Skip to main content

What Data Can the AI Assistant Access?

Dotan Asselmann

The AI Assistant works from the surgical data Theator holds for your cases — the cases you have permission to view, their metadata, the surgical video, and the AI-generated annotations and benchmarks built from that video. It does not read your hospital's records, and it does not analyze video on the spot. This article explains what it can access, what it cannot, and why a request sometimes fails.

What the AI Assistant Can Access

The assistant can search, filter, and analyze the case data and AI-generated outputs Theator stores for your procedures:

  • Your cases and case metadata — search and filter your cases by procedure type, date, surgeon, facility, operative time, and other structured case details. Results are always limited to the cases you have permission to view.
  • AI-generated insights and annotations — the procedure steps, surgical events, and milestones Theator's models mark on each case's timeline. See Intro to Insights.
  • Performance and benchmarks — case analysis with operative-time comparisons against your personal average and peer averages.
  • Video clips — specific moments in a case and short clips you can review or download. Video clips are only available when the case was recorded and the video has not been removed by your venue's retention policy.
  • Charts, tables, and exports — visual summaries of a group of cases, and CSV or Excel exports of a case cohort.
  • Collections and sharing actions — creating collections, sharing cases, and setting up report subscriptions on your behalf.
  • Help Center articles and published literature — Theator's own documentation and public medical literature (PubMed).
  • Files you upload — documents you attach directly to the conversation.

What the AI Assistant Cannot Access

Some requests fall outside what the assistant can reach:

  • It does not analyze raw video on demand. The assistant surfaces the AI annotations that were already generated when a case was processed. It does not watch a recording live to answer a new question, so asking it to re-analyze the footage in the moment will not work.
  • It does not identify or segment anatomy in real time. It cannot outline a specific structure — such as a nerve or vessel — on the video for you. It can only point you to annotations and clips that already exist for the case.
  • It does not read your hospital's records or post-op notes. The assistant works from Theator's surgical data. Clinical records such as post-operative notes are not available to it unless that data has been integrated into Theator at your venue.
  • It only returns data you have permission to view. Every answer is scoped to your own access. It cannot show cases, surgeons, or data outside that scope.
  • It does not retain patient identifiers in its long-term memory. The assistant's durable cross-session memory stores only your standing preferences — such as how you like answers formatted — and explicitly excludes MRNs, patient identifiers, and case-specific content. Conversation history from the current session is saved and accessible while the session is open, subject to your normal access permissions.

Why a Request May Fail

If the assistant declines or comes back empty, it is usually one of these:

  • The case is outside your access. The case may not be linked to you, or it may belong to a scope you do not have permission to view.
  • The case has no recorded video. Some cases are created from EHR scheduling data and have no video footage. Video clips, video-based insights, and similar actions are not possible for these cases.
  • The video was removed by your venue's retention policy. After the retention period expires, the video is permanently removed. Case metadata and any AI annotations generated before removal remain available, but you cannot view or clip the footage.
  • The case is still being processed. A recording that is still being processed has no AI annotations yet, so there is nothing for the assistant to draw on. Try again once processing completes.
  • The procedure type is not supported by AI annotation. Not every procedure type has AI models that produce annotations. The assistant can still retrieve case metadata and operational data for unsupported types, but annotation-based analysis will return empty results.
  • The data or feature is not enabled at your venue. Some data and capabilities depend on what is turned on for your organization. Contact your venue administrator or Theator support to check.
  • You asked for something the assistant does not do. On-demand video analysis and live anatomy segmentation are not capabilities of the assistant, so those requests will not return a result.

When a request fails, try rephrasing it around data the assistant can reach — a case, a procedure type, a date range, or a metric — or share feedback so the team can improve the experience.

What's Next?

Comments

0 comments

Please sign in to leave a comment.