Amazon AIP-C01: Amazon Q, Quick and Kiro in 2026
Amazon AIP-C01: Amazon Q, Quick and Kiro in 2026
Amazon Q became a familiar name for two different kinds of assistance: finding answers across company information and helping developers work with AWS and software code. By 2026, treating “Amazon Q” as one unchanged product creates confusion. AWS has redirected new enterprise-assistant deployments toward Amazon Quick, and the Amazon Q Developer IDE experience is moving toward Kiro. Some existing Q services continue, while particular new-customer paths and IDE subscriptions have changed.
For a business planning to adopt or renew an AI assistant, those distinctions are practical rather than cosmetic. A knowledge-search tool needs source permissions and citation quality. A coding assistant needs repository context, developer workflows and a tested migration plan. A platform for building a custom agent needs application-controlled tools and backend authorization. Choosing among them begins with the user’s job and the product’s current availability, not with a shared brand name.
First separate the products and their users
Amazon Q Business was designed for searching, summarizing and acting on information in connected enterprise systems. Amazon Q Developer addressed software engineering and AWS development tasks. Amazon Q features also appeared in particular AWS service experiences, such as the management console or analytics workflows. Those names do not imply that access, licenses or migration behavior is identical across products.
The significant changes can be understood as separate tracks:
| Experience | Main user need | 2026 planning position |
|---|---|---|
| Amazon Q Business | Enterprise knowledge and connected work | Closed to new customers from July 31, 2026; existing customers may continue |
| Amazon Quick | Research, enterprise insights and actions | AWS’s newer direction for capabilities related to Q Business |
| Amazon Q Developer IDE plugins | Coding assistance inside supported IDEs | New signup restrictions began May 15, 2026; support is scheduled to end April 30, 2027 |
| Kiro | Agentic, specification-driven software development | Recommended migration direction for the affected IDE workflows |
| Q Developer in the AWS Console and certain first-party experiences | AWS guidance and operational assistance | Not included in the announced IDE-plugin sunset |
Amazon Quick was introduced as Amazon Quick Suite before AWS began using the shorter product name in some current materials. For an existing Q Business account, AWS describes options to keep using the existing service or use its index with Quick. That is not a promise that every custom integration, entitlement, workflow and administrative setting will transfer unchanged. A migration still requires testing against the actual business use case.
Enterprise knowledge assistants depend on the quality of their sources
The basic enterprise-assistant task sounds simple: an employee asks a question and receives an answer based on company documents. The hard part is deciding which documents should count. A customer-support worker may need a current returns policy, an approved troubleshooting procedure and the history of the specific customer conversation. A finance analyst may need the latest close checklist, but not another team’s payroll file.
Connectors and indexing can make information searchable, but searchability is not sufficient. Each record needs an identity, source location, effective date where relevant and a permission model that can be reconciled with the employee’s actual access. If a group membership is revoked, the assistant must not continue serving restricted passages from an old index or cached interaction. If a policy has been replaced, the previous version should not silently outrank the new one just because its language matches the query more closely.
Useful answers should preserve evidence. For policy questions, a citation to the precise governing document and version is more valuable than a polished explanation with no traceable source. Employees should be able to see where an answer came from and recognize when material is missing or contradictory. A capable assistant should sometimes refuse to make a decision and instead present the unresolved records or escalate to a responsible person.
The underlying techniques overlap with retrieval-augmented generation on AWS, but a prebuilt enterprise assistant and a custom retrieval application are not interchangeable products. The latter gives developers more direct control over retrieval logic and workflows while also making them responsible for more security, testing and operational decisions.
Amazon Q Business and Amazon Quick require a migration decision
AWS’s 2026 availability change means Amazon Q Business is no longer open to new customers. Existing customers can continue using their current deployment, while Amazon Quick is presented as the newer route for broader research, analysis and automation capabilities. Organizations with Q Business already deployed should first document what they actually use: connected repositories, user groups, permissions, application actions, index contents, citation behavior, costs and retention settings.
Before moving an established workflow, test whether a user can get the same correct, permitted result on the proposed destination. A successful login and a working search box are weak migration evidence. Verify how the replacement handles an expired document, a removed employee, an unavailable connector and a request that needs an action rather than an answer. Measure the operational impact of retraining employees or changing automation owners as well as the technical migration.
New adopters should avoid designing a fresh Q Business deployment based on older promotional articles that describe it as open to all customers. They should start with AWS’s current availability guidance and evaluate Quick against the actual security, integration and licensing requirements. Existing customers, meanwhile, should avoid assuming an immediate shutdown where AWS has announced a restriction on new customers rather than termination of their current service.
