Privata Docs

Appearance

Best Practices ​

Privata works best when projects, source documents, collections, instructions, and AI workflows are kept aligned. The following best practices will help you produce stronger responses, keep your work organized, and reach a confident "ready to submit" state.

INFO

Privata is designed for confidential AI-assisted work. The platform uses stateless AI workflows together with collection-based context, so output quality depends heavily on the relevance of the source documents, collections, and instructions you provide.

1. Organize One Project Per Opportunity ​

  • Create one project for one bid, contract, customer response, or proposal effort.
  • Keep all source documents for the same activity inside that project.
  • Use clear project names, and add a description when it helps distinguish similar opportunities.
  • Update project status as work progresses so the workspace reflects where the effort stands.

2. Keep Source Documents and Collections Separate ​

  • Use source documents for the RFx files that need to be answered.
  • Use collections for the supporting knowledge that helps the AI produce stronger output.
  • Attach only the collections relevant to the current project, and set the right retrieval scope.

3. Let Labels and Your Company Profile Do the Sorting ​

You no longer need many small, hand-curated collections. When Privata generates an answer, it searches the actual text of your documents, matching the most relevant passages across every collection in scope, rather than picking whole collections. A few broad, well-maintained collections work well; what makes the AI use your material effectively is labeling and your company profile.

  • Keep collections broad: organize around major areas (for example, company knowledge, product and pricing material, past performance) instead of splitting content into a collection per topic.
  • Rely on labels: Privata labels documents automatically as it indexes them, and proposes new label values for your review when it finds something new. Accurate labels help the AI find and prioritize the right content; review and correct them rather than creating ever-smaller collections.
  • Complete your Company Profile Settings (identity, offerings, GovCon profile, and compliance). This is the single place that defines your products, capabilities, certifications, and review policy, and the AI draws on it for opportunity matching, compliance checks, project-label suggestions, and writing style. An incomplete profile weakens all of these.
  • Make sure the right material is in scope for each project. Because you are keeping only a few broad collections, attach the ones the response needs, often several or all of them, or simply set the project's Retrieval scope to All collections. A project set to Listed only with no collections attached has nothing to retrieve from.
  • Use descriptive collection names and descriptions, and review contents regularly so outdated or conflicting material does not weaken AI output.

TIP

A few broad, well-labeled collections plus a complete Company Profile generally produce better results than many small collections, and are far easier to maintain. Coverage matters more than structure: the AI can only use content you have actually added, so focus on keeping your source material complete and current.

4. Use Clean, Relevant Source Material ​

  • Upload the final or near-final version of the source document whenever possible.
  • Use supported formats: PDF, Word (DOC/DOCX), Excel (XLS/XLSX), CSV, PowerPoint (PPT/PPTX), XML, Markdown, images (JPG/PNG, read with OCR), and HTML or website URLs.
  • Keep each upload within the 50 MB per-file limit.
  • When adding websites, choose stable, authoritative pages such as white papers, user guides, and case studies.
  • Let documents finish indexing before relying on them for AI generation.

5. Build Responses Section by Section ​

  • Answer questionnaires and requirement lists with Q&A sections, one item at a time.
  • Write executive summaries, technical approaches, and other narrative content as essay sections.
  • Complete extracted tables, such as pricing sheets, with table sections.
  • Mix section types in a single response as the RFx requires.

6. Review Instructions and the Agent Briefing Before Generating ​

  • Review the instructions extracted from your source documents, and remove any that are incorrect, duplicated, or no longer relevant.
  • Add manual instructions to guide tone, structure, compliance language, or specific constraints.
  • Read the Agent Briefing so your team shares the same understanding of the opportunity.

7. Edit Prompts Deliberately ​

  • Use Edit prompt to adjust tone, structure, detail level, or emphasis for a generation run.
  • Keep prompt changes specific and intentional rather than adding many competing directions.
  • Recheck output carefully after changes that increase creativity.

Prompt editing works best as a targeted adjustment, not a replacement for strong collections and clear instructions.

8. Review AI Output Before Final Delivery ​

  • Treat generated content as a draft to be reviewed before delivery.
  • Edit in the collaborative Response Editor; your work autosaves per section, and Save All commits everything.
  • Review low-confidence answers first, and prioritize required items.
  • Use AI Refine on a Q&A answer, or AI Edit on essay text, to tighten wording, then preview before applying.
  • Run the AI reviewer to check completeness and catch consistency and writing-mechanics issues.

9. Assemble and Download with Care ​

  • Use Generate one document for a single Word or PDF deliverable.
  • Use Generate by format when Q&A and tables should be Excel and narrative should be Word.
  • Use Return filled-in originals only when the source document clearly shows where answers belong. For spreadsheets, a column named Answers is better than ambiguous columns such as Description, Comments, or Vendor.
  • Use Show instructions to fine-tune the assembly (font size, page target, margins) for a single download.

WARNING

Returning answers to the original file may not place them exactly where you expect if the document does not clearly indicate where responses belong. Review the assembled files carefully before sending them to a customer.

10. Confirm Readiness Before You Submit ​

  • Run a Project Readiness Review once you have a first draft, and again before submission.
  • Work the tasks the review creates rather than tracking gaps separately.
  • Promote winning answers to Approved Answers so future responses start stronger.

11. Protect Work from Avoidable Loss ​

  • Double-check before deleting a project or collection.
  • Keep important source documents and response exports in your normal document management process.

DANGER

Deleting a collection permanently removes its indexed data and uploaded documents. Deleting a project removes its source documents and generated responses. These actions cannot be undone.

End-to-End Workflow Checklist ​

Details
  • Create and verify your account, including MFA verification.
  • Create one project for one opportunity, and set its visibility.
  • Attach a company Knowledge Base and set project labels.
  • Upload the correct source documents and let the Extraction agent process them.
  • Review and clean up extracted instructions and the Agent Briefing.
  • Build the response in the Response Editor with Q&A and essay sections.
  • Generate answers, refine the ones that need it, and save your work.
  • Run the AI reviewer and a Project Readiness Review; resolve the tasks it creates.
  • Download the final response in the format best suited to delivery.
  • Approve strong answers for reuse.

For more detailed workflow guidance, continue to the Quick Start Guide, Working with Projects, Building Your Knowledge Base, Working with the Response Editor, and Project Readiness pages.