[Home](https://help.qlik.com/en-US/cloud-admin-playbook/Content/home.htm) > [AI services](https://help.qlik.com/en-US/cloud-admin-playbook/Content/ai_services.htm) > [Qlik Predict](https://help.qlik.com/en-US/cloud-admin-playbook/Content/ai_services/qlik_predict.htm) > Review Predict deployments

# Review Predict deployments

**Cadence:** Weekly

**Subscription model:** Premium and Enterprise editions of capacity-based Qlik Cloud Analytics; basic tier with Qlik Sense Enterprise SaaS (user-based)

**Roles:** Tenant Admin (TA), Analytics Admin (AA)

**Last verified:** 2026-05

|  | Initial | Recurring |
| --- | --- | --- |
| **Estimated time** | 30 min | 15 min |

> **Applies only if your tenant has Qlik Predict entitlement.** Quick check: admin console side nav for **ML experiments** / **ML deployments**. If absent, skip this page.

**Benefits:**

* Approval pipeline doesn't bottleneck on stale model-approval requests
* Active deployed-model count stays within tier limit
* Prediction-configuration ownership is reassigned promptly when owners depart

***

## Goal

Models must be approved before predictions are generated. Approval can be done by users with the **Approve or reject ML models** permission, or by tenant / analytics admins. Tracking deployed-model count is important because the subscription enforces a maximum number of active deployed models across all tenants within the subscription (for multi-tenant subscriptions); reaching the limit blocks new deployments until existing ones are deactivated or deleted.

## Tasks

| Task | TA | AA | DA | AuA |
| --- | --- | --- | --- | --- |
| Review pending model-approval requests | ✓ | ✓ |  |  |
| Review **Active deployed models** count vs. tier limit | ✓ | ✓ |  |  |
| Review prediction-configuration owners that need to be reassigned (for example, owner left the tenant) | ✓ | ✓ |  |  |
| After any tenant-level update, re-test forecast real-time prediction endpoints on the Machine Learning API | ✓ | ✓ |  |  |

## Gotchas

* **Forecast (vs. regression) deployments can return `FAILED_PRECONDITION`.** The error means the prediction request's preconditions aren't met. Common causes: an **association-field data-type mismatch** between the app and the deployment (for example, a key stored as text in one place and integer in another); the caller lacking **Can Edit** on the space where the deployment lives; or a **new forecast deployment failing after a tenant-level update** while existing deployments keep working. Re-test forecast endpoints after any tenant update, and verify the association field's type and space permissions before escalating. ([community](https://community.qlik.com/t5/Qlik-Predict/Qlik-Predict-Forecast-Real-time-Prediction-Failed-Precondition/td-p/2542672))
* **The legacy AutoML real-time prediction and dataset prediction endpoints were removed on February 3, 2026** and superseded by the Machine Learning API (see [Qlik release notes and deprecation notices](https://help.qlik.com/en-US/cloud-admin-playbook/Content/tenant_configuration/review_release_changes.htm#references)). Migrate any clients still calling the legacy endpoints.

## References

* [Administering Qlik Predict](https://help.qlik.com/en-US/cloud-services/Subsystems/Hub/Content/Sense_Hub/Admin/mc-managing-experiments-deployments.htm)
* [Who can work with Qlik Predict](https://help.qlik.com/en-US/cloud-services/Subsystems/Hub/Content/Sense_Hub/AutoML/who-can-work-with-automl.htm)
* [Monitoring resource consumption](https://help.qlik.com/en-US/cloud-services/Subsystems/Hub/Content/Sense_Hub/Admin/mc-monitor-consumption.htm)
* [Qlik Cloud Analytics subscription options](https://help.qlik.com/en-US/cloud-services/Subsystems/Hub/Content/Sense_Hub/Admin/subscription-options-analytics.htm)
* [Product Descriptions for Qlik Cloud Subscriptions (PDF)](https://assets.qlik.com/image/upload/v1713309774/qlik/docs/Legal/terms-and-conditions/product-descriptions-for-qlik-cloud-subscriptions_hxk8re.pdf)

**Tags:** `#weekly` `#ai_services` `#qlik_predict` `#automl` `#deployments`