DP-800 · Associate

DP-800 Cheat Sheet: Quick Reference for the SQL AI Developer Associate Exam

All three domains condensed to reference tables, plus decision tables for the security, vector-search, and change-handling features candidates mix up most.

Quick Facts

Credential: SQL AI Developer Associate
Level: Associate (Intermediate)
Domains: 3
Passing score: 700 (scaled)
Prerequisites: None formal
Renewal: Annual, free
Duration / question count / price: Not published on the DP-800 exam page as of Sept 7, 2026
Skills measured as of: March 12, 2026

On duration, question count, and price: Microsoft's general policy for Associate-level role-based exams without labs is roughly 100 minutes, and most Microsoft Certification exams run 40-60 questions at a typical Associate price of $165 USD — but none of that is stated on DP-800's own pages, so treat it as a baseline, not a confirmed figure.

1. Design and Develop Database Solutions (35-40%)

ConceptAreaQuick note
Column store indexesTable designOptimizes large analytical tables for aggregation-heavy queries
In-memory, temporal, external, ledger, graph tablesSpecialized tablesEach solves a different need: speed, history, external data, tamper-evidence, relationships
JSON columns & indexesSemi-structured dataNative JSON storage without a separate document database
PRIMARY KEY / FOREIGN KEY / UNIQUE / CHECK / DEFAULTConstraintsEnforce data integrity at the table level
SEQUENCESObject typeNumber generation independent of any one table, unlike IDENTITY
PartitioningTable & index designSplits large tables/indexes for manageability and performance
Views, scalar functions, table-valued functionsProgrammabilityEncapsulate and reuse query logic
Stored procedures, triggersProgrammabilityProcedural logic and automatic reactions to data changes
CTEs, window functionsAdvanced T-SQLReadable recursive queries and running/ranking aggregates
JSON_OBJECT / JSON_ARRAY / OPENJSON / JSON_VALUEJSON functionsBuild and shred JSON directly in T-SQL
REGEXP_LIKE / REGEXP_REPLACE / REGEXP_SPLIT_TO_TABLERegex functionsPattern matching without CLR or app-side logic
EDIT_DISTANCE / JARO_WINKLER_DISTANCEFuzzy matchingApproximate string similarity scoring in T-SQL
MATCH operatorGraph queriesTraverses node and edge graph tables
GitHub Copilot & Copilot in FabricAI-assisted toolsConfigure model/MCP options, instruction files, MCP server endpoints

2. Secure, Optimize, and Deploy Database Solutions (35-40%)

ConceptAreaQuick note
Always Encrypted / column-level encryptionData securityProtects data from anyone reading the column in plaintext, including DBAs
Dynamic Data MaskingData securityObscures a value in query results without touching stored data
Row-Level Security (RLS)Data securityRestricts which rows a query can return, not which columns
Object-level permissionsData securityGrants or denies access to specific tables, views, procedures
Passwordless access / Managed IdentitySecure accessRemoves stored credentials from connection strings
AuditingComplianceTracks who did what, for compliance review
Execution plans / DMVs / Query Store / Query Performance InsightPerformance tuningDifferent lenses on the same slow-query diagnosis
Blocking & deadlocksPerformance issuesConcurrency problems resolved via isolation levels
SQL Database Projects (SDK-style)CI/CDSource-controlled, buildable database schema models
Schema drift detectionCI/CDCatches manual out-of-band changes before deployment
Branching, pull requests, code ownersCI/CDReview discipline applied to schema like application code
Data API builder (DAB)Azure integrationAuto-generates REST and GraphQL endpoints from database objects
Change Event Streaming (CES) / CDC / Change TrackingChange handlingThree overlapping ways to detect and react to data changes
Azure Monitor / Application Insights / Log AnalyticsObservabilityRecommended monitoring targets for deployed solutions

3. Implement AI Capabilities in Database Solutions (25-30%)

ConceptAreaQuick note
External models (multimodal, multilanguage, structured output)Model evaluationChoosing a model by size, language, and output-format needs
Embedding maintenance methodEmbeddingsTable triggers, CDC, CES, Change Tracking, Azure Functions, Logic Apps, or Microsoft Foundry
ChunkingEmbeddingsHow source text is split before embeddings are generated
Full-text vs. semantic vector vs. hybrid searchIntelligent searchKeyword match vs. meaning match vs. both combined
VECTOR data type & vector indexesVector searchNative storage and indexing for embedding vectors
VECTOR_DISTANCE / VECTOR_NORMALIZE / VECTORPROPERTY / VECTOR_SEARCHVector functionsThe core T-SQL vector search surface
ANN vs. ENNVector searchApproximate (fast, scalable) vs. exact (precise, smaller-scale) nearest neighbor
Reciprocal Rank Fusion (RRF)Hybrid searchCombines full-text and vector result rankings into one score
sp_invoke_external_rest_endpointRAGCalls an external language model directly from T-SQL
Structured data to JSON, send, extract responseRAG pipelineThe three steps of a RAG flow built in T-SQL

Security Feature Quick Decision

Four security features that solve different problems but get tested against each other constantly:

If the scenario says…Use this
Hide a value in an app's query results without touching storageDynamic Data Masking
Prevent anyone, including a DBA, from reading a column in plaintextAlways Encrypted / column-level encryption
Let users query a shared table but only see their own tenant's rowsRow-Level Security
Block a role from a specific stored procedure or view entirelyObject-level permissions
Remove passwords from a connection stringPasswordless access with Managed Identity

Vector & Hybrid Search Quick Decision

The exam describes a data volume and accuracy requirement, not the acronym directly:

If the scenario says…Use this
Small vector set, guaranteed exact nearest neighbors requiredENN (Exact Nearest Neighbor)
Large-scale vector set, approximate results acceptable for speedANN (Approximate Nearest Neighbor)
Need both keyword and meaning-based relevance in one ranked resultHybrid search with Reciprocal Rank Fusion
Query is a literal keyword or phrase matchFull-text search
Query is conceptual ("find similar meaning")Vector search

Change-Handling Quick Decision

CDC, Change Tracking, CES, and SQL trigger bindings all detect that something changed — the exam tests which one fits the stated latency and detail need:

If the scenario says…Use this
Need row-level before/after values for auditingChange Data Capture (CDC)
Just need to know which rows changed since last checkChange Tracking
Need a continuous, near-real-time change stream to a target systemChange Event Streaming (CES)
Need to trigger a function the moment a specific row changesAzure Functions with a SQL trigger binding

DP-800 Cheat Sheet FAQ

What's the difference between Always Encrypted and Dynamic Data Masking?

Always Encrypted keeps a column unreadable in plaintext to anyone, including DBAs. Masking only changes how a value looks in query results — the stored value is untouched.

What's the difference between ANN and ENN vector search?

ENN guarantees exact closest matches but scales poorly at high volume. ANN trades a small amount of accuracy for speed and scale.

Does DP-800 cover Microsoft Fabric as well as SQL Server?

Yes — Microsoft's audience profile explicitly spans SQL Server, Azure SQL, and SQL databases in Microsoft Fabric as one T-SQL surface.

Can I print this DP-800 cheat sheet?

Yes — every table here is plain HTML, so a browser print or "print to PDF" renders cleanly.

MSCertQuiz sells practice-exam access for DP-800 and other Microsoft certifications; this cheat sheet is written by the same team that builds those questions.

More DP-800 Resources

Turn This Cheat Sheet Into Practice Reps

Start with 40 free DP-800 questions covering every domain above.

Start Free DP-800 Practice