Amazon introduced Q, an AI assistant for AWS customers, at its AWS re:Invent conference in Las Vegas. In public preview and starting at $20 per user per month, Q is designed to help with cloud questions and a wider range of workplace tasks, drawing on AWS expertise and information from connected business systems.
An assistant grounded in company information
Organizations can connect Q to apps and software such as Salesforce, Jira, Zendesk, Gmail and Amazon S3 storage. Q indexes that material to learn business-specific terms, structures and product names, then uses those connections to inform its responses.
Employees can ask questions about a company’s work or upload a Word document, PDF or spreadsheet and ask about its contents. For instance, a team could ask which product features customers find difficult and what improvements might help. Q can respond with citations, giving users a way to trace answers to the information behind them.
The assistant is also intended to help people work with AWS itself. Amazon says Q can account for differences in application workloads, such as how long an app runs or how often it accesses storage. In an example from AWS CEO Adam Selipsky, a user asking which EC2 instance suited a video encoding and transcoding application would receive options that considered both performance and cost.
From answers to work in progress
Q can draft or summarize materials including blog posts, press releases and emails. Through configurable plugins, it can also initiate tasks such as creating service tickets, notifying teams in Slack and updating dashboards in ServiceNow.
Before an action runs, users can inspect what Q intends to do. The assistant also links to the results so people can check whether the action was carried out as expected. This review step puts a human in the loop when Q moves beyond producing text and begins changing work systems.
Q is available through the AWS Management Console and a web app, and can also be accessed in chat apps such as Slack. Amazon describes it as able to troubleshoot network connectivity by analyzing configurations and suggesting remediation steps.
Support for developers and business teams
Q connects with CodeWhisperer, Amazon’s service for generating and interpreting application code. In supported IDEs, including Amazon’s CodeCatalyst, it can generate software tests based on a customer’s code and draft plans and documentation for feature work or code transformations. Developers can refine and execute those plans using natural language.
Amazon said a small internal team used Q to upgrade around 1,000 apps from Java 8 to Java 17 and test them in two days. The code transformation feature supports upgrades from Java 8 and Java 11 to Java 17. Support for moving from .NET Framework to cross-platform .NET was described as coming soon. Q’s code features, including transformation, require a CodeWhisperer Professional subscription.
Amazon is also incorporating Q into AWS Supply Chain and QuickSight. In QuickSight, it can suggest visualizations, reformat business reports and answer questions about the data in a report. In AWS Supply Chain, it can analyze questions about shipment delays. Amazon Connect, its contact center software, can use Q to suggest customer responses and actions, surface related support articles, and prepare post-call summaries for supervisors.
Permissions and data use are central to the pitch
Amazon says Q follows a user’s existing identities, roles and permissions: people can only receive information they are authorized to access. Administrators can also restrict sensitive topics and filter questions or answers where needed.
To reduce the chance that Q invents facts, administrators can configure it to draw only from company documents rather than knowledge from underlying models. Selipsky said the models powering Q, which include models from Bedrock and Amazon’s Titan family, do not train on customer data.
These safeguards address concerns that can make businesses cautious about generative AI, including data exposure and liability. Q’s value will depend on how well its connected information, access controls and proposed actions work in practice. The announcement describes a broad ambition: an assistant that helps people navigate AWS, company knowledge and routine work from a conversational interface.