AI agent

Your team knows how. Let agents share the workload.

When routine tasks fill the day or requests suddenly pile up, hand part of the work to Recordable’s AI agents. They use your team’s guides to work in your apps. You review the plans and supervise the work.

Requires the desktop app · macOS and Windows

Example · refund a returned order
Refund order #5012, they sent it back.
Process refund · order #5012
Click Refund, reason: returned6s
Click “Refund €89.00”Click

Verified the charge from Apr 14. I’ll wait before issuing the refund.

Auto-Accept: OFF
Details
Try approving or rejecting this example. No real actions are taken.
From know-how to action

Start with your guide. Make it today’s task.

01

Choose the task

Pick a published guide and tell the agent what you need today: a different customer, order, or set of details.

02

Review the plan

The agent adapts the guide to your request. Read the proposed steps and edit the instructions before you start.

03

Supervise the work

With Auto-Accept off, review proposed actions and approve or reject them as the agent works in your apps.

04

Check the result

Review the outcome in the app. The run history keeps the steps, actions, and screenshots for you to inspect.

01Reuse what your team knows

More requests. The same process.

A rush of returns or a queue of onboarding requests: hand the repeatable steps to agents using your team’s guides. Change the details for each request, review the plan, and keep your team focused on the cases that need their attention.

  • 01Use the same guide for the next customer or request
  • 02Edit the proposed instructions before starting
  • 03The agent reads the screen to find where to act
  • 04For a list of items, batch runs repeat the task per row
Your instruction
Refund order #5012, they sent it back.
Adapted plan3 of 5 steps adapted

Example plan · review the instructions before starting.

1Search for order #5012adapted
2Open order #5012adapted
3Click Refund, reason: returnedadapted
4Email the customer
5Close the order
02Stay in control

You decide what happens next.

With Auto-Accept off, the agent shows proposed actions and waits for approval. Approve or reject them as it works. Turn on Auto-Accept when you’re ready for actions to proceed without individual approval.

  • 01See the proposed action before approving it
  • 02Choose whether to enable Auto-Accept
  • 03Stop the run when you need to intervene
Process refund · order #5012
Click Refund, reason: returned6s
Click “Refund €89.00”Click

Verified the charge from Apr 14. I’ll wait before issuing the refund.

Auto-Accept: OFF
Details

Try approving or rejecting this example. No real actions are taken.

03Work in your existing apps

Do the work where it already happens.

The agent acts on screen through Recordable’s desktop app for macOS and Windows. It can work across browser and desktop apps using the sessions you already have open.

  • 01Start runs from the Recordable desktop app
  • 02Carry out steps across browser and desktop apps
  • 03Use the accounts you’re already signed into

AI runs use automation credits. See pricing for the trial allowance, included credits, and simultaneous run limits.

Recordable Desktop
Run progress
Process refund · order #5012
Search for order #5012
Completed
Open order #5012
Completed
Click Refund, reason: returned
Running
macOSWindows
Start with a task you know

Delegate the steps. Check the outcome.

Read about security

A good first task has a clear guide and a result you can check. Keep the first run supervised, see how the agent handles it, and decide what you’re comfortable handing over next.

Give it a clear starting point

Review the guide and update any outdated instructions before asking the agent to use it.

Make time to review

AI can make mistakes. Check the proposed actions and the final result; use the run history to understand what happened.

Know where data goes

Actions happen on your computer. Screenshots are sent to the selected AI provider to guide the agent. Read about security for the details.

Start with the work that keeps coming back.

Pick a repeatable task that fills your queue. Record it, check the guide, and try handing part of the workload to an agent.