What Is Amazon Kendra? A Guide to Its Use as an Enterprise Search Tool

Amazon Kendra is a managed enterprise search service from Amazon Web Services (AWS) that uses machine learning and natural language processing to help organizations find accurate answers buried across unstructured content, including file shares, SharePoint, Salesforce, S3 buckets, and internal databases. Rather than returning a list of links for users to sift through, Kendra interprets the intent behind a query and surfaces a direct answer, a relevant passage, or the source document itself.

Many organizations still operate Kendra today, and understanding how it works, what it costs, and where AWS is steering the platform next can help you plan a more durable enterprise search strategy. Read on for a full breakdown of Kendra’s features, pricing, and user sentiment, along with how a knowledge layer like Bloomfire fits alongside AWS’s retrieval infrastructure, whether that’s Kendra now or Bedrock Knowledge Bases going forward.

Pros and Cons of Amazon Kendra

Like any enterprise software, Amazon Kendra comes with distinct strengths and limitations. Users generally praise Amazon Kendra for the precision of its search results and its ability to integrate semantic search directly into existing applications. However, many also cite the technical overhead and unpredictable pricing that come with owning an AWS-native service.

Pros

• High-precision search results: Amazon Kendra returns specific answers and ranked passages instead of long link lists, which reviewers cite as a major time-saver over traditional keyword search.
• Fast setup for technical teams: A guided three-step process (create an index, connect data sources, and start querying) along with sample code allows engineering teams to stand up a working proof of concept quickly.
• Wide connector library: Amazon Kendra ships with more than 30 native connectors for common enterprise systems, including SharePoint, Salesforce, ServiceNow, Slack, and databases, reducing the amount of custom integration work needed to get content indexed.
• Deep AWS ecosystem fit: Native integration with services such as Amazon Bedrock, Amazon Q Business, Amazon Simple Storage Service, and Amazon Lex makes Amazon Kendra a natural building block for teams already developing on Amazon Web Services.
• Continuous relevance tuning: Amazon Kendra’s incremental learning feature adjusts rankings based on real user click behavior, improving results over time without manual retraining.

These strengths explain why Amazon Kendra has been a popular choice for teams building custom search experiences on Amazon Web Services, particularly where connector breadth and result precision matter most.

Cons

• Requires dedicated engineering resources: As an application programming interface (API)-first Amazon Web Services service, Amazon Kendra offers no ready-made front end, meaning business teams cannot configure or maintain it without developer involvement.
• Unpredictable, usage-based pricing: Costs accumulate hourly across index tier, storage units, query units, and connector sync time, which reviewers say makes budgeting difficult at scale.
• Capacity and token limits: Certain index tiers cap document counts and extracted text volume, which can restrict larger or more complex deployments unless customers move to higher-capacity editions.
• No content governance or authoring tools: Amazon Kendra indexes and retrieves whatever content already exists in connected systems, but includes no built-in tools for content creation, moderation, or lifecycle management.
• Uncertain long-term roadmap: With Amazon Kendra in maintenance mode and closed to new customers as of July 30, 2026, organizations evaluating it now need to weigh Amazon Web Services’ clear preference for Amazon Bedrock Knowledge Bases going forward.

These limitations matter most for organizations without in-house engineering capacity, and the roadmap change adds a new layer of consideration for anyone starting fresh with enterprise search on Amazon Web Services.

Need Search Without the Overhead?

See how Bloomfire delivers governed, AI-ready enterprise search out of the box.

Learn More!

Key Features of Amazon Kendra

Amazon Kendra’s feature set centers on search precision, extensibility, and integration with the broader AWS artificial intelligence (AI) stack rather than content creation or knowledge governance. The following capabilities form the foundation developers build on when embedding Amazon Kendra into a larger application or workflow.

1. Natural language search: Amazon Kendra uses deep learning models to interpret the intent of a query. It will then return direct, relevant answers rather than generic keyword matches, supporting factoid questions, descriptive questions, and mixed keyword queries.
2. Generative AI integration: Indexes can connect to Amazon Bedrock and Amazon Q Business to power retrieval-augmented generation (RAG) and conversational search experiences, using Amazon Kendra’s GenAI index as the retrieval layer.
3. Experience Builder: A low-code tool that lets teams assemble a functional search interface without writing extensive custom front-end code, making it easier to expose Amazon Kendra-powered search to internal users.
4. Relevance tuning controls: Administrators can adjust ranking based on document freshness, custom attributes, and observed user behavior to improve result quality and promote preferred answers.

Each of these features addresses a specific technical challenge in building enterprise search, from interpreting queries accurately to indexing content at scale. What Amazon Kendra’s feature set does not address is what happens to that retrieved content once an employee finds it, which is where a dedicated knowledge layer becomes valuable.

Amazon Kendra Pricing Plans: What You Need to Know

Understanding Amazon Kendra’s pricing structure is useful for existing customers managing budgets, even as Amazon Web Services steers new projects toward Amazon Bedrock Knowledge Bases. Amazon Kendra bills hourly and separates charges across index tier, storage, query volume, and connectors.

• Basic Developer Edition: Starts around $1.125 per hour, supports up to 10,000 documents or 3 gigabytes (GB) of extracted text, and is intended for proof-of-concept work rather than production use.
• Basic Enterprise Edition: Priced at roughly $1.40 per hour, supports up to 100,000 documents or 30 GB of extracted text, and includes high availability suited for production workloads.
• GenAI Enterprise Edition: Starts around $0.32 per hour, supports up to 20,000 documents or 200 megabytes (MB) of extracted text, and is purpose-built for higher-accuracy semantic and generative use cases.

