DP-800 Exam: What the SQL AI Developer Certification Covers

DP-800, Developing AI-Enabled Database Solutions, is the exam for the Microsoft Certified: SQL AI Developer Associate. It tests T-SQL database development across SQL Server, Azure SQL, and SQL database in Microsoft Fabric, plus security, performance, and CI/CD, and a block of AI skills: embeddings, vector and hybrid search, and retrieval-augmented generation driven from the database itself.

At a glance

• Exam: DP-800, Developing AI-Enabled Database Solutions.
• Certification: Microsoft Certified: SQL AI Developer Associate.
• Skills measured as of October 19, 2026: design and develop database solutions (35 to 40%), secure, optimize, and deploy database solutions (35 to 40%), and implement AI capabilities in database solutions (25 to 30%).
• Platforms: SQL Server, Azure SQL, and SQL databases in Microsoft Fabric.
• Passing score: 700 or greater. Renewal is yearly through a free online assessment on Microsoft Learn.
• Free practice: Microsoft lists a DP-800 practice assessment.
• Price: check the exam page for the current price in the United States.

Who DP-800 suits

Microsoft's DP-800 study guide describes a developer who writes T-SQL and builds databases on Microsoft SQL platforms, knows CI/CD in GitHub, uses AI-assisted development tools, and understands embeddings, vectors, and models. You work with application developers, DBAs, architects, AI engineers, and security teams.

In practice, it fits:

• SQL developers who are being asked to add semantic search or a chat feature over company data.
• Application developers who own their database layer and want AI features without a separate search service.
• BI and reporting developers with strong T-SQL who want to move toward application work.

It is less of a fit if your job is mainly administration (DP-300 targets that), data engineering in Fabric (DP-700), or building agents and AI apps on Azure (AI-103).

Microsoft added SQL AI Developer Associate as a skilling option for the Data and AI Solutions Partner designation in its June 2026 Partner Center announcements, which matters if you work at a Microsoft partner.

The three skill areas in plain terms

Design and develop database solutions (35 to 40%)

The foundation is solid database design: tables, data types, columnstore indexes, constraints, sequences, and partitioning, plus specialized tables (in-memory, temporal, external, ledger, and graph) and JSON columns with JSON indexes. Then programmability: views, scalar and table-valued functions, stored procedures, and triggers.

The T-SQL list is broader than many developers expect. Beyond CTEs and window functions, it names JSON functions such as JSON_OBJECT, JSON_ARRAYAGG, and JSON_CONTAINS, regular expression functions such as REGEXP_LIKE and REGEXP_REPLACE, fuzzy matching with EDIT_DISTANCE and JARO_WINKLER_DISTANCE, graph queries with MATCH, and error handling.

This area also covers AI-assisted development: enabling GitHub Copilot and Copilot in Fabric, choosing models and Model Context Protocol (MCP) tools in a chat session, writing GitHub Copilot instruction files, connecting to MCP server endpoints for SQL Server and Fabric lakehouses, and understanding the security impact of all of it.

Secure, optimize, and deploy database solutions (35 to 40%)

• Security: Always Encrypted and column-level encryption, Dynamic Data Masking, Row-Level Security, object permissions, passwordless access, auditing, securing model endpoints with managed identity, and securing GraphQL, REST, and MCP endpoints.
• Performance: isolation levels and concurrency, execution plans, DMVs, Query Store, Query Performance Insight, and resolving blocking and deadlocks.
• CI/CD: SQL Database Projects (including SDK-style projects) in source control, unit and integration tests, reference data, branching and pull requests, secrets management, schema drift detection, and pipeline controls such as branch policies and code owners.
• Integration: Data API builder for REST and GraphQL endpoints, Azure Monitor with Application Insights and Log Analytics, and change handling with change event streaming, CDC, Change Tracking, Azure Functions SQL trigger bindings, or Logic Apps.

Implement AI capabilities in database solutions (25 to 30%)

Three topics. Models and embeddings: evaluating and registering external models, choosing how embeddings stay current (triggers, Change Tracking, Functions, Logic Apps, CDC, or Microsoft Foundry), deciding which columns to embed, and chunking text. Intelligent search: full-text, semantic vector, or hybrid search, the vector data type and indexes, VECTOR_DISTANCE, VECTOR_NORMALIZE, VECTORPROPERTY, and VECTOR_SEARCH, ANN versus KNN, and combining results with reciprocal rank fusion (RRF). RAG: building a prompt with sp_invoke_external_rest_endpoint, converting rows to JSON, sending them to a language model, and extracting the response.

What vector search in SQL looks like

The pieces are now built into the engine. A minimal pattern: store embeddings in a vector column, fill it with a registered model, then search.

