Mira AI agent - Assessments autofill

What is Mira Privacy Specialist?

Mira's Privacy Specialist is an AI writing assistant that prepares suggested answers with evidence for assessments (RoPA, DPIA, TIA, vendor security, AI governance, and more). It uses your assessment schema, linked data sources information, existing answers, uploaded documents (e.g., vendor contracts, project slides), employee form submissions, and business details to create a draft version of your assessment with only what it can confidently defend.

Where to find it

Open any assessment → top-right Write with Mira button
If the button is disabled: you can go to "Mira AI agents" in the navigation → Assessments auto-fill and turn on the agent there. 

How it works (step-by-step)

1. Open the assessment you want to accelerate.

2. For better results:
   1. Add a data source to the assessment (and make sure the data source is filled in correctly)
   2. Add labels 
   3. Add a prompt—context or instructions specific to this assessment (e.g. what is this assessment for, what is the project, which regulation)
      

1. 1. Upload files to the reference materials - vendor contracts, project slides. You can reach the reference materials pane by clicking on"add supporting files" or "view"reference materials if you've already added them. 
      Only once Mira is done analyzing the document will they be used as context. 
      

2. Click Write with Mira once your prompt and files are ready.

3. Click Run. You’ll see a spinner; you can leave the assessment and come back later to see results. 

4. When ready, suggestions and evidence appear inline per question. You can expand the evidence to see more details 
   

5. Accept/Reject suggestions individually or for the whole assessment.

6. Click Save (standard assessment save) to persist accepted answers.

What the AI uses

• Existing answers (never overwritten)

• Uploaded files (contracts/SCCs/policies - all uploaded documents)

• Business details (set up in your settings: here's how)

• Your prompt

• Custom fields (e.g., legal basis, data subjects) + your custom values (set up in your settings)

• Regulatory guardrails based on template type (e.g., GDPR, DORA, EU AI Act)

Results

• A suggestion you can accept or reject
• Evidence and sources including: 
  • Data sources information 
  • Your uploaded documents
  • Framework links used for guardrails include GDPR Art. 30, EDPB DPIA, NIST AI RMF, EU AI Act, DORA, SOC2/ISO readiness, and others you’ve mapped by template.

• Generates minimal valid answers in the correct format.

• Honors field types (short/rich text, single/multi-select, date, etc.).

• If not confident, leaves blank.

Tips & best practices

• Upload documents first. Contracts, DPAs, or policies dramatically improve AI writing quality.

• Use specific prompts. Example: “Vendor = X; processing = support tickets; no children’s data; prefer SCCs.”

• Keep your Business Details updated. The more detailed it is, the better the AI aligns with your organization.

• Leverage closed lists. Keep your custom values tidy in Settings so AI can select the right options.

• Scope the schema. Hide questions you’ll never use to reduce noise and speed up AI writing.

Troubleshooting

• No suggestions appeared → Check assessment state (not Completed), role (not Collaborator), and that the schema has fillable fields.

• Suggestions look generic → Add/upload evidence; refine the assessment-level prompt; ensure template’s closed lists are populated.

• Wrong data types proposed → Confirm your custom field configurations and option values; re-run after fixes.

• Performance feels slow → Large files or long templates take longer; you can keep working while the job runs asynchronously.