Chat

Chat is Mindfront’s primary direct control surface. It is where people instruct the AI, inspect how work is progressing, correct its course, and collaborate with it through text or voice.

Not every job begins in chat. Work can start from a familiar surface such as email, from a business event, or from a standing order. Chat remains the common place to step into that work: ask what happened, change the goal, supply missing context, or work alongside the AI when a job needs closer human involvement.

A significant unique feature is that chats in Mindfront can take place with multiple users in the same conversation with Mindfront. This allows multiple people to collaborate around the same AI-driven work.

When work needs more than a conversational answer, Mindfront starts a job. A job is a durable run of AI work with a purpose, context snapshot, complete action history, approval state, outcome, and reflection. Jobs can begin in chat or elsewhere in the business. See How Work Starts for the complete model.

Drafts & Approval

When Mindfront proposes an action that interacts with the outside world, like sending an email or updating a CRM record, it prepares a Draft and sends it through the risk evaluation engine.

A Draft is a proposal of the action to be taken. The risk engine scores it against organizational policy and routes it to automatic execution, machine review, human approval, or refusal. When human approval is required, the authorized reviewers can approve as-is, edit before approving, or reject it and give Mindfront new instructions. See Teams & Access for how reviewer access works.

Example Job Process for Email

1. A new customer query arrives via email, creating a business event.
2. Mindfront evaluates the event and starts a job to handle the query.
3. Mindfront uses its available knowledge and tools to compose a response.
4. Because sending an email is an external action, Mindfront prepares a Draft containing the proposed reply and submits it to the risk engine.
5. The risk engine evaluates the recipient, content, authority, and organizational policy. In this example, it routes the draft to a human reviewer.
6. An authorized user reviews the draft. They can approve it as-is, edit the content, reject it, or give Mindfront new instructions.
7. Once approved, the draft executes and sends the email reply to the customer.
8. The job completes, with its decisions and actions retained in the business history.