Amazon just started shipping the very layer consulting firms want to sell you.
Bedrock Managed Agents turns model routing and fallback logic into a native cloud feature, not a judgment call you pay a firm to make.
In late April, at its What's Next with AWS event, Amazon put OpenAI's models on its Bedrock platform and, in limited preview, introduced Bedrock Managed Agents. Built with OpenAI's agent harness, the managed agents let a company stand up production-ready AI agents that run inside its own AWS environment, where each agent carries its own identity and logs every action it takes. Anthropic's models and Amazon's own models already run on the same platform, so the buyer is no longer forced to pick a cloud in order to pick a model.
Read past the product names and notice what just happened to the consulting pitch. The work that advisory firms want to sell as ongoing judgment, deciding which model handles a task, what happens when one fails, how to weigh cost against speed, is being folded into the platform as a native feature. It is becoming a setting rather than a standing engagement. When the cloud provider ships the plumbing, the value is no longer in being paid to manage the plumbing.
That reframes the question every buyer should ask before signing a substrate-management contract. As Amazon, and its rivals, keep shipping more of this natively, does your AI strategy get cheaper or more expensive over time? If it gets cheaper, you own the specification the platform executes, and the platform doing more of the work for free is pure upside. If it gets more expensive, you are paying a firm to sit between you and a capability the platform is commoditizing in real time, and you are on the wrong side of the trade.
The timing is the whole point. The advisory firms are about to publish their compute-scarcity briefings, framing the management layer as their domain. The platform is moving the other way, absorbing that layer into the base product. The asset that lasts is not the management service. It is the specification that runs on top of it, and the answer to who owns that specification.