Amazon has unveiled Q, an innovative AI-powered chatbot for AWS customers, during the keynote at the AWS re:Invent conference in Las Vegas this morning. Starting at just $20 per user per month and currently in public preview, Q can address inquiries such as “How do I build a web application using AWS?” With its training spanning 17 years of AWS knowledge, Q not only provides potential solutions but also offers insights into why users might consider these recommendations.
“You can easily chat with Q, generate content, and execute actions,” stated AWS CEO Adam Selipsky on stage. “It understands your systems, data repositories, and operations, making interactions seamless.”
AWS customers can customize Q by integrating it with organization-specific software and applications such as Salesforce, Jira, Zendesk, Gmail, and Amazon S3. By indexing the connected data and content, Q learns about the business’s organizational structure, core concepts, and product specifics.
For instance, companies can ask Q to analyze product features their customers find challenging and suggest improvements. Similar to ChatGPT, users can upload files (like Word documents, PDFs, or spreadsheets) and pose questions about their contents. Q then utilizes its vast network of connections, integrations, and company-specific data to provide comprehensive answers, complete with citations.
Q extends its functionality beyond simple Q&A. It can generate or summarize content such as blog posts, press releases, and emails, in addition to executing actions on users' behalf via configurable plugins. This includes automatically creating service tickets, notifying teams in Slack, and updating dashboards in ServiceNow. To ensure accuracy, Q prompts users to review actions before they are executed and provides links for validation of results.
Accessible through the AWS Management Console, the web app, and popular chat applications like Slack, Q possesses a deep knowledge of AWS offerings. It understands the nuances of app workloads on AWS and can suggest appropriate AWS solutions based on specific criteria, such as whether apps run for brief intervals or access storage infrequently.
During the keynote, Selipsky provided an example of an application requiring high-performance video encoding. If asked for the most suitable EC2 instance, Q would generate a response factoring in both performance and cost-effectiveness. “I genuinely believe this will be transformative,” Selipsky remarked. “Our goal is to empower various professionals in different fields with the benefits of Amazon Q.”
In addition to its robust capabilities, Q can troubleshoot issues, such as network connectivity challenges, by analyzing configurations and recommending remediation steps.
Q integrates seamlessly with CodeWhisperer, Amazon's application code generation service. Within compatible IDEs (such as Amazon's CodeCatalyst), Q can formulate tests to benchmark software based on customer-specific code. It can also draft plans and documentation for implementing new features or upgrading existing code frameworks, transforming tasks into actionable steps using natural language processing.
Selipsky shared that a small internal team successfully utilized Q to upgrade around 1,000 applications from Java 8 to Java 17—and test them—within just two days. Currently, Q's code transformation features support only the upgrade of Java 8 and Java 11 applications to Java 17, with plans to extend compatibility to .NET Framework in the near future. However, access to Q's development features necessitates a CodeWhisperer Professional subscription, though the timeline for any changes to this requirement remains unclear.
Amazon is also expanding Q's integration with its first-party products like AWS Supply Chain and QuickSight. Within QuickSight, Q can offer visualization options for business reports, automatically reformat them, or respond to inquiries about data referenced in reports. In AWS Supply Chain, users can query, “What’s causing the delay in my shipments?” and receive up-to-minute analyses in return.
Q will soon be incorporated into Amazon Connect, the company's contact center software. With Q's assistance, customer service agents will be able to receive suggested responses to customer inquiries, including proposed actions and relevant links to support articles, without needing to type queries manually. Additionally, Q will generate post-call summaries for supervisors to track follow-up actions.
Selipsky emphasized throughout the keynote that every answer Q generates and action it takes is fully controllable and filterable. Q ensures that users only access information they are authorized to see, while administrators can impose restrictions on sensitive topics, filtering out inappropriate queries and responses as necessary.
To combat the common issue of "hallucinations"—instances where Q may produce inaccurate information—administrators can configure it to source responses solely from company documents rather than generalized knowledge models. This safeguards proprietary data, as the models that drive Q—comprised of various technologies from Amazon’s AI development platform, Bedrock, including the proprietary Titan family—do not utilize customer data for training.
These assurances are particularly relevant for enterprises that approach generative AI cautiously due to liability and security concerns. Several organizations have imposed bans or restrictions on tools like ChatGPT, highlighting fears regarding data usage and potential leaks.
“If a user lacks permission to access information without Q, they won't gain access through Q, either,” Selipsky asserted. "Q recognizes and respects your existing identities, roles, and permissions. We will never use sensitive business content to train the underlying models."
This careful emphasis on privacy positions Q as Amazon’s response to similar offerings from Microsoft and Google, such as Copilot for Azure and Duet AI for Google Cloud. Both services function as chat-driven assistants that assist cloud customers in configuring apps and troubleshooting potential issues.
While it bears similarities to these competitors, Q appears to be more comprehensive, addressing a broader spectrum of business intelligence, programming, and configuration needs. Ray Wang, founder and principal analyst at Constellation Research, he views it as the “most important” announcement at re:Invent thus far, underscoring its significance in supporting developers with AI tools.
Ultimately, the true test lies in Q’s performance and whether it lives up to Amazon's ambitious promises.
Ranking
