TL;DR – What you’ll learn
- Highlights of AI‑powered agent and model innovations from Snowflake, Google Cloud, Precisely, Zenlytic, Fivetran, Zerve, and Dataiku.
- How structured & unstructured data meet generative AI in scalable platforms—with demos across Cortex Agents, Gemini models, AI storytelling, data orchestration, and flexible agent workflows.
- Key decision criteria: governance, model performance, real‑time interaction, analytics transparency, infrastructure automation, and end‑user trust.
🆕 Demo Highlights: Who Presented What
🧠 Snowflake – Cortex Agents with Cortex Analyst & Search
Speaker: James Chawarly, Sr. Developer Advocate @ Snowflake
- Agents seamlessly query both structured and unstructured data.
- Cortex Analyst generates SQL from semantic models; Cortex Search pulls relevant info from documents.
- Users get accurate results, dashboards, SQL code, and revision feedback via a single interface.
- Example: Insurance claims agent interface that counts open claims, surfaces stats via SQL, and retrieves specific policy citations with Data‑to‑Charts visualizations.
🧩 Google Cloud – Gemini 2.0/2.5 via Vertex AI
Speaker: Holt Skinner, Developer Advocate @ Google Cloud AI
- Gemini 2.5 Pro supports up to 2M‑token contexts, multimodal inputs, real‑time APIs.
- Adds built‑in chain‑of‑thought reasoning, grounding via Google Search with URL citations, function calling, long‑form structured outputs.
- User demo showed reasoning logs, error debugging, and controlled layout generation.
- Tools: Gemini dev SDK (
google-genai), Vertex AI Studio, and GitHub-hosted notebooks covering vision, code generation, math reasoning, and grounded responses.
🌐 Precisely – Spatial Context via Data Graph API
Speaker: Noan Young, Sr. Director, Product @ Precisely
- GraphQL interface unifies parcel, building, flood, foot‑traffic, POI datasets.
- Use cases: site project evaluation—foot traffic, competitor density, risk exposure—within a single spatial query.
- Demo: Enriching a Denver address with building/business attributes and mapping Vegas neighborhoods for location‑based forecasting.
⚙️ Zenlytic – Conversational Data Science & Agent‑First Workflow
Speaker: Ryan Jensen, CEO & Co‑founder @ Zenlytic
- Users ask in plain language (chat or voice); agents (“Zoe”) respond with insights, visuals, and optional SQL for validation.
- Power users can trigger clustering, segmentation, Python-driven charts—all within the conversation.
- Examples: ROI breakdown by campaign, persona clustering (“Premium Spenders”, “Subscription Enthusiasts”) with scatter visualizations.
🔁 Zerve AI – Multi-Agent Orchestration and Compute Abstraction
Speaker: Greg Michaelson, Co‑Founder & CPO @ Zerve AI
- Zerve OS is a code-native canvas that abstracts infrastructure, orchestration, and distributed compute. It supports Python/R/SQL and transforms casual instruction into DAG workflows.
- Agents understand project context, generate pipelines, provision compute, fix errors, and scales tasks using The Fleet abstraction.
- Demo: CSV-based “probability of default” model workflow: agent-created blocks, error corrections, EDA, modeling, and runtime orchestration—all visually traceable.
- Zerve provides a free Community Tier that includes serverless compute, Git integration, multi-language support, and public canvas templates.
📊 Dataiku – Trend Detection & Visual Storytelling
Speaker: Valentina Pina Vivas, Sr. Data Scientist @ Dataiku
- Point‑and‑click LLM prompt engineering, schema cleaning, automation.
- Live dashboards reacting to social media trends with built‑in model triggers.
- Auto-generated slide decks populated from connected, live datasets, with AI-powered story flow.
- Additional features: Slack alerts for anomalies, branded AI chat UI, and conceptual AI‑inspired visual art for ice cream.
🔍 Questions Answered in This Session
- How can enterprises unify structured and document data for LLM-driven agents?
Platforms like Snowflake’s Cortex use semantic catalogs and unified APIs to bridge SQL and text retrieval. - What enterprise-grade LLMs support long-context real-time workflows?
Gemini 2.5 Pro in Vertex AI with reasoning capability, function calling, and grounded search support. - How to integrate geospatial context natively?
Precisely’s Graph API merges multiple vetted external datasets with native location awareness for richer insights. - How do business users access data insight without SQL?
Zenlytic’s conversational agent interface abstracts complexity yet delivers summaries, charts, and code reference. - How do you operationalize AI by automating data ingestion and infrastructure?
Zerve automates workflow generation, orchestration, compute scaling, and pipeline deployment from plain-language instructions. - How do you enable flexible, GenAI-ready data connectivity at enterprise scale?
Fivetran’s quick connector rollout, schema tracking, and syncing into lakes or warehouses. - How can storytelling and data trend insights be automated for executives?
Dataiku provides live dashboards, slide automation, prompts, and scenario alerting tied to underlying data.
✅ Executive Takeaways for Data & AI Leaders
- Governance matters: Choose platforms with catalogs, RBAC, citation-style retrieval, and modular workflow control (e.g. Cortex, Vertex, Zerve, Precisely).
- Build trust: Accompany agent answers with SQL transparency, citations, visual context, and feedback loops.
- Support real-time and large-context workflows: Models like Gemini Pro deliver highly capable, multimodal, low-latency reasoning.
- Start use-case first: Begin with a focused domain like claims, marketing, or site evaluation, then expand with successful demos.
- Streamline pipelines and computing: Tools like Fivetran, Zerve, and Dataiku help reduce time-to-data and shift engineering focus to insight.
- Automate storytelling: Dynamic slide decks, alert triggers, conversational UIs, and interactive graphs speed executive adoption of insights.
