Built into every workspace
Your agents and teams, in one shared workspace.
Every Vocion workspace comes with its agents and teams. Ask a question or hand over work, and the right agent takes it. Your people see the same work the agents do.
- 01
Ask any workspace
Ask a question or hand over work, in chat or in Slack.
- 02
The right agent takes it
A team's lead routes the request to the agent that owns it.
- 03
Answers cite what they read
Here, the contracts folder in Google Drive.
Teams
A lead agent routes the work.
A team is a group of agents. Give it a lead to route requests to the agent that owns the work, and a person accountable for it, or let it inherit the workspace owner.
- A lead routes the work
- The lead answers for the team in chat.
- An owner when you want one
- Set per team, or inherited from the workspace.
Review
The work that needs a person comes to one place.
Approvals, questions and recommendations land in Needs you. Each one says which agent recommends it and why, and cites the documents it read.
Browser
Work on sites that have no API.
Some of the work lives behind a sign-in: a vendor portal, a government site. Ask for it in chat, or tag @browser, and an agent does it in a browser while you watch beside the conversation. What it finds comes back as a table. An automation does the same task for every row of a list, each row in a browser of its own.
- It stops before it changes anything
- Anything that saves, sends, pays or deletes waits for your approval. Since 6.0, so does creating, inviting, enrolling, archiving or activating.
- Passwords stay out of the model
- The browser types them. The agent only sees a placeholder.
Follow-through
Finish what is owed from where you read it.
A follow-up, a to-do or an owed touch opens with its verbs. Do it hands it to the team, and anything that goes out to someone waits for you. Done and Drop close it with one click and Undo. Records listed under an answer can be marked done without opening them. The agent closes records the same way.
- Its age first
- An owed thing from three months ago calls for a different decision than one from yesterday.
- Your type decides
- Mark a type x-owed: true in its YAML. Products and companies stay as they are.
Read
What is owed, and how long ago.
Do it
The team drafts it. You approve what goes out.
Done
Closed with Undo, and on its history.
Record
Every run keeps a record.
What the agent did, what it read, who approved it and what it cost. Open it when someone asks why. A team's records are open to that team.
How a role runs
One loop runs every agent.
Define the role. Run the work. Review where judgment matters. Measure the result. Improve from what reviewers approve.
- 01
Define
A role, its tools, its rules and an owner.
- 02
Run
It does the work in your systems.
- 03
Review
A person approves what matters.
- 04
Measure
Quality, cost, cycle time, outcome.
- 05
Improve
Edits become rules a person signs off.
- Then back to Define, with what the team learned.
Define
Agents, teams, tools and rules are YAML in your git. Agents answer from the documents and records you connect, and cite them.
Review
Anything that leaves the building waits for a person. Autonomy is earned per action.
Earned autonomyImprove
An edit becomes a proposed rule. A person approves it before any agent uses it.
The learning loopRun
Your whole workforce, in one place.
A roster of who runs what, with an owner, a budget, quality and status. Everything that needs a person lands in one list.
- Support ReplyOwner · Dana64%92%Running
- Sales Follow-upOwner · Sam38%88%Running
- Weekly ReportOwner · Priya12%95%Paused
Owner · budget used · quality · status
- Reply to Acme about the late orderApproval
- Which price list for Contoso?Question
- Renewal reminders paused by DanaPaused
- Let CRM updates run on their ownRecommendation
An owner for every agent
Set per team, or inherited from the workspace owner.
A budget for every agent
Soft and hard caps per agent, daily or monthly, plus one cap for the whole workspace.
One list for people
Approvals, questions, recommendations and paused work, each with who is asking and why.
Measure
Measured like any team.
Manage agents on performance, not usage: quality, cost, cycle time and outcomes, per team.
| Team | Quality | Cost / run | Cycle time | Outcome |
|---|---|---|---|---|
| Support | 92% | $0.38 | 40 min | 81% same day |
| Revenue | 88% | $0.52 | 2 h | 14 follow-ups sent |
| Finance | 95% | $0.21 | 1 day | Close on time |
Illustrative. Sample data, not a customer result. Run a starter and measure your own week.
Quality
How often reviewers approve as is, and eval scores on your own cases.
Cost
Model, tokens and cost on every run, against the agent's budget.
Cycle time
From request to done, including the wait for a person.
Outcomes
The number the team already tracks, like same-day resolution.
Turn it on
Already on in every workspace.
Agents and teams are the core of every workspace, so there is nothing to install. Your agents, teams and the apps a workspace adds are declared in workspace.yaml and applied from git.
Connects to
- Slack
- Google Drive
- Gmail
- Notion
- HubSpot
- GitHub
- All integrations
# workspace.yaml
plugins: [wiki, data-rooms]
# Agents, teams and chat are always on.
# Plugins add capability, and the apps that own them.Personal and shared
Every person gets a private workspace. Teams share the rest.
Your models, your keys
A workspace can use its own model key and is billed on its own vendor account.
Ready-made agents
Hire a controller, an operations analyst or a people ops partner from the Marketplace.
Get started
See Vocion run your team's real work.
Bring one job your team does every week. In the demo we walk through it on sample data shaped like yours, with the record it keeps.
For your team: a demo built around your job, on sample data.
For your engineers
Run it yourself
The requirements, the AWS path and the repo.