Agentic Studio

The conversational and generative surface: chat, AI-run workflows, projects, and a drafting editor that all work over your data.

Agentic Studio (formerly AI Workspace) is the conversational and generative surface of the platform. It is where you ask questions in plain language, let the assistant run multi-step work, group related work into projects, and draft documents with AI help. Everything here works over your operational data, grounded in what you are authorized to see.

Chat

Chat is the front door. Ask a plain-language question and the assistant answers using your data.

Plain-language questions. Ask about your operation the way you would ask a colleague.
Multi-turn. Conversations carry context, so you can ask a follow-up instead of restarting.
Generated artifacts. Answers can include richer output such as tables, code, diagrams, maps, documents, slide decks, and images, opened in a side panel you can preview, copy, or download.
Model choice. A picker lets you pick the assistant's model, or leave it on Auto to let the platform choose based on the request.
Grounded and cited. When the assistant draws on the compliance corpus, the sources appear as citation chips beneath the answer.
Skills. Start from a reusable skill instead of a blank prompt. See Skills.
Stop and steer. Stop a running answer at any point, then edit your message or ask again.

For step-by-step use, see Using chat.

Projects

A project groups related chats under shared instructions and reference files, so you start fresh short chats instead of growing one long one. A project is a set of tabs around one body of work.

Overview. The project's instructions, which ride into every chat started inside it.
Tasks. A five-column board for the project's work, so a plan and the chats that carry it out live in the same place.
Memory. The visible, editable record of what the assistant has learned about the project. You can add to it, correct it, or remove an entry.
Chats. Every chat started inside the project.
Files. Reference files that become standing context for every chat in the project. Before anything uploads, you see a report of what will and will not be used.
Notes. Short written notes kept alongside the work.
Skills. The skills this project's agents may use. Pin a few to keep a project on a deliberate, short list.
Artifacts. The board of pinned output, so the results worth keeping are easy to find later.

See Projects.

Maps and locations

The assistant works with locations directly in chat: match an address to coordinates, compute a route and drive time between two places, compare drive times across several stops, and show the area reachable within a set drive time. Results render as an interactive map artifact you can pan, zoom, and pin to a project board. See Maps and directions.

Documents and exports

Output does not have to stay in the browser. Generated artifacts export to PDF, Word, Excel, CSV, and PowerPoint, and a chat reply can be exported to PDF or Word. Diagrams render inside exported documents rather than being dropped. Every export is recorded in the audit log, and the exported content itself is never stored by the export step.

Skills

A skill is a reusable capability the assistant can put to work: a set of instructions, the reference material that goes with them, and the tools the skill is allowed to use. Browse the catalog, read a skill's full instructions before you run it, run one in a fresh chat, or pin a short list of skills to a project. Tenant admins can add new skills through a draft-and-approve review flow. See Skills.

The editor

The editor is a rich-text drafting surface with AI assistance built in.

Format text, build tables, and insert images.
Use AI completion to continue, fix, extend, reduce, simplify, summarize, or translate a selection.
Your work autosaves as you type.

See The editor.

AI-run workflows and automation

The assistant can do more than answer. For a request that spans specialties, the Delegator hands parts of the work to specialist agents and folds the results back into one answer, each delegated step shown as it starts and finishes.

Your workspace starts with a roster of named specialists (General, Builder, Analyst, Scribe, Planner, Researcher, Illustrator, Compliance, and Ops, coordinated by a project manager agent). You can pin one specialist for a chat instead of letting the request route itself.

The roster is yours to shape, not a fixed list. A tenant admin can edit an agent's persona and the model it runs on, duplicate an agent, add a new one, or disable one the team does not need. Every edit goes through a review step before it takes effect, and each agent card shows whether it is stock or customized.

Connector-driven automation (scheduled and event-triggered syncs) keeps your data current in the background; that side is documented under Integrations.

How answers stay grounded

The assistant is grounded in the data you are already authorized to see. It does not get a wider view than you have, and it does not reach beyond your authorized data. When it cites regulatory facts, it shows the source so you can check it.

What it is not

It does not make dispatch decisions for you. It surfaces information and drafts work; you decide.
It does not reach beyond your authorized data, and it never crosses into another organization's records.
It is not the compliance authority. For cited, tier-ranked regulatory facts and screening, that is Compliance Link.