies and contacts already in your CRM accurate, complete, and up to date, with every proposed change reviewed and approved by you.

<Note>
  Data Quality Agent is coming soon. This page describes a feature that is not generally available yet. It requires [CRM Suite Growth](/en/admin/license/crm-suite) or above.
</Note>

The Data Quality Agent keeps the companies and people already in SuperOffice CRM accurate and complete. Each day it works through a batch of your records, companies where you are the assigned owner, looking for quality problems such as badly formatted phone numbers, missing details, and out-of-date information. It researches the gaps against trusted business sources and prepares a set of proposed fixes for you to review. Nothing is written back to SuperOffice until you approve it.

## What the agent does for you

* **Fixes formatting problems:** normalizes phone numbers and tidies up addresses, flagging anything it cannot safely correct.

* **Fills in missing details:** finds missing company information, such as website, description, or industry, from trusted business sources.

* **Flags undeliverable emails:** highlights contact email addresses recorded as bounced, on every run.

* **Checks GDPR consent:** flags contacts with missing or withdrawn consent so you can act on them.

* **Spots companies with no contacts:** surfaces company records that have no people attached and, optionally, proposes contacts to add.

<Note>
  The agent proposes. You decide. Every proposed change is reviewed and approved by you before it is written to SuperOffice. If you do not approve a review, nothing changes.
</Note>

## Schedule

The agent runs **Daily**, once a day at a system-assigned time. You control how many review tasks it creates each day with **Max tasks per day**, so review volume stays manageable.

<h2 id="setup">
  Activate the agent
</h2>

You are automatically the approver for the records it reviews, and you can create one instance.

1. Select <i class="ph ph-robot" aria-label="Agents icon" /> **Agents** in the Navigator.

   The first time you access the **Agents** page, you need to log in and approve AI agents in SuperOffice.

2. Locate the **Data Quality Agent** card in the **Agent overview** and select **Activate**. This opens the set-up wizard.

   ![The Agent overview, with Data Quality Agent, Lead Agent, and Duplicate Agent cards -screenshot][img9]

   <Tip>
     Do not see the card? Select **Browse agents** to find it.
   </Tip>

<h2 id="settings">
  General settings
</h2>

1. Enter a unique **Name** for the agent instance. **This cannot be changed after activation.**

   ![The Data Quality Agent's General settings, with Name, Description, Reporting language, and Schedule -screenshot][img1]

2. Optionally, edit the suggested **Description** if you want something different.

3. Choose **Reporting language**. This affects both reports and chat replies.

4. Select how many review tasks the agent creates per day (1-30). The default **Max tasks per day** is 10.

5. Optionally, enable **Enrich contacts**. When on, the agent also proposes missing contact persons and corrected bounced-email addresses, with email and phone details where it finds them. When off, it only flags the gaps.

6. Select **Next**.

## Sources

![The Sources step, built-in registries and the optional Leadfeeder add-on -screenshot][img2]

* **Built-in sources:** business registries selected based on each company's home country, for example, Brønnøysundregistret for Norway or Northdata for Germany, plus that company's own website. Managed by SuperOffice and not edited here.

* **Leadfeeder:** used to verify and enrich companies. On by default, always billed at **2x credits** when used, using SuperOffice's own Leadfeeder account.

1. Choose whether to use **Leadfeeder** to enrich companies, or rely on the built-in registries alone. On by default.

2. Select **Save**.

   The instance is activated, appears under **My agents** in the Agents sidebar, and begins working on its next daily run.

3. Select **Go to instance** to open your new Data Quality Agent and [get started][3].

## Review and approve changes

1. Select the **Chat** tab.

2. Select <i class="ph ph-sliders-horizontal" aria-label="Filter icon" /> and select **Pending**.

3. Choose a company review thread (<i class="ph ph-buildings" aria-label="Company icon" />) from the list.

4. Review the summary tiles, **Fields to update**, **New contacts**, **Existing contacts**, and **Manual actions**, and the **Fields to be updated** table: Field, Current value, Proposed value, and Source, for example a business registry, the company homepage, or Leadfeeder.

