Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
AWS Adds Chat-Built Flows and Multi-Agent Orchestration to Quick - OpenSmartRoute
AWS Adds Chat-Built Flows and Multi-Agent Orchestration to Quick
AWS released updates enabling chat-described workflows and multi-agent orchestration for better process automation and collaboration.
Key points
New features support flows that coordinate multiple specialized agents simultaneously.
Flows can analyze data, perform calculations, draft reports, and be tested in chat.
Group permissions and write approvals control what agents can do across teams.
Features are available immediately in all regions where Quick is offered.
Why it matters: This update enables faster, more reliable automation of recurring tasks, reducing dependence on individual employees and increasing process consistency.
By OpenSmartRoute editorial · written through the router by llm-small
From Unite.AI - “AWS Quick Adds Chat-Built Flows and Multi-Agent Orchestration”
Introduction - AWS releases a broad update to Quick with chat-built workflows and multi-agent orchestration
On November 20, 2026, AWS announced a major update to its Quick platform. This update introduces new features that allow users to create workflows through chat descriptions. It also adds the ability to run multiple specialized agents at the same time. These changes aim to improve process automation and team collaboration.
This update makes Quick more flexible and easier to use for teams. It supports building, testing, and refining workflows directly in chat. It also enables organizations to coordinate different AI agents for complex tasks. These features are available immediately in all regions where Quick is offered. They help organizations standardize recurring work and manage multiple AI tools more effectively.
Conversational Flows - Describing processes in chat to create, test, and refine workflows
The new chat-based workflow feature allows users to describe a process in natural language. They can then have Quick automatically build a flow based on their description. This flow can be run, tested, and improved within the same chat session. Users see a step diagram alongside the chat, which helps review and understand the process.
These conversational flows can include various elements. They can coordinate multiple agents, skills, tools, structured data, knowledge, and applications. Users can start flows from web, desktop, or mobile interfaces. The platform supports longer workflows and larger data volumes. This makes it easier for teams to automate complex, repetitive tasks without needing detailed technical knowledge.
The process of building flows in chat is designed to be straightforward. For example, a user might describe a monthly report process. Quick then creates a flow that analyzes data, compares results, flags variances, drafts commentary, and prepares a report. The user can review each step, make adjustments, and approve the final output. This approach reduces reliance on individual employees’ private checklists and promotes consistency across teams.
Flow features and capabilities - Handling calculations, data analysis, drafting, and longer processes
Wotton Kearney deploys its AI platform Harvey across 13 offices after a successful pilot with 94% adoption, aiming for measurable client benefits.
The updated platform introduces deterministic steps alongside AI-led drafting and analysis. Deterministic steps perform fixed calculations and checks, such as applying spreadsheet formulas or verifying data integrity. These steps help ensure workflows are repeatable and verifiable.
AWS emphasizes that these features support longer-running processes and larger data sets. For example, a monthly reporting flow can analyze multiple regions in parallel. It can compare results with predefined plans, apply thresholds to identify significant variances, and generate summaries automatically. These processes can be run repeatedly with minimal manual intervention, saving time and reducing errors.
The platform also supports drafting and drafting review. For instance, after analyzing data, an agent can draft commentary or summaries. Users can then review, request modifications, or approve the output directly in chat. This streamlines collaboration and speeds up decision-making.
The new features aim to make workflows more reliable and consistent. They enable teams to automate routine tasks while maintaining control over the process. Longer workflows and larger data volumes become manageable, helping organizations scale their automation efforts.
Real-world examples - Monthly reporting, data work, and collaboration benefits
AWS provides a fictional example of a finance team using these new features. The process involves analyzing regional data, comparing it with plans, and flagging variances. Fixed spreadsheet formulas are used for calculations, and thresholds are set to identify significant differences. The workflow then drafts commentary and prepares an Excel report for review.
A person examines the flagged regions and requests changes in chat. Once satisfied, they approve the report for distribution. Colleagues can then run the same flow again, ensuring consistency. This reduces dependence on individual employees' private checklists and promotes standardization.
Another real-world example comes from a Latin American fintech provider. The company’s head of data reports that nearly four times as many employees now work with data each month. They have achieved this without expanding their data team. One specific use case saves 80 hours of work per month. Employees across the business can build agents and flows under shared governance, making data work more accessible.
This broader participation in data tasks is a key benefit of the update. It allows organizations to scale their automation efforts without hiring more specialists. Teams can collaborate more effectively, and workflows become more transparent and repeatable.
The platform’s new interface combines a prioritized feed of emails and messages with a daily briefing. Users can ask questions or have Quick draft replies. Longer tasks continue even after the browser closes. When users return, they see completed results and pending actions from agents. The library feature collects recent work, favorites, pinned items, and Quick-created files. Conversations, agents, and files sync across devices, supporting seamless collaboration.
