Atlan open-sources its analytics-agent playbook—but the artifact link is still missing

Atlan says it has open-sourced the playbook its forward-deployed engineers use to build analytics agents, following more than a year of work with enterprise teams. The announcement came during the company’s Oct. 28 Context Conference session on the “Post-AI Data Stack,” alongside speakers from Workday, DocuSign and Upwork. The first-party conference page says the playbook covers governed metrics, certified queries, evaluations and learning loops for making analytics agents accurate, reliable and cost-effective.

The important caveat is discoverability. At the time of this review, Atlan’s conference page says the playbook was open-sourced but does not link to a repository, package, license or versioned document. The page offers an on-demand recording through a form. Until the artifact itself is linked, practitioners can verify the announcement and the stated scope, but not inspect the implementation, reproduce the evaluations or assess the license.

The proposed stack starts before SQL generation

Atlan frames the session around infrastructure that supplies agents with business meaning before they answer questions. The agenda names five components: governed metrics, certified queries, evals, learning loops and the wider context needed to connect them. That is a useful boundary for analytics-agent teams because it separates model choice from the controls that determine whether an answer is reproducible and reviewable.

The same event included a Verizon architecture session and an ecosystem discussion involving Snowflake, Google, Microsoft and Anthropic representatives about portable context objects and protocols. Atlan also described a “company brain” model in which shared skills, business rules and ontologies are treated like code, with releases, versioning, relationships and usage logs. These are Atlan’s descriptions of conference content, not independently benchmarked product results.

Enterprise evidence is present—but numbers are not yet public

Atlan’s companion page for an MIT Technology Review Insights report says the study polled 250 executives across four regions, 12 industries and organizations with annual revenue from $100 million to more than $5 billion. It says the report examines AI maturity, blockers, ownership and how Verizon and Workday build and test context for agents. The public landing page does not expose the survey findings themselves without a form, so this article does not infer percentages or outcomes from the study.

For practitioners, the next verification step is concrete: look for the promised playbook artifact, its license, the evaluation datasets and the definitions behind “accurate” and “cost-effective.” Open-sourcing a field guide could turn enterprise anecdotes into an inspectable operating method. For now, Atlan has published the scope and the participants; the reproducible artifact still needs a public address.