-- Store embeddings next to the rows they describe
ALTER TABLE dbo.Products ADD embedding VECTOR(1536);

-- Fill them using an external model registered with CREATE EXTERNAL MODEL
UPDATE p
SET p.embedding = AI_GENERATE_EMBEDDINGS(p.Description USE MODEL MyEmbeddingModel)
FROM dbo.Products AS p;

-- Exact (KNN) search: compare the question with every row
DECLARE @q VECTOR(1536) =
    AI_GENERATE_EMBEDDINGS(N'waterproof hiking boots' USE MODEL MyEmbeddingModel);

SELECT TOP (5) p.ProductID, p.Name,
       VECTOR_DISTANCE('cosine', p.embedding, @q) AS distance
FROM dbo.Products AS p
ORDER BY distance;

For larger tables, add an approximate (ANN) index and query it with VECTOR_SEARCH:

CREATE VECTOR INDEX vix_products_embedding
ON dbo.Products (embedding)
WITH (METRIC = 'cosine', TYPE = 'DiskANN');

SELECT TOP (5) WITH APPROXIMATE
    p.ProductID, p.Name, r.distance
FROM VECTOR_SEARCH(
        TABLE = dbo.Products AS p,
        COLUMN = embedding,
        SIMILAR_TO = @q,
        METRIC = 'cosine'
    ) AS r
ORDER BY r.distance;

Three details worth knowing for the exam and for real projects:

• Availability differs by platform. According to Microsoft's vector search documentation, the vector data type and vector functions are generally available in SQL Server 2025, Azure SQL Database, Azure SQL Managed Instance, and SQL database in Fabric. Vector indexes and VECTOR_SEARCH are generally available in Azure SQL Database, SQL database in Fabric, and Managed Instance on the Always-up-to-date policy, but still preview in SQL Server 2025, where you enable PREVIEW_FEATURES first.
• The syntax changed. The VECTOR_SEARCH reference marks the old TOP_N parameter as deprecated; current indexes use SELECT TOP (N) WITH APPROXIMATE and must order by distance ascending. Many blog posts still show TOP_N.
• Latest-version vector indexes need at least 100 rows before you can create them, so seed enough test data.

The study guide notes that most questions cover generally available features, but the exam may include commonly used preview features too.

A six-week study plan with labs

You can practice for free on your own machine: Microsoft's What's new in SQL Server 2025 page says the Standard Developer and Enterprise Developer editions are free for development and test use. To practice features that are generally available only in the cloud, such as the latest vector indexes, use an Azure SQL Database or a SQL database in Fabric, with a budget alert set.

1. Week 1, modern T-SQL. Store product data as JSON, query it with JSON_VALUE and OPENJSON, aggregate with JSON_ARRAYAGG, validate formats with REGEXP_LIKE, and find near-duplicate names with EDIT_DISTANCE.
2. Week 2, special tables and programmability. Build a temporal table and query its history, a ledger table, and a small graph with MATCH. Wrap logic in procedures with proper error handling.
3. Week 3, security and performance. Add Row-Level Security per customer, mask emails with Dynamic Data Masking, encrypt a Social Security number column with Always Encrypted, and turn on auditing. Then create a blocking chain on purpose and diagnose it with DMVs and Query Store.
4. Week 4, CI/CD and APIs. Put the schema in an SDK-style SQL Database Project in GitHub, add tests, deploy through a pipeline, and detect drift. Expose a table and a stored procedure through Data API builder as REST and GraphQL.
5. Week 5, embeddings and search. Register an external model, chunk long text with AI_GENERATE_CHUNKS, generate embeddings, run KNN and ANN searches, then combine full-text and vector results with RRF.
6. Week 6, RAG and review. Call a language model with sp_invoke_external_rest_endpoint using retrieved rows as JSON context, configure a GitHub Copilot instruction file and an MCP connection, then take the practice assessment and revisit weak areas.

Common mistakes

• Treating it as an AI exam. Between 70 and 80% of the weight sits in classic database development, security, and deployment.
• Skipping CI/CD. SQL Database Projects, branching, and drift detection are listed skills with real depth.
• Ignoring platform differences. Know which vector features are preview on SQL Server 2025 and generally available in Azure SQL and Fabric.

How CertTulen Academy can help

CertTulen Academy runs live, MCT-led online training with hands-on labs. Our DP-800: Develop AI-Enabled Database Solutions course prepares you for this exam, and like our other professional certification courses it includes an exam voucher, mock practice tests, and lab access; see the course page for details. If your T-SQL needs strengthening first, Microsoft SQL Server Advanced is a good foundation. Planning a broader move into data roles? Read our guide to Microsoft certifications for career changers, or contact us for a batch that suits your time zone.