October 2025 Summaries

CData Connect AI enables Claude to securely query live LinkedIn business data through the Model Context Protocol, providing governed access without copying or transferring information. The integration supports analysis of company pages, engagement, followers, job postings, advertising, and network performance, allowing marketing, recruiting, and business development teams to ask natural-language questions and receive timely insights. Setup requires a LinkedIn Developer account, OAuth credentials, Marketing Developer Platform access for certain data, and appropriate scopes such as r_organization_social for company pages or r_ads and rw_ads for advertising. Users create a LinkedIn connection in Connect AI, authenticate through OAuth, then add the CData Connect AI connector within Claude. The guidance emphasizes validating results against LinkedIn analytics, managing API rate limits, monitoring performance, maintaining audit logs, and following governance and GDPR requirements. Claude cannot directly connect to LinkedIn or autonomously post content or send messages, but it can analyze performance data and assist with drafting content.

CData Connect AI uses the Model Context Protocol (MCP) to connect AI agents such as Claude, ChatGPT, Gemini, and Copilot directly to live API data, avoiding data replication and the delays associated with traditional extract-and-query pipelines. The described setup involves creating an API connection in Connect AI, defining tables and JSON response structures, selecting an authentication method, assigning permissions, and exposing the configured endpoint through Connect AI’s MCP server for use in an AI assistant. Using a public countries API as an example, the workflow enables agents to translate natural-language requests into structured API calls and return real-time results, such as country, population, capital, or currency information. The platform also provides autoscaling, query optimization, caching, monitoring, audit logs, and table-, column-, and row-level access controls, while supporting authentication management and enterprise compliance frameworks including SOC 2, ISO/IEC 27001, and GDPR.

Generative AI can deliver measurable business value when large language models are connected securely to live, governed enterprise data rather than used as isolated tools. Three prominent applications are conversational analytics, in which employees ask natural-language questions of CRM, finance, marketing, or HR systems; AI agents that automate multi-step workflows such as invoice processing, sales follow-up, and customer refunds; and knowledge assistants that search internal documents, summarize information, and draft communications. These use cases require more than an LLM, relying on data connectivity, access controls, agent frameworks, context stores, and user-facing integrations to provide accurate information and safely take action. The material presents CData Connect AI as an integration layer that supplies read/write access to hundreds of business systems through governed connections and MCP, aiming to reduce custom integration work and help organizations move AI pilots into scalable production deployments.

CData Connect AI is presented as a no-code platform for connecting Atlassian Jira with Anthropic’s Claude through the Model Context Protocol, enabling users to query live Jira data in natural language without relying on dashboards, JQL, or data replication. Jira provides project, issue, sprint, workflow, and reporting data, while Connect AI maintains source-system authentication, permissions, and governance as Claude interprets requests about ticket status, blockers, sprint progress, capacity, velocity, and cross-project performance. The setup involves creating a Jira connection in Connect AI using options such as OAuth, API tokens, basic authentication, or SSO, then enabling the CData Connect AI connector in Claude.ai. The guidance recommends defining data scope and permissions before deployment, using specific ticket, sprint, and field references for more accurate queries, and maintaining integrations through performance monitoring, dedicated accounts, audit logging, connector ownership, and regular permission reviews.

CData Connect AI can link Anthropic’s Claude to Xero’s cloud accounting data through the Model Context Protocol, enabling users to ask natural-language questions about invoices, payments, payroll, bank transactions, contacts, and financial performance without manual exports or database queries. The platform retrieves live Xero data through secure APIs while preserving existing authentication, permissions, governance controls, and optional tenant-level restrictions rather than copying data to an external database. Setup requires Xero credentials, appropriate data-access permissions, a defined scope of entities to expose, and selection of PKCE, OAuth, or OAuthClient authentication, followed by configuring a Xero connection in Connect AI and authorizing its connector in Claude. Claude translates user prompts into structured requests that Connect AI sends to Xero, returning current results for queries such as overdue invoices, payroll expenses, and revenue comparisons. The guidance also recommends monitoring API usage and response times, maintaining audit logs, using caching and batching for higher volumes, validating outputs against expected records, and reviewing credentials, permissions, pop-up blockers, and connection logs if authentication fails.

