Use AI to create elements

Compliance officers and IT admins who want to unlock AI-assisted drafting in DPMS — the sparkle-icon button that auto-fills descriptions, purposes, and data categories on ROPA records, assets, and TOMs — need to complete two configuration steps in IT Settings first: enabling the AI toggle and entering valid provider credentials. This article walks you through both.

The AI autogenerate feature in DPMS lets you fill in complex fields — descriptions, processing purposes, data categories, and more — with a single click on any record's detail view. Before that sparkle-icon button lights up for your team, however, an IT administrator needs to do two things in IT Settings: switch on the AI feature and tell DPMS which external AI service (Anthropic, OpenAI, or Google) to call. Both steps live under IT Settings › AI, and both must be completed for AI to work. Skipping either one leaves the button greyed out or non-functional for everyone.

How to open it

In the main left-hand sidebar, click IT Settings. A secondary navigation panel appears on the left side of the screen. Click the AI group to expand it. You will see two sub-items:

AI — the on/off switch and industry context setting
AI Credentials — the provider, API key, and model selection

Click whichever sub-item you need. You must have IT Settings read access to view these pages. Editing requires IT Settings edit access; if you only have read access, the Edit button will be visible but greyed out with a tooltip explaining you cannot make changes.

What you see

Both sub-pages share the same layout. At the top of the content area is a small breadcrumb showing your position — for example, IT Settings › AI › AI Credentials — with left and right chevron arrows that let you jump between the two sub-sections without going back to the sidebar.

Below the breadcrumb is a card with a title and an Edit button in the top-right corner. The card lists the current settings as labelled rows. On the AI page you see the feature status (Active or Inactive) and, when AI is on, the selected industry. On the AI Credentials page you see the provider name, model details, token limits, and an Active/Inactive status. If no credentials have been entered yet, the card simply reads "No active provider configured."

Neither card lets you change values in place — all editing is done through the Edit button, which opens a dedicated edit form.

Working with this screen

Enabling AI for the first time

Start on the AI sub-page. The card will show "AI: Inactive." Click Edit. The edit form opens with an AI toggle in the off position.

Flip the toggle on — a blue colour confirms it is active. As soon as you do, an Industry dropdown appears below the toggle. This tells DPMS what sector your organisation operates in (for example, Healthcare or Financial Services) so that AI-generated text uses appropriate regulatory language. Click the dropdown, find your sector, and select it. If you click Save without choosing an industry, the dropdown border turns red and a "Required field" message blocks the save — so make sure to pick one before saving.

Click Save. DPMS stores your settings and returns you to the read-only card, which now shows "AI: Active" and your chosen industry. Important: at this point the sparkle button is still not functional for your team. You also need to configure credentials (next section).

Configuring your AI provider credentials

Navigate to AI Credentials (either click it in the sub-navigation panel or use the right chevron arrow in the breadcrumb). Click Edit.

The first thing you see is the AI Provider dropdown. Select your provider — for example, "Anthropic." The form immediately adds the fields specific to that provider: an API Key field and a Default Model dropdown.

Paste your API key into the API Key field. Then open the Default Model dropdown and select the model your organisation has been provisioned for (for example, "claude-3-5-sonnet-20241022"). As soon as you pick a model, two lines appear below the dropdown in small grey text — Context Window and Max Output — showing the token limits for that model. These are for your information; they are saved automatically with the credential.

Finally, flip the Activated toggle at the bottom of the form to on (it turns blue). This is the switch that actually makes DPMS use the credentials you have just entered. A credential can be saved but left inactive — which is useful if you want to prepare a configuration before a go-live date.

Click Save. DPMS stores your credentials, updates its internal state to mark AI as available, and returns you to the read-only credentials page. From this moment on, the AI sparkle-icon button becomes active for all users who have the appropriate permissions on their records.

Switching to a different AI model (or rotating your API key)

If you need to update your model selection or replace an expired API key, navigate to AI Credentials and click Edit. The form loads with the current provider already selected.

To change the model only: Open the Default Model dropdown and select the new model. The token-limit information below the dropdown updates immediately. Leave the API key field as it is (it shows a partially masked preview of the saved key — for example, "sk-\*\*\*\*\*") and click Save.

To rotate the API key: Click inside the API key field. The masked preview clears and the field becomes a standard password input, ready for you to type or paste a new value. Paste your new key and click Save. The old key is replaced.

Heads up: The moment you click into the API key field, the masked value disappears. If you click the field and then navigate away or click Save without entering a new key, DPMS will save a blank key and AI generation will stop working. Always paste your new key immediately after clicking into the field.

Switching to a different AI provider

If your organisation decides to move from one provider to another (for example, from Anthropic to OpenAI), go to AI Credentials › Edit and change the AI Provider dropdown.

