Amazon Q

Also known as: Amazon Q Developer · Amazon Q Business

On this page 7

Amazon Q is the family of generative AI assistants from Amazon Web Services that answers questions, writes and transforms code, and completes tasks using an organization’s own data and cloud resources. Within AI models & assistants, it is best understood not as a single foundation model but as an assistant product layer built on Amazon Bedrock, which routes requests to underlying models and wraps them with retrieval, permissions and enterprise controls.

The two main branches are Amazon Q Developer, aimed at software engineers and cloud operators, and Amazon Q Business, aimed at employees who need answers from internal documents and applications. AWS also embeds Q capabilities into services such as QuickSight and Amazon Connect, so the name describes a shared assistant brand rather than one interface.

How Amazon Q processes a request

Amazon Q answers a request by combining a large language model with retrieved context and the caller’s identity. The user prompt is enriched with relevant material, such as indexed documents, open source files or AWS account metadata, and the model generates a response that is filtered by guardrails and access rules before it is shown.

This design follows the retrieval-augmented generation pattern. Instead of retraining a model on customer data, Amazon Q indexes that data separately and fetches the passages relevant to each question. AWS states that customer content in the business tiers is not used to train the underlying models, which is a central reason enterprises consider the product.

The main processing stages are summarized below, together with the component responsible for each one.

StageWhat happensResponsible component
AuthenticationThe user identity and group membership are resolvedIAM Identity Center or IAM
RetrievalRelevant documents, code or resource data are fetchedIndex, connectors or workspace context
GenerationA foundation model drafts the answer or codeModels served through Amazon Bedrock
FilteringTopic blocks, permissions and safety rules are appliedGuardrails and admin controls
ActionOptional plugins or agents perform a taskPlugins, agents and tool integrations
Main stages of an Amazon Q request

Amazon Q Developer and Amazon Q Business compared

The two branches share infrastructure but serve different users and data sources. Amazon Q Developer grew out of Amazon CodeWhisperer, the earlier AWS code completion tool, while Amazon Q Business is an enterprise assistant that connects to repositories such as SharePoint, Confluence, Salesforce, Amazon S3 and many other systems through managed data source connectors.

The following table compares the branches on the properties that most often decide which one an organization adopts.

PropertyAmazon Q DeveloperAmazon Q Business
Primary userDevelopers, DevOps and cloud administratorsKnowledge workers across departments
Main surfacesIDE plugins, command line, AWS consoleWeb app, browser extension, chat integrations
Typical dataSource code, AWS account resources, documentationWikis, tickets, file shares, CRM records
Core tasksCode completion, debugging, code transformationQuestion answering, summaries, simple workflows
Access controlIAM or Builder ID identitiesDocument-level permissions from connected sources
Comparison of the two main Amazon Q branches

A distinctive feature of Amazon Q Developer is its set of software development agents. These agents can implement a multi-file feature from a natural-language description, generate unit tests, review code for security issues and perform large code transformations, such as upgrading Java applications from older releases to a newer long-term support version.

Where Amazon Q sits among AI assistants

Amazon Q competes most directly with Microsoft 365 Copilot for workplace knowledge and with GitHub Copilot for coding assistance. Its differentiator is deep integration with AWS services and accounts, which matters most to organizations that already run their infrastructure on AWS.

The table below positions Amazon Q against neighboring assistants in the same topic, focusing on ecosystem and grounding source rather than on model quality, which changes quickly between releases.

AssistantHome ecosystemMain grounding sourceTypical buyer
Amazon QAWSCompany data and AWS resourcesAWS-centric enterprises
Microsoft 365 CopilotMicrosoft 365Microsoft Graph contentMicrosoft 365 tenants
GitHub CopilotGitHub and IDEsRepository and editor contextDevelopment teams
NotebookLMGoogleUser-uploaded sourcesIndividuals and small teams
Amazon Q and related assistant products

Unlike model families such as Claude, Gemini or Llama, Amazon Q is not a model that customers download or call directly. It is an application that hides model selection from the user, so its answer quality depends on both the underlying models and the quality of the retrieval index.

Deploying and configuring Amazon Q

A successful deployment of Amazon Q depends more on data preparation and identity setup than on prompt design. Administrators create an application, connect data sources, map users and groups, and define which topics or actions are blocked.

  • Identity integration: connect IAM Identity Center to the corporate directory so that answers respect existing group membership.
  • Connector scope: index only repositories with clean, current content, because outdated pages produce outdated answers.
  • Admin guardrails: configure blocked words, blocked topics and whether the assistant may answer from general model knowledge.
  • Subscription tiers: assign per-user plans, since lower tiers restrict features such as plugins and agentic actions.

The following decision table helps match common situations to a sensible starting configuration.

SituationRecommended choiceReason
Team writes code for AWS workloadsAmazon Q Developer in the IDECombines code help with AWS resource context
Staff search scattered internal documentsAmazon Q Business with connectorsRetrieves answers while honoring permissions
Legacy Java or .NET estate needs modernizingDeveloper transformation agentsAutomates repetitive upgrade steps
Analysts need dashboards from questionsQ features in QuickSightGenerates visuals from natural language
Choosing an Amazon Q setup

Limitations and common mistakes with Amazon Q

Amazon Q shares the core weaknesses of generative AI: it can produce confident but wrong answers, and retrieval only reduces this risk without removing it. Its permission model is only as accurate as the access control lists in the source systems, so overshared files remain overshared in the assistant.

SymptomLikely causeFix
Answers cite outdated policiesStale documents remain in the indexArchive old content and schedule regular syncs
Users see content they should notOverly broad permissions in the sourceAudit source access lists before indexing
Generated code fails reviewSuggestions accepted without testingRequire tests and normal code review
Confusion about product namesAWS has reorganized the brandingCheck current documentation for each feature
Frequent problems and their fixes

A further practical point is naming drift. AWS has repeatedly renamed and regrouped parts of the Q family, moving some capabilities into newer products, so tutorials written a year earlier may describe menus and tiers that no longer exist. Teams should treat current service documentation as authoritative.

Stripped to the essentials

Amazon Q is a helpful assistant built by AWS that people can chat with at work. Programmers use one version to get help writing and fixing code. Other employees use a second version to ask questions about company documents, such as travel rules or product manuals.

Think of it as a well-read colleague who only opens the files you may open. It looks up the relevant pages first and then writes an answer in plain language. It can still be wrong, so important answers deserve a quick check.

Where it is applied

A mid-sized insurance company runs most of its systems on AWS and keeps claims procedures in Confluence and contracts in SharePoint. The IT team deploys Amazon Q Business with both connectors and links it to the corporate directory, so claims handlers see only the documents their department may read.

In parallel, the development team of about 40 engineers enables Amazon Q Developer in its IDEs and uses the transformation agent to upgrade a dozen internal Java services to a newer runtime. Engineers still review every change and run the existing test suites.

After three months, claims handlers report finding procedures in seconds rather than searching several wikis, and the upgrade project finishes weeks earlier than planned. The team also discovers several overshared folders during the permission audit, which turns out to be a valuable side effect of the rollout.