CData Connect AI is presented as a managed Model Context Protocol platform that extends n8n’s low-code, open-source workflow automation capabilities with governed, real-time, semantic access to more than 300 enterprise data sources. Rather than relying on disconnected APIs, data replication, or local exports, it queries systems such as Salesforce, Snowflake, NetSuite, BigQuery, and Zendesk in place, enabling AI agents and workflows to reason about business entities and take actions such as updating records, creating tasks, and triggering processes. The platform emphasizes enterprise governance through source-level permissions, identity propagation, audit logging, and support for standards including SOC 2, ISO 27001, HIPAA, and GDPR. A universal MCP endpoint allows organizations to configure data access once for multiple workflows and agents, while semantic models provide contextual understanding of schemas, metadata, and relationships. An example churn-detection workflow combines CRM, product usage, and support data to assess risk and automatically alert account teams while logging all actions, illustrating the proposed shift from isolated automation projects to scalable, business-aware enterprise orchestration.

Data platforms are increasingly consolidating ingestion, transformation, storage, and governance capabilities, offering convenience while potentially increasing vendor dependency and reducing customer flexibility. CData argues for a component-based, standards-driven integration approach that enables organizations to connect systems across Microsoft, AWS, Google Cloud, on-premises environments, and major data platforms using drivers, pipelines, replication, and change data capture. The company emphasizes that meaningful openness includes visibility into how data is moved, processed, governed, and deployed, rather than reliance on opaque vendor-managed pipelines. It also highlights the continuing importance of hybrid environments, including legacy platforms such as IBM DB2 iSeries, and the need to synchronize their data with modern analytics and AI platforms such as Databricks, Snowflake, and Microsoft Fabric without sacrificing security or reliability. CData presents predictable pricing, transparent performance, interoperability, and customer control as essential principles for sustainable data integration as enterprises continue to operate across multiple clouds, platforms, and legacy systems.

Presto is a distributed SQL engine that federates queries across diverse data sources, while Snowflake is a cloud data warehouse designed for independently scalable compute and storage, secure analytics, and columnar data processing. The material recommends combining Presto’s exploratory, federated-query capabilities with replication into Snowflake for production analytics, using batch ETL or ELT, incremental replication, change data capture, or streaming depending on data freshness and workload needs. CData Sync is presented as a no-code integration platform for connecting Presto and Snowflake, mapping schemas and data types, scheduling jobs, detecting schema drift, monitoring logs, and using Snowflake COPY INTO for parallel loading. Performance guidance includes query pushdown, parallel paging, bulk operations, Snowflake clustering, and autoscaling warehouses, while security recommendations cover OAuth, SSO, Kerberos, TLS encryption, AES-256 storage encryption, least-privilege access controls, and audit logging. The pipeline can also support AI feature stores, LLM access to live Snowflake data through Model Context Protocol, and multi-cloud replication scenarios.

CData’s tenth Vibe Querying episode demonstrates how CData Connect AI, Model Context Protocol (MCP), and Microsoft Copilot Studio can be combined to deploy conversational AI agents within Microsoft Teams. Using a remote MCP server, the agent connects to live enterprise sources such as Salesforce and Zendesk, allowing users to ask natural-language questions, analyze data across systems, retain conversational context, summarize support issues, and create CRM tasks without changing applications. The demonstration shows cross-platform account and ticket analysis, opportunity-risk identification, automated task creation from an email-reported issue, and Copilot Studio activity logs that expose generated SQL, tool calls, error handling, and self-correction. The approach is presented as a way for support, sales, customer success, and other teams to reduce manual reporting and context switching while turning Teams into a shared data and workflow hub.

CData Foundations 2025 highlighted Databricks ecosystem priorities around real-time, federated, and agentic analytics, with speakers emphasizing that AI agents and operational decision systems need low-latency access to current enterprise data. Databricks is advancing serverless access, Delta tables, Unity Catalog, and Lakehouse Federation to support governed, cross-system data use, while CData positions its connectors, change data capture, APIs, and virtualization capabilities as tools for providing live access to SaaS platforms, databases, and other operational systems. The sessions also stressed self-service and low-code pipeline development to reduce dependence on engineering teams, handle issues such as schema drift and authentication, and enable domain users to access data more quickly. Federation and query pushdown were presented as ways to avoid unnecessary data duplication while improving experimentation, resilience, and real-time insight, with CData and Databricks together framed as supporting governed, AI-ready data pipelines and analytics workloads.

