DigitalOcean, a cloud platform company, has announced tools meant to help AI assistants learn from their own mistakes. The tools record what an assistant does, flag where it falls short, and test fixes before real customers run into the same problem. Some are open for developers to try now. Others are limited to select users.
What changed
DigitalOcean calls these software helpers agents. The company says it built earlier products so agents could think and act. The new pieces are meant to help them improve. Learning Services has three parts:
- Insights records what an agent does.
- Signals spots failures and performance gaps.
- Simulations tests possible fixes in realistic practice conversations before they go live.
One commenter described the goal this way: a failed user session can become a fix before the four-hundredth person hits the same wall.
The data side
Agent Data Services is a managed place to store what an agent works with. It processes and indexes documents, video, audio, logs and agent sessions, then makes them searchable in one catalog. DigitalOcean says it includes built-in governance and access controls, meaning rules about who can see what.
It also feeds the learning tools. Signals writes sessions into it. A tool called Optimizer pulls training and test data from it only where the customer expressly opts in.
A separate upgrade, the Inference Router, picks the AI model best suited to each task. It balances cost and quality, and it reruns the same tests when a new model version arrives.
Who can try it
Insights, Signals and Evaluations, which include Simulations, are in public preview. That means developers can try them now, though the products are still being refined. Inference Router optimization and Agent Data Services are in private preview.
What to watch
Business owners using AI chat helpers could see fewer repeat mistakes. One commenter, Jonathan Saring, noted that once records of an agent's actions feed learning, they stop being just logs and become production data. That raises practical questions about where customer information is kept and how it is used.
The material does not say what the tools cost or when they will be generally available.
The takeawayIf your business uses an AI assistant, ask your provider how its conversation records are stored and whether they are used for training.