Thesis capabilities

This overview explains the purpose of the product’s capabilities. The available features, connections, and client applications must be confirmed for the specific deployment; additional connections and automation require configuration and the appropriate permissions.

Thesis can receive recordings in several ways. The choice depends on where meetings take place and materials are stored.

Method	Purpose	Requirements
File upload	Process an existing audio or video recording.	A supported file and permission to upload it.
Vinteo connection	Receive recordings from selected conferences and, where supported by the source, return results to it.	A configured source, access, and rules for receiving recordings.
External connector	Receive recordings from another system; a TrueConf connection is documented.	A separately installed and configured connector. Support for other systems must be confirmed separately.
Meeting agent	Connect an agent to a video meeting to obtain its materials.	A supported connection method and a configured profile.
Integration with other systems	Send recordings from another system and retrieve processing status and results.	A prepared integration with the required permissions.

Important: uploading a package of materials does not mean that the transcript is ready. The method used to retrieve the result depends on the integration. An agent that joins a video meeting and an Assistant agent that prepares answers from materials perform different tasks.

Add an audio or video recording approved for processing to obtain a transcript: the spoken text with timestamps. Find a passage and check it against the source material.

Result: text that is easier to work with than a full recording.

• Upload an audio or video file. With a configured connection, recordings can arrive from a conferencing system.
• View text with timestamps and separate speaker turns.
• Compare the text with the audio, correct errors, and save versions.
• Use AI suggestions when proofreading: check changes against the recording, accept suitable suggestions, and save the corrected version.
• Use a glossary when improving and editing the transcript to follow the organization’s preferred spelling of names and specialist terms.

When speaker identification is configured, Thesis can match voices to saved participant profiles. Users check the speaker labels and correct them if necessary. Separating voices alone does not reliably establish participants’ names.

Example: find a discussion of a launch date, listen to that passage, and check the text and speaker label.

Obtain a concise summary, review the main topics and agreements, and correct errors. Prepare a final document from the verified materials.

Result: a starting point for meeting minutes or a message to participants, rather than an automatically approved document.

• Select an available summary template.
• Check topics, decisions, and questions against the transcript.
• Generate a new version of the result if needed.

Teams can configure a shared format for meeting outcomes by choosing a template, the required sections, and instructions. This helps keep the results of recurring meetings consistent.

When configured, summaries can use approved organizational reference documents as additional context. Information from those documents does not, by itself, constitute an agreement reached by meeting participants.

Important: correcting the transcript does not automatically update the summary. After substantial edits, the summary needs to be generated again and reviewed.

Where this feature is available, one or more examples can be used to prepare an organizational document template. The Assistant helps fill it using selected materials or a completed meeting recording. Before downloading or saving the result, check the values, unfilled fields, and whether the formatting matches the selected example.

Decisions, action items, and open questions can be identified in the meeting outcomes. Users check their meaning and manually clarify the next steps.

With the appropriate permissions, users can correct an identified item’s wording, change its type to a decision, action item, or open question, or remove an item identified in error. Check these edits again after generating a new summary.

• Assign an owner and due date, or take responsibility for an unassigned item.
• Mark an item as completed or reopen it.
• View the items assigned to them in a dedicated section.
• Receive reminders about overdue open action items when notifications are configured.

Result: it is clear who will continue the work after the meeting. Assigning an action item does not, by itself, grant access to the source recording.

1. Create a draft from the selected summary.
2. Check the text, assignees, and due dates.
3. Approve the document.
4. Select a workspace and confirm the publication audience.
5. Publish the minutes as a separate action.

Result: a standalone document for participants. Approval does not mean publication, and publishing the minutes does not grant access to the source recording.

Find a recording or a passage in a transcript. Ask the Assistant a question about selected materials available to the user and verify the sources behind its answer.

Example question: “Which questions remained open after the discussion?”

• Select specific accessible meetings, documents, or workspaces. Check the sources shown under “Assistant is reading” before asking a question; for example, keep two meetings selected to compare their outcomes.
• Include accessible documents in the question’s context.
• Refine the answer with follow-up questions and check its sources.
• Start a separate conversation from a selected answer in your own conversation to explore another direction. The original conversation remains available; generated files and confirmed actions are not copied into the new one.

Result: an answer based on work materials, beyond simply finding a file name. If the sources conflict or do not contain an answer, the conclusion needs clarification.

Use a personal or team collection of documents as a source for search and Assistant answers. A conversation recording is not required.

• Upload documents approved for processing to a folder and wait until they are ready for search.
• Select materials for a question, find a relevant condition, or compare information across several documents.
• Open the answer’s sources and check quotations against the text.
• Where configured Assistant workflows are available, perform recurring reviews of materials using a defined procedure.

Where a workflow workshop is available, an authorized specialist can prepare a workflow for a recurring task, test it with examples, and share it with colleagues. A shared procedure does not replace checking each answer.

Example: compare two sample descriptions of request handling and list differences in their steps and deadlines, citing the sources.

Optical character recognition (OCR) makes scanned documents usable for search and Assistant answers when the feature is configured at the site. Recognized passages must be checked against the original, especially numbers and complex tables.

Where recognition is configured, images such as photographs of text pages can also be added to the Assistant’s materials. Once processed, they can be searched and used for questions; check the results against the original image.

Important: uploading a file does not, by itself, mean it is ready for search. If several editions are stored, the edition containing the passage must be checked before drawing a conclusion.

Bring meetings and documents on a project or topic together in a workspace. Work with outcomes, action items, and open questions, and return to previous discussions.

Result: a shared record of work on a topic. When adding materials, check who will be able to access them.