At CData Foundations 2025, AWS Senior Technical Account Manager Harshit Kohli argued that enterprise AI adoption is constrained less by model capabilities than by organizations’ ability to provide fresh, governed, high-quality data from dispersed systems. He emphasized that AI infrastructure must accommodate data volume, variety, and real-time velocity, while successful deployments should follow staged, measurable initiatives rather than rushing into large-scale implementations. CData positions its AWS integrations as support for this data foundation by simplifying ingestion into Amazon S3 and Redshift, enabling live connections from enterprise applications to Amazon Bedrock models through Model Context Protocol, and integrating with AWS Glue, Athena, and SageMaker. Its approach aims to reduce pipeline complexity, data movement, and cloud costs through push-down queries, caching, incremental synchronization, and predictable pricing, helping organizations combine governed historical data with current operational information for AI applications.

SF Tech Week highlighted a shift from AI chatbots and pilots toward enterprise-scale automation through autonomous agents that can perform complex, multi-step tasks using live business data. At OpenAI Dev Day, the company introduced AgentKit and Agent Builder for visually designing, evaluating, and deploying agent workflows, alongside an Apps in ChatGPT SDK intended to let custom enterprise applications, dashboards, and systems operate within ChatGPT conversations. IBM also introduced BeeAI, an open-source multi-agent orchestration framework emphasizing interoperability, observability, modularity, and governance for hybrid or regulated settings. The announcements underscore that effective enterprise AI depends not only on foundation models and agent frameworks but also on secure, reliable connectivity to SaaS platforms, databases, ERP and CRM systems, and data warehouses, with the source positioning MCP-compatible integrations such as CData Connect AI as infrastructure for accessing data across more than 300 sources.

SQL Server remains widely used for analytics and operational workloads because of its Microsoft ecosystem integration, hybrid and multi-cloud deployment options, security and compliance capabilities, and recent performance improvements such as Intelligent Query Processing, columnstore indexes, and in-memory OLTP. Modern data pipelines use both ETL, which transforms data before loading, and ELT, which performs transformations within SQL Server, while supporting sources ranging from SaaS applications and enterprise databases to streaming and IoT systems. Transformation approaches include in-flight processing, SQL push-down, and in-database execution, while loading patterns span batch jobs, change data capture for incremental replication, micro-batches, and reverse ETL. The guide groups available platforms into Microsoft-native tools such as SSIS and Azure Data Factory, SaaS ELT services such as Fivetran, Hevo, and Skyvia, and self-hosted or hybrid products including CData Sync, Qlik Replicate, and Informatica, each with differing connector coverage, pricing, deployment, and management trade-offs. It recommends evaluating tools through connector depth, benchmarked performance and CDC latency, scalability, total cost of ownership, and security and deployment controls, noting that driver-based connectors and connection-based pricing may simplify maintenance and cost forecasting for high-volume integrations.

Delta Lake has become a widely used open table format for lakehouse architectures by combining low-cost Parquet storage with transaction logs that provide ACID guarantees, schema enforcement, versioning, deletes, and support for concurrent batch and streaming workloads. Its open interoperability allows a single Delta table stored in Amazon S3, Azure Blob Storage, ADLS, or Google Cloud Storage to be queried by platforms including Databricks, Microsoft Fabric, Apache Spark, Trino, and Presto, reducing the need for duplicate data copies or format conversions. CData Sync’s latest release writes replicated data directly into open Delta tables and provides controls for partitioning, schema evolution, deletion handling, compaction, and cloud storage targets. The approach supports Databricks and Fabric integrations, including Fabric’s OneLake and Open Mirroring options, while platforms such as Snowflake, BigQuery, and Athena can generally access the underlying Parquet files but do not natively use Delta transaction logs or their associated reliability features.