Amazon Q Developer IDE users have a specific deadline
AWS announced that new Amazon Q Developer IDE-related accounts and new paid subscriptions would stop being available from May 15, 2026. Existing eligible subscriptions can continue adding users, and customers with the relevant Q Developer Pro or Kiro subscription can retain IDE-plugin access through the transition window. Support for the affected IDE plugins and paid subscriptions is scheduled to end on April 30, 2027. The announcement also includes IDE-based code transformation capabilities delivered through those plugins.
Those are not universal dates for every experience containing the words “Amazon Q Developer.” AWS specifically states that its developer assistance in the AWS Management Console and named first-party experiences, including supported chat-app integrations, is not part of the IDE-plugin sunset. A team should inspect exactly where each developer currently uses the product before deciding that the entire Q Developer service has been discontinued.
Kiro is AWS’s recommended direction for developers who relied on the IDE plugins. It provides an agentic environment oriented toward specifications, structured implementation, repository context and repeatable development workflows. A migration should inventory the features developers actually depend on: inline suggestions, chat, project-level instructions, extensions, authentication, generated code review and code transformation. Switching tool names without testing the development process may disrupt a team’s normal work.
For a coding organization, useful migration tests include generating a change against an existing codebase, preserving established test conventions, explaining a security-sensitive modification, and handling an incomplete or failing build. Human developers still need to review commits, protect secrets, check dependency licenses and run project tests. Agentic coding does not remove the normal responsibility to verify software changes.
Governance is more than avoiding disallowed answers
A knowledge assistant can create two separate risks. The first is retrieving information a user was never entitled to see. The second is responding poorly to information the user was entitled to receive. These require different controls. Identity-aware indexing and retrieval protect access boundaries; content and quality evaluations address unsafe language, unsupported claims and inappropriate recommendations. A content filter cannot reverse an unauthorized search result.
Where a custom Bedrock application is appropriate, Guardrails policies can complement application authorization through selected input and output checks. A ready-made enterprise product has its own supported governance features, which must be evaluated on their actual terms rather than assumed to be identical to controls in a separately developed Bedrock application. Administrators should also evaluate how prompts, source passages and action logs are stored and who can inspect them.
Action-taking workflows raise the stakes. An assistant may be allowed to summarize a customer refund policy but not issue a refund on its own. If an authorized service does permit an action, it needs to validate the employee’s role, the customer record, allowed monetary limit, approval requirement and current transaction state. The language model’s recommendation is evidence for a decision, not the authorization or the proof that the action succeeded.
Measure productivity by resolved work, not impressive demonstrations
Enterprise AI adoption is often justified by examples of faster answers and smoother collaboration. Those benefits may be real, but a demo with tidy documents is not a measurement of production value. Pick recurring tasks with a known baseline: locating an approved procedure, preparing a case summary, answering a routine internal question or generating a first pass at a software change. Measure correct task completion, time spent checking the result, escalations and the frequency of inaccurate or unauthorized responses.
A support assistant that saves a minute on easy questions but sends employees down the wrong policy path on difficult ones may have little net value. A developer assistant that accelerates code generation but increases review and rework time may disappoint. Cost calculations should include subscriptions, index capacity, integrations, security reviews and human corrections where appropriate; a per-user price alone is not the full operating cost.
Build an evaluation set that includes both success cases and expected refusals. Test revoked access, document updates, conflicting records, tool timeouts and misunderstandings of user intent. For a bespoke generative workflow, foundation-model evaluation should cover retrieval, generation, authorization and any downstream action separately. Aggregate satisfaction cannot reveal whether one high-impact business step repeatedly fails.
Know when to use a product and when to build a system
Ready-made enterprise tools can reduce the work required to connect knowledge and deliver a usable interface. They make sense when the available connectors, governance options and supported workflows match the organization’s needs. A custom solution may be appropriate when a team needs specialized retrieval, a unique approval flow, complex tool orchestration, model routing or detailed control over each transaction. That flexibility carries responsibility for infrastructure, incident response and ongoing evaluation.
For example, an organization might use Amazon Quick for cross-source workplace research but build a separate customer-facing assistant on Bedrock because it needs precise transaction contracts and product-specific entitlements. The Bedrock production architecture then involves retrieval, scoped tools, retries, tracing and business-state validation that are distinct from an employee’s research interface. Choosing two different solutions can be more coherent than forcing a single brand into every use case.
The important 2026 lesson is that Amazon Q’s original umbrella name no longer describes one static procurement decision. Amazon Q Business has a newer enterprise path through Amazon Quick, Amazon Q Developer IDE users have a dated migration route through Kiro, and unaffected AWS assistance experiences should not be mislabeled as discontinued. A sound decision keeps those product boundaries clear, tests current entitlement and capability details, and evaluates whether employees actually complete their work more correctly and efficiently.
- Category: All Certifications Amazon
- Tags: AI, Amazon Q, Enterprise, Generative