Heads up: Changing the provider immediately erases all previously entered values in the form — API key, model selection, and the Activated toggle — without any warning prompt. There is no confirmation dialog. If you change the provider by accident, click the back arrow at the top of the form to return to the read-only view without saving.

Once you select the new provider, fill in all the required fields for that provider and flip the Activated toggle on before saving.

Note for OpenAI users: If you select a "reasoning model" (such as o1 or o3) from the Default Model dropdown, an additional Reasoning Effort dropdown appears below it. If you later switch back to a standard model, this field disappears and its value is removed from the saved data.

Note for Google users: Selecting Google as the provider shows a Vertex AI toggle. Enabling it reveals two additional fields — Project and Location — needed to route requests through Google's enterprise Vertex AI platform. If you disable the Vertex AI toggle, those fields are hidden and their values are not submitted.

Reviewing the AI configuration (auditors and DPOs)

If you have read access but not edit access, you can still view all settings. The Edit button will be visible but disabled with a tooltip. Navigate to AI to check the feature status and industry context, then use the right chevron arrow in the breadcrumb to jump to AI Credentials to confirm the provider, model, token limits, and active status. All these values are visible in the read-only cards without entering edit mode.

Note that API keys are never shown on the read-only page — only non-sensitive fields are displayed. This is by design.

Field reference

AI (toggle) — Enables or disables the AI autogenerate feature across the entire platform. When off, the industry dropdown is hidden. When on, an industry must be selected before saving.

Industry — The business sector that provides context for AI-generated content. Required when the AI toggle is on. Options include Healthcare, Financial Services, and others. Choosing the wrong industry will result in AI-generated text that uses irrelevant regulatory language.

AI Provider — The external AI service DPMS will call. Currently supports Anthropic, OpenAI, and Google. Changing this field resets all other credential fields.

API Key — The secret key that authenticates DPMS's requests to the chosen AI provider. Stored encrypted; shown partially masked in edit mode. Clicking the field clears the mask.

Default Model — The specific AI model to use for generation requests (for example, "claude-3-5-sonnet-20241022" for Anthropic). Selecting a model automatically populates the Context Window and Max Output values.

Context Window — Shown as an informational label below the Default Model dropdown once a model is selected, and as a read-only field on the credentials index page. Expressed in tokens. A higher number means the AI can process longer records in one go.

Max Output — The maximum length of a single AI-generated response, in tokens. Informational; set automatically based on the model chosen.

Reasoning Effort (OpenAI reasoning models only) — Appears when a reasoning-capable model (o1, o3, etc.) is selected. Controls how much computational effort the model applies. Only visible for compatible models.

Project / Location (Google Vertex AI only) — The Google Cloud project and region to use when routing requests through Vertex AI. Required when the Vertex AI toggle is on.

Activated — Determines whether the saved credential is actually used by DPMS. A credential can be fully configured but left inactive. Only one active credential is used at a time.

How this connects to the rest of DPMS

Everything configured here has immediate, platform-wide consequences. The two most important outcomes are:

The AI sparkle-icon button on every record. Once the AI toggle is on and an active credential is configured, the autogenerate button becomes available on ROPA records, asset records, TOMs, vendor records, and any other element that supports AI-assisted drafting. If either condition is not met, the button remains disabled for everyone.
The industry context and token limits. The industry you choose here is sent with every AI generation request across the platform, shaping the language and regulatory references in all generated content. The context window value determines how much of a record's existing content the AI can read before generating a response — which affects quality for long, complex records.

After finishing this screen, you do not need to do anything else system-wide. Users can immediately start using the AI autogenerate button on their records. If you later need to disable AI for a period — for example, during a contract renewal — you can either flip the Activated toggle off on the credentials page (quickest, leaves configuration intact) or turn off the AI toggle on the general settings page.

Tips & common pitfalls

Heads up: Enabling the AI toggle on the general settings page is not enough on its own. You must also go to AI Credentials, enter your provider's API key, and flip the Activated toggle on. Both pages must be configured for the sparkle button to work.

Heads up: Clicking into the API key field clears the masked preview immediately. If you do this by accident and click Save without entering a new key, your AI integration will break. Always paste the new key straight away, or use the back arrow to cancel.

Switching providers wipes the form. There is no undo. If you accidentally change the provider dropdown, click the back arrow at the top of the form — not the browser's Back button — to discard changes.
Industry validation only fires on Save. The Industry dropdown is required when AI is active, but you won't see the validation error until you attempt to save. Fill it in before clicking Save to avoid the red error message.
The read-only page never shows API keys. If AI stops working and you suspect the key is wrong, you must go into edit mode, click into the API key field, and re-enter the key. There is no "test connection" button to verify the key without making a live generation request.
Inactive credentials mean no AI, everywhere. A fully configured credential with the Activated toggle turned off is treated by DPMS exactly the same as having no credential at all — the sparkle button will not work for anyone.