Exam-style
Three requests land against the same Bedrock-backed summarisation service in one week. The front-end team wants a hosted web app with sign-in and a chat component wired to the backend, without building the hosting, the auth flow or the API client by hand. The partner integration team wants the request and response contract agreed and mocked so both sides can build against it before either writes code. An operations analyst wants to reorder the steps of a summarise-then-classify pipeline, and occasionally add a condition, without waiting on a release. Which set of tools fits the three asks in that order?
Reveal the answer
C. AWS Amplify for the front end, an OpenAPI specification for the partner contract, and Amazon Bedrock Prompt Flows for the analyst
Each ask names its own tool if you read what the asker is trying to avoid. The front-end team is avoiding hosting, sign-in and hand-written API clients, which is AWS Amplify to develop declarative UI components: hosting, an auth flow and a chat component that binds to the backend without plumbing written by hand. The partner team is avoiding the fixed cost of two teams guessing at each other, which is what OpenAPI specifications for API-first development approaches buy: one document both sides generate clients, mocks and validation from, agreed before either side commits code. The analyst is avoiding the release train, which is Amazon Bedrock Prompt Flows for no-code workflow builders: a visual canvas where prompt, condition and Lambda nodes are wired together, versioned and published without a build. The tempting swap is a Step Functions state machine for that last one, because the pipeline really is a sequence of steps with a branch in it, and Step Functions runs sequences of steps with branches. What the analyst asked for was a change they can make themselves, and a state machine is application code: the definition lives in the repository, moving a step is a pull request and a deployment, and the analyst is back where they started. Durability, retries and long-running fan-out would earn a state machine here; nobody asked for those.
Q. A hosted chat UI with sign-in, an API contract agreed before either side codes, and a pipeline an analyst can reorder without a deployment. Three asks, one week. Which tool for which?
A. AWS Amplify for the hosted interface, an OpenAPI specification for the contract, and Amazon Bedrock Prompt Flows for the reorderable pipeline.
Why? These are the three routes to accessible AI interfaces, and they solve different bottlenecks. Amplify removes the work between a working model call and a usable app: hosting, an auth flow, and declarative UI components including a chat interface that talks to the backend without a hand-rolled client. An OpenAPI document removes the wait between two teams, since a contract that generates mocks and clients lets both sides build in parallel and settle the response shape, including whether it streams, before anyone writes the handler. Prompt Flows removes the deployment from the loop for someone who is not shipping code: nodes on a canvas, a version published when the change is ready, no build. Reach for a Step Functions state machine when the pipeline needs durable execution, retries or a run measured in hours. Reach for a visual builder when the person changing the pipeline is not the person with commit access.