Integrations

The team works where your items already are. Most work needs no new software, and no data leaves your systems unless we agree it in writing first.

Get a plan

Inside your own tools

Your helpdesk, CRM, store admin, back office or labeling tool. Our people work your queue directly, and results land where your process already expects them.

Best for: Support, catalog, records and review work that already lives in a tool.
How it works: You create accounts for our people, with only the permissions the task needs
The team works your queue, following your guidelines
You switch off access whenever you choose
You set up: Accounts and permissions
Your guidelines and examples

In Deepen AI's tools

Annotate, Calibrate, Curate and Validate: Deepen AI's own platform for image, video, LiDAR and multi-sensor data. Results export to ASAM OpenLABEL, COCO, KITTI, Pascal or your own schema.

Enterprise programs

Best for: Labeling, calibration, curation and validation of sensor and multimodal data.
How it works: We agree in writing how data is shared, where it's stored and when it's deleted
The team works in Deepen AI's tools
Results come back in the format your pipeline reads
You set up: Data sharing terms
Your taxonomy or rubric

In SageMaker Ground Truth or Amazon A2I

Our people join a work team in your private workforce and sign in to your labeling portal. Your task templates, S3 data and code stay as they are; only the workforce changes.

Ground Truth and A2I migration guide

Best for: Existing AWS pipelines, including those that used the Mechanical Turk workforce.
How it works: You add our people to a private work team
You point labeling jobs and review workflows at that team
Results land in your S3 output, as before
You set up: A private work team
New A2I workflows for the team

Through a shared sheet or storage

Your process writes items to a spreadsheet, folder or storage bucket you share with the team. The team works them and writes results back in the columns or format you set.

Best for: Batches, backlogs, and pipelines that already write items to a file.
How it works: You share the location with the team
Items arrive there on your schedule
Results are written back for your process to pick up
You set up: Access to the shared location
A column or file format for results

When your code produces the work

Pipelines that create work in code connect to the team in one of three ways.

Write items to a shared sheet or bucket: Your code writes each item; the team writes the result back; your code reads it on its next run.
Use a review queue in your own tool: If your system already has a screen or queue for items that need a person, our people work it through accounts you control.
Keep Ground Truth or A2I: For existing AWS customers, your code keeps calling the services it already calls, with our team as the workforce.

Access stays yours

Accounts you create, with only the access the task needs
Access you can switch off at any time
Data stays in your systems unless we agree a move in writing
An NDA before we see real data

Security and data handling

Questions

Can our code send work to the team?

Yes. Your code writes items to a shared sheet or bucket and reads the results back, sends them to a review queue in your own tool, or keeps calling Ground Truth or A2I with our team as the workforce.

Which tools can your team work in?

Most web-based tools your team already uses: helpdesks, CRMs, store admins, back offices, spreadsheets and labeling tools. Tell us yours when you request a plan, and the plan says how the team will work in it.

Do we need to install anything?

Usually not. The team works through accounts you create in the tools you already use.

How do results come back to us?

Where your process already expects them: in your tool, in your S3 output for Ground Truth and A2I, or written back to the shared sheet or bucket.

Can we use more than one route?

Yes. A program can mix them, for example review in your helpdesk and a weekly batch through a shared bucket.

Tell us where the work lives

Get a plan

A plan within 2 working days. No sales call.