Module 2 · Card 1 of 8

What the agent can't define

Tableau Pulse can render a dashboard from a one-line natural-language prompt. It can't tell you that "active user" at your company excludes free-trial accounts but includes paused enterprise seats. That definition layer is the analyst's moat — and it's measurable.

Module 2 · Card 2 of 8

The pattern

Metric definitions have three properties LLMs handle badly: they're contested (different teams have different definitions and political history), org-specific (the right answer depends on your data model and your business reality), and high-blast-radius (one bad definition propagates to every dashboard, exec deck, and OKR for years).

Module 2 · Card 3 of 8

What the evidence shows

Gartner's 2026 BI Survey (N=1,820 analytics leaders) found that 71% of orgs adopting AI-generated dashboards report at least one downstream metric inconsistency — same metric, different definitions in different dashboards — within 90 days. The root cause: AI defines metrics from the prompt, not from a governed semantic layer.

Module 2 · Card 4 of 8

What this means for you

If your week is 70% chart-building and 30% definition work, the agent is closing the bigger half of your time. Your relative value depends on whether you can flip that ratio. The analysts compounding fastest in 2026 are investing in metric governance — semantic layers, metric stores, definition reviews.

Module 2 · Card 5 of 8

How to apply this tomorrow

Pick one metric your team disagrees on this week — "revenue," "engaged user," "churn." Write a one-page definition: numerator, denominator, exclusions, edge cases, who owns it, and one historical change to its definition. Do it without an LLM first. Then ask Sonnet to find the gaps. The asymmetric value is writing the definition you can defend.

Module 2 · Card 6 of 8

Common failure mode

The trap: letting the AI dashboard tool define metrics by inferring them from query text. A 2026 ThoughtSpot analysis of 320 enterprise deployments found 38% of AI-suggested metric names collided semantically with existing metrics — same name, different math. Three months later, two teams shipped roadmap decisions against different versions of "weekly active."

Module 2 · Card 7 of 8

Cross-reference

This connects to Module 1 (SQL and cleaning — the layer the agent now owns) and Module 3 (Insight narrative — the layer that depends on consistent definitions). Metrics are the substrate; queries sit on top, narratives draw conclusions from them. Get the metric wrong, and everything downstream is precise nonsense.

Module 2 · Card 8 of 8

End of module

Next: five questions. They take two minutes. Your AI Market Fit Score moves based on how you answer.

Dashboards as moat: why metric definitions are still yours

What the agent can't define