• Collect project meetings in a shared feed.
• Keep workspace documents in a shared list or organize them into folders. Where transferring personal materials is available, add a personal file or folder after checking who will gain access.
• Discuss materials in a private or shared conversation, selecting the audience in advance.
• With a configured source, automatically add new recordings from a selected room.

Example: a project workspace contains previous team meetings, work documents, and discussions. Adding a recording may give workspace members access to its materials.

Save a personal note, set a reminder, and prepare for the next discussion using the accumulated materials. The specific automation options must be confirmed for the version in use.

• Return to the workspace’s unresolved questions.
• Ask the Assistant what to include in the next agenda.
• Review the proposed agenda and add current questions.

In the web interface, users can dictate an Assistant question or a personal note when voice input is available and the browser has permission to use the microphone. After recognition, they review the text and choose an action: send the question, save the note, or set a time and confirm a reminder. Speech recognition does not perform these actions on its own.

The Today page brings together upcoming meetings, new materials, action items, reminders, and recording processing status. It provides a convenient way to return to current work.

The My section brings together a user’s own uploads, saved accessible meetings, documents, conversations, notes, and reminders. An important meeting can be added to a personal collection for quick access. This bookmark does not copy the recording or retain access after permission is revoked.

Result: the next discussion starts with the context preserved. A meeting reminder is not the same as an automatically prepared, detailed briefing.

Connect an approved organizational calendar to see upcoming meetings on the Today page. CalDAV and Exchange connections are supported; available sign-in methods depend on the organization’s configuration.

• Select calendars to read and refresh their events.
• If needed, link an event to a meeting series to refer to its materials.
• Set a notification relative to the start or end of an event.

Result: the schedule helps users return to meeting materials at the right time. Thesis reads the calendar; events are still created and edited in the original calendar system. Linking an event to materials does not grant additional access to them.

With the required connection configured, a recording agent can be scheduled for a meeting from a calendar or an email invitation. In the personal Assistant, users can also ask an agent to join a selected meeting or provide a link and time. Before the request is saved in the chat, the user reviews and confirms the time and link. Whether the agent can join depends on support for the meeting platform and the configuration; check its connection status.

Where email integration is available and permitted by the organization, Thesis can display meetings from invitations and save attachments from new emails to selected folders according to configured rules. With the user’s separate consent, messages from the connected mailbox can be used as a source for questions in the personal Assistant.

The connected mailbox is read-only: Thesis does not send replies from it or modify its messages. Before saving attachments to shared folders, check who will be able to access them.

Automation can provide a reminder or prepare an answer based on selected materials. These are different results.

Capability	What the user receives
Reminder	A notification containing predefined text or a link, for example about a meeting, a review of minutes, or an overdue action item.
Personal Assistant agent	An answer based on selected sources, generated on a schedule, before a calendar meeting, or once its recording is ready; the result can be explored further in a conversation.
Workspace or organization agent	An answer prepared on a schedule for a specified audience, taking recipients’ permissions into account. Available sources depend on the type of agent.

An agent is configured with a task, sources, and a trigger. A post-meeting run depends on the recording being ready for the agent to process, rather than simply on the call ending. A person reviews the prepared answer; the agent does not approve or publish minutes on its own.

Notifications are available in the Thesis inbox. Depending on the type of automation and its settings, delivery through email, browser notifications, a personal messaging service, or a team chat may be configured. The available channels vary; before enabling external delivery, check the recipients and permission to share the content. A notification does not grant access to the source materials.

Result: users can return to materials according to a configured schedule or event. Calendar connections, agents, and delivery channels are enabled separately; their availability and suitability for a workflow must be checked at the specific site. If a run or delivery does not occur, its status can be checked in the automation history.

Administrators manage users and connected sources within their assigned permissions.

• Manage users, groups, roles, and access to materials.
• Connect recording sources and configure processing rules.
• Monitor processing status, investigate errors, and restart processing when permitted.
• View usage and processing statistics within their authorized scope.
• Configure media storage and arrange backups, updates, and recovery according to the procedures for the deployment.

Result: the organization manages access and system maintenance. An administrative role name does not, by itself, imply access to every recording. Usage statistics are not an assessment of employee performance; retention periods for media, text, and published documents must be considered separately.

The main Thesis interface runs in a browser and is supplied with the server. Visible actions depend on the version and the user’s permissions.

The separate Thesis Client is intended for working with recordings, uploading files, viewing results, and asking the Assistant questions. The documented Android and iOS capabilities include a voice recorder and local audio drafts that are explicitly uploaded to the server.

Application availability: before using a separate client for Android, iOS, a browser, macOS, or Windows, confirm whether an installation package has been released for the deployment, which server version it supports, and which actions are available.

A local audio draft is not yet a processed recording on the server. It does not imply full offline operation of every feature or synchronization between devices. The separate client and the built-in web interface may have different capabilities.

A verified transcript or summary can be exported to a file, including Word (DOCX) or PDF. Where the corresponding features are available, a result can be shared through a revocable link or published back to a connected Vinteo source.

Where Assistant file creation is available, a result based on selected materials can be prepared as a separate Word document or Excel spreadsheet. The reviewed file can be downloaded or saved to personal documents or a team workspace for later use. Check the audience before saving it to a workspace.

Before sharing, select a verified version and check the content, the recipients, and permission to distribute it. Publishing back to the source is a separate action; check its result afterwards.

A revocable link provides access to the permitted recording materials. Revoking the link ends access through it but does not delete materials already downloaded or copied.

• Quality depends on the recording, speech clarity, and processing settings.
• Assistant answers must be checked against their sources; a claim without supporting evidence must not be treated as an established fact.
• Permission to view material does not imply permission to share it with third parties.
• This overview does not guarantee specific processing times, accuracy, formats, or identical functionality across all devices.

Thesis · Usage examples · Getting started