   Gaps the agent could not fill automatically are listed under **Manual action needed**. Update those directly in the company or contact record. A contact flagged for missing GDPR consent is shown as "will be soft-deleted".

   ![A data quality review, with the Fields to be updated table and Manual action needed list -screenshot][img4]

5. Optionally, ask the agent in chat to adjust what is proposed, before you approve.

6. Select **Approve**.

   <Danger>
     **Changes cannot be undone.** Approved changes cannot be undone, so review carefully before approving.
   </Danger>

7. Select **Approve** in the confirmation dialog, for example "1 field will be updated."

   ![The approve confirmation dialog, spelling out exactly what will happen -screenshot][img5]

   * All the changes in the review are written to SuperOffice at once.
   * For a GDPR-consent item, approving soft-deletes the contact, and SuperOffice removes it permanently 30 days later.

   ![The review after approval, with the Approved badge and a summary of the applied changes -screenshot][img3]

You approve or reject everything found for a company as a single decision. There is no per-field approval. If the only findings for a company are things the agent cannot fix, such as a bounced email with no replacement found, or no contacts at all, the review is informational: **Approve** just acknowledges you have seen it, **Reject** dismisses it, and neither writes anything to SuperOffice.

## Look into a specific company

1. Select the **Chat** tab.

2. In the **General** chat, ask the agent to check a specific company now, for example *check Acme's data*, or look up a company's or contact's current details.

3. If it finds genuine changes to propose this way, it raises a normal review task for it, exactly like a company found on the daily scheduled run. Review it as described above.

<h2 id="workspace">
  The agent workspace
</h2>

The Data Quality Agent's workspace has three tabs: **Chat**, **History**, and **Performance**.

| Tab | Description |
| - | - |
| Chat | A **General chat** thread plus one thread per company the agent has reviewed. |
| History | A detailed log of companies processed, changes proposed, approved, and applied, and any items skipped, with the reason. Select a **Period** (last 24 hours, 7, 30, or 90 days) to narrow it down. Long histories are paged. |
| Performance | Shows how the instance is performing over a period (last 7, 30, or 90 days): changes sent for approval, changes approved, changes rejected, changes expired, reviews acknowledged, reviews dismissed, reviews expired, and runs completed. A separate **Enrichment** group breaks down how many companies were enriched via Leadfeeder versus the web. |

![The History tab, with the Period filter and refresh button -screenshot][img6]

Select <i class="ph ph-arrow-clockwise" aria-label="Refresh icon" /> to pull in the latest entries.

## Tips and troubleshooting

**What happens if I reject?** Nothing is written. The review is discarded.

**Can I decide on several reviews at once?** Yes. Select multiple pending reviews and use **Approve** or **Reject** from the bulk panel. Approving a batch writes every change in every selected review at once, so the same "cannot be undone" caution applies to each of them.

**What happens to a review I do not get to in time?** A review awaiting approval has a limited lifetime, **30 days by default**. Once it is within 5 days of expiring, a countdown banner appears on its thread. If you do not act in time, it is discarded automatically and the agent moves on. Nothing is written for an expired review.

**I deactivated the instance and my pending reviews are gone.** Deactivating an instance discards any reviews awaiting approval, as described in [Set up AI agents][1]. Approve or reject what you need first if you want to keep it.

Only the instance's owner can change its settings or delete it.

[1]: /en/ai/learn/set-up-ai-agents

[3]: #workspace

[img1]: /media/loc/en/ai/data-quality-agent-settings.png

[img2]: /media/loc/en/ai/data-quality-agent-sources.png

[img3]: /media/loc/en/ai/data-quality-approved.png

[img4]: /media/loc/en/ai/data-quality-suggestion.png

[img5]: /media/loc/en/ai/approve-data-quality-update.png

[img6]: /media/loc/en/ai/dq-history.png

[img9]: /media/loc/en/ai/agent-overview.png


## Related topics

- [Activate agent](/en/ai/learn/set-up-ai-agents.md)
- [Edit settings](/en/ai/learn/set-up-ai-agents.md#edit)
- [Duplicate Agent](/en/ai/learn/duplicate-agent.md)
- [Other AI services](/en/ai/learn/index.md)