PostgreSQL data pipelines increasingly require near-real-time processing, change data capture (CDC), and support for workloads such as AI, machine learning, vector search, and geospatial data, while facing challenges from schema drift, extensions, and managed-cloud restrictions. The guide evaluates 10 ETL/ELT tools—CData Sync, Fivetran, Hevo Data, Airbyte, Stitch, Matillion, Integrate.io, Talend Open Studio, Pentaho, and Apache NiFi—using criteria including connector coverage, PostgreSQL edition and extension support, performance, deployment flexibility, security, compliance, and pricing. It contrasts common pricing approaches, noting that row-based plans can become unpredictable at high volumes, while connection-, compute-, subscription-, and open-source models offer different tradeoffs. CData Sync is presented as a driver-based, PostgreSQL CDC-capable option for hybrid and secure deployments, whereas Fivetran emphasizes managed SaaS ingestion, Airbyte and open-source tools favor customization and deployment control, and Matillion focuses on cloud-warehouse transformations. Organizations are advised to assess on-premises, cloud, or hybrid requirements; confirm support for extensions such as PostGIS and pgvector; test schema-change handling and CDC performance; and conduct a defined proof-of-value before a broader implementation.

CData Connect AI is presented as a managed Model Context Protocol platform that connects ChatGPT, custom GPTs, actions, and agent workflows to live enterprise data from systems such as CRM, ERP, finance, support, SaaS applications, databases, and APIs. It aims to improve response accuracy and reduce hallucinations by providing real-time, business-specific context, including semantic definitions, relationships, metrics, and rules, while avoiding data replication into warehouses or vector stores. The platform applies existing user identities and permissions, row- and column-level controls, SSO, and audit logging to govern data access and constrain automated actions. Through one connection, organizations can support use cases such as executive dashboards, sales and support risk analysis, domain-specific GPTs, controlled record updates, and cross-functional reporting, with live queries translated into source-native SQL or API requests.

CData Connect AI is presented as an enterprise Model Context Protocol platform that connects Claude to live business data from more than 300 sources while applying centralized security, governance, and performance controls. It enables Claude’s Artifacts, Projects, Claude Code, and autonomous agents to query data in place rather than relying on replicated datasets, helping outputs such as dashboards, analyses, compliance reports, and automated workflows reflect current source-system information. The platform emphasizes identity-based permissions, row- and column-level filtering, audit logging, SSO integration, and a unified policy framework across all Claude use cases. Its semantic modeling capabilities are intended to preserve business definitions, relationships, and calculation logic so Claude can interpret organizational data accurately and reduce hallucinations. CData also highlights SQL pushdown, selective data transfer, caching, and broad connector support as mechanisms for scalable performance, positioning Connect AI as a managed layer that makes Claude’s AI capabilities more suitable for governed enterprise deployment.

CData Foundations keynotes argued that enterprise AI is entering an operational phase in which success depends less on model novelty than on trusted, connected, context-rich data. Speakers from CData, ServiceNow, Google, Sisense, and Argano identified four related requirements: reliable governed data, semantic consistency across systems, strong security and observability controls, and real-time connectivity to enterprise applications. They emphasized that AI agents need business context beyond field names and data types, while governance should enable accountable experimentation rather than delay adoption until data is perfect. CData positioned its platform of enterprise connectors, a universal semantic layer, and Connect AI, a managed Model Context Protocol offering, as infrastructure for exposing system metadata, relationships, and data securely to AI tools. The event’s central message was that organizations can begin practical AI initiatives with existing, imperfect data estates if they prioritize dependable connections, context, and controls as they scale.

Enterprise IT leaders seeking to expand AI adoption must balance rapid business demand with security, governance, accurate data access, scalability, and measurable return on investment. The text argues that AI systems should connect directly to live, authoritative enterprise sources such as ERPs, CRMs, financial databases, Salesforce, NetSuite, and Snowflake, rather than relying on disconnected information that can produce outdated or hallucinated responses. CData Connect AI is presented as a platform for connecting AI assistants and tools including Claude and ChatGPT to governed data while preserving existing authentication, single sign-on, role-based access controls, source permissions, auditing, and compliance requirements such as GDPR and HIPAA. Organizations can establish centralized AI accounts, configure shared data connections, have individual users authenticate with their own source credentials, deploy the CData connector organization-wide, and monitor access through both Connect AI and AI-provider administrative interfaces. The proposed approach aims to reduce unsanctioned AI use, provide reliable natural-language access to business data, and support secure AI deployment across teams and multiple AI platforms from a unified data layer.