Runwork
All Use Cases

Visible AI work

AI working where nobody can see it

Every important action in your workspace is written down and not edited afterwards. When somebody asks what the AI did, there is a record rather than a recollection.

Powered By

Audit Logs
Oversight Permissions
Integrations
MCP Servers Workspaces

The Challenge

  • People are using AI on company work and nobody has a list of where
  • Somebody connected an AI tool to the CRM months ago. You find out when something goes wrong
  • Your board asks what AI is doing in the company, and the honest answer is that nobody knows
  • The AI did something six weeks ago and there is no way to go back and see what
  • Every answer you give about AI in your company is somebody's recollection of it

The Runwork Solution

  • Every important action is written to a record that is not edited afterwards
  • You can search that record by person, by what happened, or by when
  • The AI tools, integrations and MCP servers connected to the workspace are registered, so the inventory is a list rather than a memory
  • Each entry says who did it, when, and what changed, including the values before and after
  • Everyone has a role, and the role decides what they are able to do

What your team gets

The question you can now answer

Somebody asks which AI tools reach your customer data. That is a list you open, rather than an email thread asking around the office.

The change nobody remembers making

A permission changed, or a record did. The entry says who did it, when, and what the value was before it changed.

An inventory that exists on its own

The AI tools, integrations and MCP servers in the workspace are registered as they are added, so the list is there whether or not anyone kept one.

What you read is what was written

Entries are added and not edited afterwards, so an entry from six weeks ago says what it said six weeks ago.

Something to hand to a reviewer

When a security questionnaire asks how you know what happened, the answer is a record that already exists rather than one assembled for the occasion.

Roles you can see and change

Everyone in the workspace is an owner, an admin, an editor or a viewer. You can see which, and somebody with the right role can change it.

A Day in the Life

9:00 AM. A question from someone senior. Which AI tools can reach customer data? Previously that would start a week of asking around. The connected integrations and MCP servers are registered in the workspace, so it is a list, and you read it.

10:30 AM. Something changed and nobody owns up. A record has the wrong value in it. You open the audit log, filter to that record, and the entry says who changed it, when, and what it was before. That closes the question in a minute instead of a meeting.

12:00 PM. A new person joins. You give them a role. Owner, admin, editor or viewer decides what they are able to do, and the change itself is written into the record like everything else.

2:00 PM. The security questionnaire. A customer's security team asks how you would know if something happened. You do not have to build an answer, because the evidence they are asking about has been accumulating since the day you started.

4:00 PM. Something from six weeks ago. You need to know what a workflow did in July. The entries are still there and they still say what they said in July, because entries are added and not edited afterwards.

5:30 PM. Nothing was reconstructed today. Every answer you gave came out of a record that was already written, rather than out of somebody's memory of it. That is the difference this page is about.

Frequently Asked Questions

How do I know which AI tools have access to our workspace?
They are registered as they are connected. The AI tools your team uses, the integrations connected at the workspace and the MCP servers available to them are all listed, so the inventory exists as a consequence of people connecting things rather than as a document somebody has to maintain. That is the difference between an inventory that is current and one that was current.
What exactly gets recorded?
Important actions: creates, updates and deletes on your data, team and permission changes, authentication events, API calls, and automation runs. Each entry carries who did it, when, and what changed, including the previous value. Ordinary reading is not recorded, so nothing is written when somebody opens a record or runs a query. Some read-side actions are recorded deliberately, because they are the ones you would want kept: opening or downloading a conversation somebody shared, and exporting your data or the log itself.
Can somebody change the record afterwards?
No. The log is append-only: entries are written once and are not edited later. That is the property that makes it worth anything as evidence, because a record that can be tidied up is a record that has to be trusted rather than read.
Do you support single sign-on?
Not today. It is on the roadmap, and we would rather say so here than let a page about knowing what happens in your workspace imply an authentication feature we do not have yet. What ships now is access control set per resource for a person, a group or a department, and every access change written into the record.
Who can do what?
You decide, per thing. A person, the groups they are in and their department can each be given or refused access to a particular app, a set of records, a connected integration. Roles remain as starting points so nobody begins with an empty grid. This page is about the record rather than the controls, so the short version is the one that matters here: whatever you decide, and every later change to it, is written into the log like any other action.

Related Use Cases

Who Uses AI What the Money Bought Work That Runs

Start keeping what your team figures out.

Start a 14-day free trial. No credit card required.

Start Free Trial Explore More Use Cases