You can see the query. Can you trust it?

A test for anyone reading about Claude Dashboards this week: search the same publication for its last "dashboards are dead" piece. Dashboards have been declared dead many times, and they keep shipping because people still need a curated view of the numbers that matter.

In Claude, ChatGPT and most data products, the time from question to dashboard is now close to zero, so more people build more of them. That only helps decisions if the numbers on those dashboards are right, and the faster they appear, the harder that is to check. The cost per chart has gone down to the floor. The risk per chart is the exciting new metric that is bouncing back up in its place.

Readable is not the same as right

Claude Dashboards tries to make checking easier. Click any figure and the query behind it appears, and Claude will explain it to anyone who doesn't read SQL. That transparency is a real step forward. That would be enough if everyone had the context about data, metrics and processes that a few analysts hold today.

A query can be readable, right, but ultimately still wrong though. It can join a deprecated table, apply last quarter's definition of revenue, or read from a pipeline that failed overnight. Nothing in the SQL tells the viewer which of those happened, and the dashboard looks just as confident either way.

In enterprises where governance and dashboard creation are evenly matched, human-speed processes dashboards spawl out and inconsistencies compound. In recent years as AI has come into BI tools, we’ve seen creation speed up, and pushing dashboard creation the last mile to end users in chatbots is only going to make these challenges worse and harder to recover from.

The gap is context

Writing SQL is no longer the bottleneck and the truth is it hasn’t been for a long time. Semantic layers, and user-friendly BI tools have been solving that same problem for years. The hard part is knowing which table is certified, which data is healthy, which definition the business uses today, and what each person is allowed to see.

AI also speeds up change. Teams now create metrics, data models and logic as fast as they query them, so context that people maintain by hand falls behind.

Without context, agents guess. They explore schemas and retry queries, spending compute and tokens on an answer that may not match the one a colleague got yesterday. Anthropic says Dashboards will use semantic models exposed through a connector, so the product is built to consume context but that is no different to how every BI tool interacts with semantic layers. The real open, and evergreen question is where that context comes from, and whether it stays current. Connecting every agent to every source over MCP leaves each agent to roam data estates without guardrails, reconcile conflicting answers itself, with more tool calls and tokens every time.

How Euno solves context

Models will keep getting better at inferring context. Doing it consistently, across every team and every query, is a long way off, and organisations won't build lasting trust in AI without that consistency.

Euno maps an organisation's data with end-to-end column-level lineage, from source systems through the warehouse, transformation and semantic layers to BI. Claude (and now Claude Dashboards) or any other agent connects to Euno and gets what it needs (and what it is allowed to have) to query the right, healthy data directly with the right logic. It maps the current state on the first pass and then adds institutional knowledge and suggests new governance rules and knowledge as logic drifts, and the people who own that logic approve them.

Claude Dashboards will only continue to expose the context gap that exists in enterprises implementing AI on their data. But its adoption can in turn help organizations close that same gap. Each dashboard it generates with Euno’s context provides insight into what people are looking for and where additional or updated context can deliver even better results. Euno continuously learns from the queries that all agents make, allowing it to maintain context and improve context at the same velocity that agents consume it.

Time-to-dashboard is now close to zero. Euno's job is to bring time to a trusted answer down with it.

‍