Skip to main content
For AI agents: a documentation index is available at https://docs.coverbase.com/llms.txt. This page is also available in markdown by appending .md to the URL.
This page is part of the User Guides collection and the Coverbase AI agent (Beta) set. It sits beside The Coverbase AI agent (Beta), Controls, approvals and governance and Example walkthroughs. To start a run, see Working with Coverbase AI.
Coverbase AI is in Beta. Your Coverbase team turns it on for your organization.
Coverbase AI can read, create and change records in almost every part of Coverbase, as you and with your permissions. Each example below is an instruction you can type, what Coverbase AI does with it, and what it stops for you to approve. The mistake people make most often is writing an instruction that names no record. “Review the findings” leaves Coverbase AI to search. “Review Snowflake’s open findings”, or the same words typed on Snowflake’s page, starts on the right record.

How to read the tables

The You approve column uses four values. They describe a run that asks before writes, which is how a run starts unless you turn on Don’t stop to ask. The details of each value are on Controls, approvals and governance. Every example also needs the permission the matching page needs. Without it, the step is refused and the report says so.

Playbooks

Coverbase AI picks a playbook for each instruction, or writes a custom plan when none fits. The playbooks are listed under Playbooks on the Coverbase AI home page, each with a ready-made instruction.

Questions and research

Coverbase AI answers a question without changing anything. It reads the record you are on first, then your other records, then the public web when the question reaches outside Coverbase. Each record it cites is a link, and each claim from the web names its source.

Vendors and services

Intake and onboarding

When Coverbase AI fills in an intake request, it asks you only what your organization alone knows, such as what data the vendor will hold, at most three questions. It answers questions about the vendor’s own practices from research, or leaves them blank with the evidence that would answer them.

Inherent risk and risk scoring

Assessments

Reviewing results and controls

Questionnaires

Coverbase AI never edits a vendor’s response itself.

Control sets and risk domains

Each control set checkpoint says how many assessments in progress use the set.

Findings and remediation

Contracts, obligations and the risk register

Radar

A reassessment Coverbase AI starts from a signal is a draft. It sends nothing, and you pick the assessment plan and run it.

Workflows

In Watch it work, the Workflows editor shows each step being added while the run builds it.

Dashboards, views and exports

When no chart can count what you asked, Coverbase AI says so and gives the numbers from the records instead.

Work queue, notifications and collaboration

Documents

Coverbase AI reads PDFs, Word documents, workbooks, CSV files, emails and text files you attach, and says so when a file has no text it can read. See Files and records.

Screening and offboarding

Configuration (admins)

Each change needs the permission its settings page needs. Coverbase AI reads the current value first and says the change as old against new.

Many records at once

A set of more than 25 records for child runs, or more than 50 for a batch, is narrowed with you first.

Getting started and standing work

To teach every later run a rule, type it to a live run starting with “always”, “never” or “from now on”, or add it under Configuration > Coverbase AI Memory. See Memory. A standing agent runs a job like these on a schedule or when something happens, such as a Radar alert or an overdue finding, and a workflow step Ask Coverbase AI to… can hand a record to one. See Standing agents.

Where else you can start or answer a run

  • Slack and Microsoft Teams. A direct message that starts with Coverbase AI: starts a background run as you. When a run needs you, the approval can reach you there.
  • Email. A needs-you email has one button per option, and you can reply to it in your own words.
  • Your AI assistant. The MCP server starts, answers, steers and stops runs.

What Coverbase AI does not do yet

These modules are not available to Coverbase AI in Beta: RFPs, Inspect, risk scenarios, sending a contract for signature, and allowlisting a screening match. A run that touches one says so in its report and does the rest of the job.

Troubleshooting

Example walkthroughs

Six end-to-end runs, step by step.

Controls, approvals and governance

What each approval value means and who sets it.

Working with Coverbase AI

Start a run, answer what it asks, steer it and read the report.