Introducing Zenitya Artifacts

We started Zenitya Artifacts after noticing the same pattern across several teams we worked with: an analyst, a script, or an agent would arrive at a genuinely useful answer, and then a person would spend the rest of the afternoon turning that answer into something the rest of the business could actually read. The analysis was finished, however the work of presenting it had only just begun, and that second half rarely received the attention it deserved.

##The gap between an answer and a report

When it comes to business analysis, the answer is almost never the deliverable. A number in a notebook, a query result in a terminal, or a long message from an agent might be correct, however it is not yet a report that a stakeholder can open, scan, and act on. This means that someone has to take the raw output and rebuild it as a readable artifact, usually by copying figures into slides, exporting charts as screenshots, or writing a summary by hand. In our experience this step is where most of the time goes, and it is also where mistakes tend to creep in, because the presentation layer is reconstructed from scratch every single time.

This problem has become more visible as more of the analysis itself moves to agents and automated workflows. A Slack bot, a scheduled job, or an internal script can now produce a competent answer on demand, however the output still arrives as text or a raw payload, which is difficult to share with someone who simply wants the result. The fact that the analysis is automated makes the manual formatting step feel even more out of place, because everything around it has been automated except the part that a human still has to assemble.

##What Zenitya Artifacts does

Zenitya Artifacts is the output layer that sits between the analysis and the people who need to read it. In simple terms, an agent, bot, script, or scheduled job sends structured content, and Zenitya Artifacts renders it into a consistent report at a shareable URL. The team that receives the link does not see a notebook or a payload, they see a page with the right blocks (e.g., metrics, charts, tables, and recommended actions) arranged in a way that reads like a report rather than a data dump.

• A schema describes the report as a set of trusted blocks, so the agent provides content while the organization controls layout, theme, and which components are allowed.
• Reports are temporary by default and shareable by link, so an ad hoc question does not have to become a permanent dashboard.
• Access and visibility are controlled at the workspace level, so a report can be shared without exposing the systems behind it.

##Why a schema, and not free-form output

An obvious alternative would be to let the agent generate the report directly, for example as HTML or Markdown. However, free-form output tends to be inconsistent, hard to govern, and difficult to keep on-brand across many teams and clients. This is because every generation is a fresh attempt, and small differences in wording or structure accumulate into reports that no longer look like they came from the same organization. By rendering from a schema instead, we let the agent decide what to say, while the workspace decides how it is presented, and this way the content stays flexible, however the presentation stays predictable.

The agent should own the content of a report, and the organization should own its form. Most of the problems we have seen come from collapsing those two responsibilities into one.

##Where we are starting

The first release is deliberately focused. It renders schema-defined reports from a JSON specification, applies a workspace theme, and returns one controlled URL that can be shared with a team or a client. Snapshot reports work first, and dynamic data blocks, which refresh selected sections from live sources, are in progress. We would rather ship a narrow product that does one thing reliably than a broad one that does many things loosely, and the roadmap from here (e.g., custom workspace themes and finer access controls) is already visible from this foundation.

If your team already uses agents, scripts, or scheduled jobs to answer business questions, Zenitya Artifacts is meant to remove the last manual step between that answer and the people who need it. In summary, we are not trying to replace the analysis, we are trying to give it a readable destination. We will write about each piece as it lands, and we are glad to have you reading this early.

##Further reading

• OpenAI: Introducing Structured Outputs in the API
• Gartner: 75% of new analytics content will use generative AI by 2027