Additional storage units, query units, and connector sync time are billed separately on top of the base index rate, and a free tier offers up to 750 hours in the first 30 days on select editions, though connector usage falls outside that allowance. Organizations still running Amazon Kendra should map out expected document volume and query traffic carefully, since costs are often underestimated until storage, query, and connector charges are combined at production scale.

Amazon Kendra Reviews: What Users Are Saying

On Gartner and PeerSpot, users consistently highlight Amazon Kendra’s search accuracy as its strongest asset, with reviewers noting it delivers precise, natural language answers instead of long lists of documents to sift through. Many reviewers also praise how easy it is to get started, describing setup as straightforward for teams already comfortable with AWS. Several users also call out the natural language processing quality as a standout feature that meaningfully improves how quickly employees and customers find what they need.

Recurring criticism on both platforms centers on cost and technical overhead. Reviewers frequently cite pricing as the most important drawback, noting that the usage-based model becomes harder to predict as document volume and query traffic scale up. Several also mention token and capacity limits as restrictive for larger deployments, and a common theme across reviews is that getting full value out of Amazon Kendra requires ongoing engineering attention rather than a one-time setup.

Overall, sentiment across G2 and Capterra leans positive on search quality and ease of initial implementation, but consistently cautious on pricing predictability and the technical investment required to scale and maintain the platform, concerns that carry extra weight now that Amazon Kendra has moved into maintenance mode.

Amazon Kendra’s Future and Its Path to Bedrock Knowledge

As organizations weigh Amazon Kendra against newer options, Amazon Bedrock Knowledge Bases is the service AWS now positions as its primary managed retrieval offering for generative artificial intelligence and search use cases. Amazon Bedrock Knowledge Bases is designed to implement the full retrieval-augmented generation workflow, from ingestion and chunking to vector-based retrieval and prompt enrichment. It then connects directly to Amazon Bedrock foundation models.

The migration path, however, is not a simple lift-and-shift. Amazon Bedrock Knowledge Bases supports only seven native data sources compared to Amazon Kendra’s more than 30 connectors, and Amazon Web Services’ own migration guidance confirms that features such as faceted search, query suggestions, custom synonyms, spell checking, incremental learning, and custom document enrichment have no direct equivalents and require custom workarounds. 

For teams currently on Amazon Kendra or evaluating new enterprise search infrastructure, this makes Amazon Bedrock Knowledge Bases the more future-proof foundation, but one that still demands real engineering investment to reach feature parity. This is exactly the kind of gap a governed knowledge layer can help offset.

How Bloomfire Complements Amazon Kendra

Amazon Kendra and Bloomfire are not competing for the same job, and that holds true whether an organization is running Amazon Kendra today or migrating toward Amazon Bedrock Knowledge Bases. Amazon Kendra is a retrieval engine that developers wire into applications to index and surface enterprise content, while Bloomfire is where that content gets curated, verified, and turned into knowledge people actually trust. 

Rather than competing head-to-head, the two can work in tandem. AWS’s retrieval infrastructure handles broad, cross-repository indexing, and Bloomfire provides the governed, human-facing layer that keeps the found content accurate and useful.

• Curated knowledge instead of raw retrieval: Amazon Kendra returns whatever exists in its indexed sources, including outdated or conflicting versions of the truth. Bloomfire’s self-healing knowledge base actively flags duplicate or stale content and routes it for review, so the knowledge feeding into any search experience stays reliable over time.
• A front end built for business users: Amazon Kendra requires custom development to become a usable interface for employees. Bloomfire ships with a ready-made, no-code destination where non-technical teams can browse, ask questions, and contribute knowledge without engineering support.
• Trustworthy, cited AI answers: Amazon Kendra can support retrieval-augmented generation-style answers only once paired with Amazon Bedrock or Amazon Q Business and custom-built citation logic. Bloomfire’s Synapse conversational AI already generates sourced, verifiable answers from certified content, reducing the engineering lift needed to get trustworthy artificial intelligence responses in front of employees.
• Governance for indexed content: Organizations using AWS search infrastructure to pull from Amazon Simple Storage Service, SharePoint, or Salesforce content can route that same content through Bloomfire’s approval workflows and expiration rules first, ensuring what eventually gets surfaced is accurate and current rather than simply “whatever was in the source system”.
• Guided rollout versus a pure engineering lift: Where standing up Amazon Kendra or Amazon Bedrock Knowledge Bases is fundamentally a software development project, Bloomfire’s implementation team guides organizations through structured onboarding, so the knowledge layer sitting alongside AWS’s search infrastructure does not require the same internal engineering investment.

Pairing the two in this way lets AWS-native teams keep the retrieval infrastructure they have already built, or are migrating toward, while giving business users the governed, trustworthy knowledge experience that infrastructure alone is not designed to provide.

Is Amazon Kendra Right for Your Organization?

Amazon Kendra remains a capable option for existing customers already invested in it, particularly engineering-led organizations that want accurate, natural language search embedded in custom applications, chatbots, or customer-facing products. For organizations starting fresh, Amazon Web Services’ own guidance points toward Amazon Bedrock Knowledge Bases as the more durable foundation, even though it currently trades away much of Amazon Kendra’s connector breadth and specialized features. In either case, pairing that retrieval infrastructure with a governed knowledge layer like Bloomfire addresses the gaps that infrastructure alone cannot close: content curation, citation, and a destination employees actually want to use.

Don’t Let Your Search Strategy Lag

Kendra’s shift is the moment to build a knowledge layer that lasts. Let’s talk.

Talk to our Experts