Dataiku — Cobuild general availability (June 2026)

Tag: S-2026-06-16-dataiku-cobuild-ga Type: article (secondary news roundup summarising a vendor press release) Author(s): Tim King, Solutions Review (roundup) — relaying a Dataiku press release Date of source: 2026-06-16 (underlying Dataiku release; surfaced in the Solutions Review week-of-26-Jun-2026 roundup) Date ingested: 2026-06-27 Authority weight: low — secondary roundup of a vendor press release; capability claims are vendor marketing, not independently verified, and the primary Dataiku release was not fetched directly. Raw file: S-2026-06-16-dataiku-cobuild-ga.md. External URLs in the raw stub.

What it claims

Dataiku has announced general availability of Cobuild, an “AI building agent” that converts plain-language business intent into governed, production-ready AI projects “without requiring code or bypassing enterprise controls.” Cobuild sits on the “Platform for AI Success” Dataiku launched earlier in 2026. The claim is that business users can describe an objective in natural language and Cobuild automatically assembles inspectable workflows, data connections and models inside a single governed environment that orchestrates analytics, machine learning and AI agents across multi-vendor stacks.

The governance angle is “governing the build”: keeping citizen/agentic AI development inside enterprise guardrails by default, rather than monitoring it only after deployment. Dataiku Govern is on Paul’s AI-governance watchlist.

Notable quotes

“Dataiku has announced general availability of Cobuild, an AI building agent that turns plain-language business intent into governed, production-ready AI projects without requiring code or bypassing enterprise controls.” — Solutions Review roundup, 26 Jun 2026.

”… automatically assembles inspectable workflows, data connections, and models inside a single governed environment that orchestrates analytics, machine learning, and AI agents across multi-vendor stacks.” — Solutions Review roundup, 26 Jun 2026.

What’s speculative vs. asserted

• Asserted (verifiable existence): that Dataiku announced Cobuild general availability and positions it as a governed, no-code AI building agent on its “Platform for AI Success.”
• Vendor marketing (label as such): “governed, production-ready,” “without bypassing enterprise controls,” “inspectable workflows” — these are vendor capability claims relayed by a roundup, not independently tested.
• Not claimed / not in scope: no named regulated-FS reference customer; no statement that Cobuild’s governance output has been assessed against EU AI Act documentation, SS1/23 / SR 11-7 model-risk, or BCBS 239 evidentiary thresholds.

Topics this feeds

• AI Governance Platforms — adds the “governed agentic build” sub-theme (governance at build-time, not just run-time) to the category synthesis.

Open questions raised

• Does a “governed by default” no-code build agent produce model-risk evidence (lineage, approvals, documentation) that satisfies SS1/23 / SR 11-7 and EU AI Act Art. 11/12 — or does it shift, rather than remove, the assurance burden onto verifying the guardrails themselves?
• How does build-time governance (Dataiku Cobuild) sit alongside run-time governance (OneTrust/IBM observability) in a coherent three-lines-of-defence control map?