AI Assistant
Ask questions about your data in plain language and get answers, charts, and files.
Intelligent AI assistance, right inside Plex.
Conversational intelligence that knows your Plex context — ask questions about your data and get instant, grounded insight without leaving your screen.

Where you meet it
Open the assistant from Assistant in the DataMagik navigation, or from the datamagik extend browser extension while you are working inside Plex. Ask a question in plain English and it answers — no SQL, no scripting.
An assistant can be made screen-aware: give it the live Plex screen as readable JSON and it can answer about the record actually in front of you, rather than asking you to describe it. That is a deliberate setting, not a default — an assistant whose message template does not reference the screen context never receives it.
Replies come back as text, as rendered charts, or as files you can download. Conversations are kept on your own device for one day, and clearing one is a single click.
Not one assistant — as many as you need
Administrators create assistants and pick, per assistant, its persona, its model, who is allowed to talk to it, and which capabilities it is offered. People choose between them from a selector at the top of the chat.
A system message sets who the assistant is and how it behaves. The model list reflects your company's own active AI key, and defaults to your provider's default when you would rather not pin one.
Tick only what this assistant should be able to reach: scripts, SQL and ODBC, Plex ERP, lookup tables, local databases, dashboards, applications, automations, schedules, email templates, document templates, allowed domains and documentation. The model is offered exactly the boxes you ticked — nothing else.
Beyond the broad families, you can expose specific scripts as tools, so an assistant can run a named, reviewed piece of your own logic rather than improvising one.
Restrict an assistant to holders of a particular role, or leave it open to everyone. Chatting at all requires the Builder or Viewer permission.
Governance
Assistants run on the AI provider key your company configures. No external workflow tool sits in the middle, and without an available model the assistant simply does not answer.
An assistant's autonomy is a setting. On Supervised it asks before it performs a write, so the model proposes and a person decides.
Writes performed by an assistant land in the AI Audit Log. Where credentials are exposed as a tool family, the assistant sees credential names only — never a secret value.
The two are easy to confuse, so here is the line between them. DataMagik AI is the assistant your team talks to during the day: it answers questions about live data, explains what it is looking at, and — inside the tool families you allowed — takes action. Its output is an answer, a chart or a file, in a conversation.
DataMagik Invoke is the builder. You describe an artifact and it produces a durable one: a report, a dashboard, an application, or a scheduled emailed report that keeps running after the conversation ends.
Scoped to the data you choose, running on your own AI provider key, with every write on the record.
Book a demoAsk questions about your data in plain language and get answers, charts, and files.
Configure the native end-user assistant: its persona, model, who can use it, and which tools it may call.
Bring your own OpenAI, Anthropic, or Google Gemini key to power the AI Assistant and AI features.
See how many AI tokens your company is using, the estimated cost, and which users and models drive it.
Review every write action the AI assistant has proposed — executed or rejected — with who, when, and what.