← Back to Analytics
Weekly · Analytics · Week of June 8, 2026

Analytics tools converge on the AI agent as the new front door to data

ai-agentsmcpgenerative-analyticsopen-standardsenterprise-plumbingmonetization
Generated 2h agoDrawn from 7 products

The week in analytics

The dominant pattern across analytics this week is the agent becoming the primary interface to data. Hex is rebuilding the notebook around its AI agent — generative data apps from a prompt, repos and connected systems as agent context, and a move to act as an MCP client that reaches into the rest of the stack. Deepnote is opening its workspace as native context for Codex. Neo4j shipped a CLI explicitly framed for agent use. Chord re-architected its Copilot's SQL generation. Four products, four different starting points, all converging on the same idea: the unit of work is shifting from the artifact a human authors to the answer or app an agent assembles, with humans steering context and review.

Running underneath that is a quieter, steadier theme of platform breadth and enterprise plumbing. Elastic's 9.4 pushed observability and AI orchestration forward in one release; Neo4j poured most of its cycle into Aura capacity, billing APIs, and governance; NocoDB kept widening from spreadsheet-database into project management. The split is clear — the high-velocity names are betting on generative and agentic surfaces, while the rest harden cores, fill out tiers, and chase parity.

Leaders

Hex is the week's clear leader at velocity 7.5 with two sparks. Generative Data Apps lets users build a working data app from a prompt, recasting Hex's core output from authored notebooks to generated applications, while becoming an MCP client opens the agent's context onto external systems through a standard protocol. Together they move the product's center of gravity from 'write the cells' to 'describe what you want.'

Elasticsearch shipped the broadest single release: 9.4 promotes Workflows to GA, expands Agent Builder, and adds native Prometheus and PromQL support. Three surfaces — observability metrics, orchestration, and agent retrieval — advance together, pushing Elastic past 'search engine' framing into both observability and AI-infrastructure territory.

Neo4j concentrated its momentum on Aura, but the directional move is neo4j-cli, a Labs-graduating tool framed as a unified entry point for developer, human, and agent use — the changelog leans on 'AX' (agent experience). Paired with Cypher 25 advancing toward the GQL standard, it is Neo4j's clearest bid yet to be addressable by autonomous agents.

NocoDB broadened beyond its Airtable-style core with a Gantt View that turns date-based tables into draggable project schedules with dependencies and milestones, plus Shared Pages for external publishing. The marquee features landing first on paid and Enterprise tiers signal a deliberate push to monetize the open-source base.

Deepnote wired Codex natively into the workspace, so an agent's exploration starts from a team's real notebooks, scheduled analyses, and integrations rather than a blank slate. It is the payoff of a year spent hardening reproducibility — Git sync, run snapshots, AI usage metering — into context an agent can trust.

Wildcards

Fairing is running against the sector's instinct to keep data in-app. Its Shopify Analytics integration syncs post-purchase survey responses into Shopify Order Metafields, and the Hazel integration pipes the same data into Hazel's analytics — a deliberate bet that merchants will analyze response data wherever they already work rather than inside a separate Fairing dashboard.

Fulcrum is the off-pattern name on capability: while peers chase AI surfaces, it spent the week hardening a field-data-collection core across iOS, Android, and web — tracking geometry, map rendering, and a web video-upload ceiling raised to 5GB. Cross-platform reliability, not new surface area, is the whole story here.

Themes that compounded

  • The AI agent is becoming the front door to data, via MCP clients (Hex), native Codex context (Deepnote), agent-framed CLIs (Neo4j), and Copilot re-architecture (Chord).
  • Open standards are gaining ground: MCP for agent context and GQL/Cypher 25 for graph queries both surfaced as deliberate bets this week.
  • Enterprise plumbing is a steady undercurrent — Neo4j's billing API and capacity bumps, Elastic's multi-branch maintenance — beneath the flashier AI work.
  • Monetization-by-tier is explicit, with NocoDB gating Gantt and other marquee features to paid and Enterprise plans.
  • Generative output is replacing authored artifacts, as Hex shifts from notebooks to prompt-built apps and Deepnote reframes notebooks as agent context.

Watch this week

Watch whether the agent-context race produces concrete reach rather than framing. Hex says it expands what the agent can build and where it pulls context from; Deepnote points to more tools beyond Codex able to read workspace context; Neo4j's AX framing implies an MCP server and structured tool definitions on top of neo4j-cli. The tell will be whether next week's releases name specific external systems and tools these agents can actually act through. On the platform side, watch for Elastic's 9.x backport churn to continue and for NocoDB's paid-gated features to start their community-edition trickle-down — the pace of that trickle is the clearest read on how hard it is leaning into monetization.