Multi-agent orchestration - Running several specialized agents at once with independent instructions
The update introduces multi-agent orchestration, which allows several specialized agents to work together on a single request. Each agent retains its own instructions, knowledge, and context. This setup enables domain experts to maintain their own agents independently.
When a request is made, Quick calls multiple agents simultaneously. Each agent processes the request based on its own instructions. The platform then combines the responses and presents them to the user. This approach allows different parts of a task to be handled by the most suitable agents.
For example, in a shipping-cost inquiry, one agent might forecast costs using planning data. Another agent could handle vendor contracts and procurement details. Both agents work on the same question but address different aspects. The user then reviews the combined response and decides what to do next.
This multi-agent setup improves efficiency and accuracy. It allows organizations to pool expertise and maintain control over each agent’s instructions. Teams can update individual agents without affecting the entire system. The result is a more flexible and scalable automation environment.
New connectors and integrations - Meta ads, Amplitude, and other data sources
AWS has added new connectors to Quick, expanding its integration options. The Meta ads connector uses Meta’s Model Context Protocol server. It enables natural-language campaign creation, management, and analysis. Users can describe their campaigns in chat, and Quick handles the rest.
Amplitude’s MCP (Measurement, Collection, and Processing) server is now supported. It allows questions about product usage, customer behavior, and conversion rates. These connectors help teams analyze marketing and product data more easily.
Existing access permissions still apply to these connectors. Administrators decide which teams can add or use each integration. Write actions, such as creating a Meta campaign or changing a budget, require explicit permission. This ensures control and security across the organization.
Reusable skills, which were previously managed on desktop, are now accessible in the web app. Teams can browse, install, create, and manage instructions for common tasks. These skills include checking contracts, formatting updates, or other routine activities.
AWS also previews prompt guardrails. Administrators can filter content, block certain topics or words, and detect prompt attacks. They can test policies before enforcing them. Guardrails help prevent inappropriate or unsafe outputs and protect personal information.
These new connectors and controls give organizations more options to extend automation and maintain governance. They support a broader range of data sources and use cases while keeping security tight.
Administrative controls - Permissions, guardrails, and shared governance for teams
The update enhances administrative controls for managing teams and workflows. Group permissions and usage limits help define what agents and flows team members can use. Administrators can set policies to control access and prevent misuse.
For example, an admin might restrict certain agents to specific teams or limit the number of hours an agent can run. They can also block data sources like file uploads or restrict connections to certain databases. These controls help enforce organizational policies and compliance.
Prompt guardrails are in preview, offering additional safety features. Admins can filter content, block specific words or topics, and detect prompt attacks. They can test these policies in a detect-only mode before applying them broadly. This reduces the risk of unsafe outputs or data leaks.
Shared governance allows teams to collaborate on building and managing agents and flows. Permissions and roles can be assigned to ensure that only authorized users make changes. This setup helps maintain control while enabling broader participation in automation efforts.
AWS’s approach emphasizes security and compliance. It provides tools to define what each team or user can do, ensuring that automation remains aligned with organizational standards.
Availability and data security - Features available immediately, data remains in customer environment
All these new features are available immediately in every region where Quick is offered. AWS states that customer data stays within the customer’s environment. The platform does not use the data to train other models or for any external purpose.
This focus on data security means organizations can adopt the new features without risking data leaks or privacy issues. The platform supports compliance with organizational policies and regulations. Customers retain control over their data and how it is used.
The features include local data handling, permission controls, and audit logs. These measures help organizations meet security and privacy requirements while automating workflows and managing AI agents.
AWS’s commitment to data security ensures that organizations can leverage these advanced automation tools confidently. They can build and run workflows knowing their data remains protected and under their control.
How OpenSmartRoute helps - Routing requests through the platform for better control and security
OpenSmartRoute is an open-source router designed for AI requests. It directs each request to the most suitable model, agent, or tool from a predefined catalog. The platform scores options based on quality, cost, speed, and safety. Teams set the importance of each factor per request.
Using OpenSmartRoute, organizations can enforce strict rules. For example, personal data can be pinned to on-premises models. Requests can be limited by region or cost. The system learns from outcomes, giving more traffic to well-performing models and less to those that fail.
When a new model is added, it becomes part of the catalog and competes for traffic in future requests. The platform works with any OpenAI-compatible provider, open-weight models served locally, and various tools and agents. It also provides a savings ledger that shows the cost of each routed request compared to the most expensive option.
OpenSmartRoute enhances control and security for teams automating AI workflows. It ensures requests follow organizational rules and optimizes for quality and cost. The platform also detects prompt injection and personal data before requests leave, maintaining privacy and safety.
By integrating with the new AWS features, OpenSmartRoute can route requests to the most appropriate agents or models based on the workflow. It supports complex orchestration while keeping data and operations secure.