Your A2I review loop, run by a trained team
Amazon A2I is closed to new customers, and its Mechanical Turk workforce closed on 30 September 2026. Keep your loop with our team as its private workforce, or move review into your own tools.
$5 per person-hour A plan within 2 working days No sales call
The reviews you ran in A2I
A2I sent the uncertain few to people. The same work goes to a trained team, with a reason for every decision.
What changes when you move
Your model, your conditions and your task templates stay the same. Who reviews, and who checks the reviews, changes.
| Topic | A2I with the MTurk workforce | Deepen AI Services |
|---|---|---|
| Who reviews | Anonymous workers, different from task to task | The same trained team, with a named team lead |
| Instructions | Read once from the task template | Trained into the team, and updated with your lead |
| Quality | Checks you built yourself | Known answers, QA checks and a weekly report, run by us |
| Unclear items | Each worker decides alone | The team lead decides, or asks you |
| How you pay | Per task, per reviewer | Per reviewer hour: $5 for standard review |
Two ways to move
Choose the one that fits your pipeline. Both start with the same trained team, and you can start with one and switch later.
Keep A2I, change the workforce
Our people join a private work team in your account. You create a new human review workflow with the same task template, and your code, conditions and S3 output carry on.
Ground Truth and A2I workforceMove review into your own tools
Your code writes uncertain items to a review queue in your own tool, or to a shared sheet or bucket. Reviewers write decisions back, and your code reads them on its next run.
How work reaches the teamYour data stays in your systems. We work under NDA, with the access you set. Deepen AI holds SOC 2 Type II, ISO 27001, ISO 9001 and TISAX.
Who checks the reviews
With the Mechanical Turk workforce, quality checks were yours to build. Here, a team lead and QA reviewers run them, and you see the results every week.
Every reviewer's accuracy is measured
Items you've already answered are hidden in the real work, so we know how accurate each person is.
Nobody guesses on unclear cases
The team lead decides, or asks you. Your answer joins the guidelines.
You see the results every week
A weekly report shows work done, QA results and what's improving.
How we start on your work
No sales call. Your task template and examples become the team's training.
What to bring from A2I
- Your task template and reviewer instructions
- Examples of reviews you trust
- Your daily volume, and how fast decisions must come back
- Which route you want: keep A2I, or move review into your tools
- 5 minutesTell us the task
What your loop reviews, how many items a day, and how fast you need decisions.
- Within 2 working daysGet a plan and a price
Reviewers, hours covered, the route, start date and the total price, by email.
- 2 weeksRun the pilot
We train the team on your template and examples, set up your route, and review real items with two weekly reports. Then you decide.
Pricing
$5per person-hour
Standard review in English. You pay for the hours reviewers work, with a trained team, a team lead, QA checks and a weekly report included.
For example, if one reviewer checks 90 document extractions an hour, that's about 6 cents a review. Your plan shows the numbers for your loop.
Set before you pay. Your plan shows the team size, hours covered and total price.
Starts with a two-week pilot on your real items, paid upfront, with two weekly reports.
Then you decide. After the pilot, you choose whether to continue.
Questions
Is Amazon A2I shutting down?
AWS says A2I and SageMaker Ground Truth are no longer open to new customers, and no new features are planned. Existing customers can keep using them. What has gone is the Mechanical Turk workforce, which closed with MTurk on 30 September 2026.
Can we keep our A2I human review workflows?
Yes. Create a new human review workflow with the same task template and our team as its private work team, then point your code at it. Your conditions and S3 output stay as they are.
Why do we need a new workflow?
The SageMaker API can't change the workforce on an existing flow definition, so moving to a private team means a new workflow. Its new name also changes the S3 output prefix, so update anything that reads it at the same time.
Should we keep A2I or move off it?
If your loop works and you're an existing customer, keeping it is the quickest move. If you're planning for the long term, moving review into your own tools means you no longer depend on a service that's closed to new customers. We can do either, or one and then the other.
Do you handle the built-in Textract and Rekognition reviews?
Yes. Those tasks arrive in the labeling portal like any other, and the reviewer's answer goes to your S3 output as before.
How do reviewers know when A2I tasks arrive?
AWS documents task notifications for Ground Truth but not for A2I, so your team lead agrees a schedule for working the portal that meets your turnaround, and it's written into your plan.
How fast do decisions come back?
You set the turnaround and the hours you need covered when we plan the work. We staff the team to meet them, and both are written into your plan, with the price, before you pay anything.
Keep your review loop running with a trained team
A plan within 2 working days. No sales call.