AI Transparency

• Company Bot answers questions about your company’s policies, places, and procedures, how the app works, and your own records, such as your hours this week, your shifts, and your time off. It also does everyday tasks you could do yourself in the app, such as requesting time off, cancelling your own request, or picking up an open shift, once you confirm them. It never answers questions about another person and never acts for anyone else.
• AI Automations are workflows your workspace builds: when something happens at work, such as a shift going unfilled, a time off request coming in, or someone joining, a set of steps runs on its own. Most steps follow fixed rules the workspace sets. Some steps use AI: writing a message, answering a yes or no question to choose which path a run takes, or planning steps with Ask AI.
• Ask AI, part of AI Automations, turns a request in plain words into a plan with one step per tool, such as requesting time off or finding cover for a shift, and does the steps once you approve.
• Chat translation shows Team Chat messages in the language each reader picks for themselves, with the original a tap away. Each message is translated once into the languages its readers picked. An admin can turn it off for the whole workspace.
• AI Scheduling turns a manager’s description of the week into shifts to fill, then fills them.
• Course Creator drafts a training course from a topic or a procedure, and rewrites a module more simply.
• AI Reports and summaries summarize survey comments, and the weekly brief for managers groups the questions Company Bot could not answer into articles to write.

Under Google Cloud’s terms for Vertex AI, Google does not use what we send to train or improve its models. We do not train models on your workspace either. If we change a model or a provider, we update this page and our Subprocessors list first.

Each request carries only what that feature needs, and only what the person asking is allowed to see:

• Company Bot: your question, your own earlier questions in the conversation, your role, the articles, announcements, events, and requests you are allowed to read, and your own facts, such as your hours, schedule, time off requests and balances, and training. When you ask it to do a task, it also sees the choices that task needs, such as your leave types or the open shifts you could pick up.
• AI Automations: only for a step that uses AI, the words the workspace wrote for that step and the facts of the run it belongs to, such as the person’s name, the shift, or the request. Steps that do not use AI send nothing to a model.
• Ask AI: your request and any photo or PDF you attach, and the workspace facts a plan needs: the team you can see, sites, departments, positions, leave types, upcoming shifts, and, for managers, what is waiting for their decision.
• Chat translation: the text of a message and the languages its readers picked. Not the sender’s name, the chat’s name, or any photo or file.
• AI Scheduling: the manager’s description of the week, the workspace’s positions and sites, and how many people hold each position. People are then placed on shifts by fixed rules (availability, rest between shifts, weekly hours, certifications, and fewest hours first), not by the model.
• Course Creator: the topic or text the manager gives it.
• Summaries: the comment text of a survey (without names; anonymous surveys never link a person to an answer), and the questions Company Bot missed.

The models never see pay rates, bank details, tax forms, passwords, or anything from another company’s workspace.

• Company Bot does not store your questions or its answers. When it cannot answer, it saves the question and who asked so the workspace’s managers can write the missing article. When you confirm a task, it keeps a record of what it did, when, and the result, visible to you and your managers.
• Chat translation keeps each translation with its message, beside the original, for as long as the message is kept. Editing a message replaces its translations, and deleting it removes them.
• AI Automations keep a record of every run: who or what it was about, each step it took, when, and what happened, visible to the workspace’s managers.
• Ask AI keeps each request, its plan, and what each step did, visible to you and your managers, so there is always a record of what it did for you. Steps can be undone where the tool allows.
• AI Scheduling and Course Creator save what they make as drafts in the tool, like anything a person writes.
• We log how much text each request used, for cost and limits, not what was said.

• Ask AI changes nothing until the person asking approves its plan, and each step runs with that person’s own permissions. It cannot do what they could not.
• Company Bot does a task only after you tap Confirm on it, and it runs with your own permissions, so it cannot do anything you could not do yourself. Where the app allows, you can undo it.
• AI Scheduling makes drafts; only a manager publishes a schedule. Course Creator makes a draft a manager edits and assigns.

AI Automations run on their own once someone with permission publishes them, and keep running until they are switched off. Each step runs with the permissions of the person who published the automation, so an automation cannot do what that person could not. Steps that approve time off or timesheets, decide timecard changes, publish a schedule, remove a shift, change certifications, or invite people ask a person to approve first, unless an admin changes that. An admin can set each kind of step to ask first, to run on its own, or to be off. An automation can be tried on one person without changing anything, and any run can be stopped at any time.

No AI feature hires, fires, disciplines, rates, or sets the pay of anyone. Time off that approves itself does so by a fixed rule the admin sets (a request this many hours or shorter), not by AI. Where an automation uses an AI step to choose a path, that path can only do what the workspace built into it, and the workspace is responsible for reviewing what it does, as our Terms of Service require.

AI features are limited to 60 requests a person an hour, and to a daily amount per workspace that grows with the number of active people. The limits keep costs fair and stop misuse.

If your workspace allows it, you can connect your own AI assistant to Vantage Workforce through our assistant connector (the open Model Context Protocol). The assistant then acts as you: it can read your schedule, time off, hours, and training, search what you can read, and, if the workspace allows actions, request time off or send a chat message for you. The assistant is run by its own company under its own terms. You can disconnect it at any time.

Language models can be wrong while sounding sure, can miss something in a long document, and can misread a request. Check AI output before you rely on it, especially anything about pay, hours, or policy. If an answer is wrong, tell your admin so the source article can be fixed.

Questions about our AI features: [email protected]. Our Privacy Policy covers how all personal information is handled.