AWS Quick Adds Chat-Built Flows and Agents

Amazon Web Services has released a broad update to AWS Quick, its artificial intelligence workspace for work, adding business workflows that can be built entirely in conversation, orchestration across several specialised agents at once, and a redesigned web application. AWS announced the changes on 20 November, saying the features are available immediately in every region where Quick is offered.
Quick is the company's workspace for turning everyday business work into repeatable processes. The central idea of this release is that the way one person does a task should become a process the whole team can rely on, rather than staying in that person's head or on a private checklist.
Workflows built by describing them in chat
Users can now describe a process in chat and have Quick build a flow to run it, then test and refine that flow in the same conversation. A step diagram beside the chat shows each stage, so the person who described the process can check that it matches how they actually work. Flows can coordinate agents, skills, tools, structured data, knowledge and applications, and they can be started from the web, desktop or mobile app.
AWS illustrates the idea with a fictional finance team. A manager prepares a monthly report explaining changes to costs and margins, pulling last month's results, comparing them with the annual plan in a spreadsheet, flagging regions that moved significantly and drafting commentary. With Quick, she describes that routine in chat, the flow is built and tested with her, and colleagues can then run the same flow at the next month-end, including a new starter at their first month-end, so the report is built the same way each time.
New deterministic steps handle fixed calculations and checks alongside agent-led drafting and analysis, which AWS says makes repeated workflows consistent and verifiable, with support for longer-running processes and larger data volumes. A real customer example comes from Pomelo, a Latin American financial technology provider, whose head of data, Rodrigo Marini, said nearly four times as many people now work with data each month without the data team growing, and that one use case already saves 80 hours of work per month.
Several agents answering one question
The update also adds multi-agent orchestration. With it, the agent receiving a request can call several specialised agents at the same time, and Quick combines what they find into one response. Each agent keeps its own instructions, knowledge and context, so the people who know that work best can build and maintain it.
AWS gives the example of asking why shipping costs rose in one region. A forecasting agent built on the team's planning data and a procurement agent built on the procurement team's vendor contracts address different parts of the same question, and the combined response shows what each agent contributed, leaving the user to decide what to do next.
The updated web app puts what needs attention on the home screen as soon as a user signs in, with a daily briefing and a prioritised feed of email and messages. Users can ask about an item or have Quick draft a reply for review. Scheduled and longer tasks keep running after the browser is closed, and a Mission Control view surfaces completed results and agents waiting for input when users return. Conversations, agents and files synchronise across web, desktop and mobile.
New connectors and tighter controls
Quick has added connectors for Meta ads and Amplitude. The Meta ads connector uses Meta's ads Model Context Protocol server for natural-language campaign creation, management and analysis, and Quick asks for permission before it creates a campaign, changes a budget or takes any other write action in an ad account. The Amplitude connector supports questions about product usage, customer behaviour and conversion. Across connectors, existing access permissions still apply.
Reusable skills, previously managed on desktop, can now be browsed, installed, created and managed in the web app. Prompt guardrails are in preview, letting administrators filter content, block denied topics and words, detect prompt attacks and protect selected types of personal information, and test a policy in detect-only mode before enforcing it. Administrators can also assign group-based permission and usage-limit profiles, for example capping agent hours by group, and control which data-source types people can create, update or share. Those controls are available to Professional and Enterprise subscribers.
Quick has also added Japanese, German, Simplified and Traditional Chinese and Vietnamese. AWS says customer data remains in the customer's environment and is not used to train anyone else's models. The company is offering Quick free for three months, with up to 10 teammates invited at no cost.