Issue #2 · May 13, 2026

The dashboard stopped being the product.

Tableau put a release out last week that admitted the dashboard is no longer the product. Read it twice. Then notice Power BI’s May update shipping Copilot summary buttons on the visual header. Then notice Microsoft Fabric Data Agents going GA at FabCon. Then notice Sigma’s May 8 release moving Agents to deployment. Four vendors, one move, same week.

The pivot is the semantic layer. Definitions, lineage, governance, business context — the thing nobody could sell on its own for a decade — is now the floor every agent stands on. The dashboard becomes the fallback when the agent can’t answer.

For builders the seat being defended is no longer “best chart UI.” It’s “best grounded answer.” Vendors investing in semantic infrastructure are about to win the next two years of BI budgets. The ones still polishing visualization chrome are getting consolidated into the renewal RFP.

What’s actually moving this week

1. Tableau went agentic in public. On May 5, Salesforce announced the Tableau Agentic Analytics Platform — a Knowledge Engine, Model Context Protocol support, and a co-led Open Semantic Interchange initiative with Snowflake and dbt Labs. The framing is the news. Tableau stopped pitching itself as the destination and started pitching the semantic substrate. If you sell against Tableau, the question your renewal champion will ask next is whether you have a semantic story their data team believes. Most competitors don’t, yet.

2. Pinterest cleared $1B with AI doing the lift. Q1 revenue up 18% YoY, MAUs at 631M, and the AI shopping assistant is powering ~80B monthly searches with about half labeled commercial-intent. Then they bought tvScientific. Read it as: small-platform ad businesses that ship working AI features inside the buyer flow are pricing differently than the ones still selling dashboards next to their inventory.

3. Microsoft Fabric Data Agents shipped GA. Announced at FabCon. Same architectural move as Tableau — a natural-language agent on top of OneLake, configured by domain experts, embeddable in Teams and Copilot Studio. If you’re building a BI tool that isn’t on Fabric, your 2027 enterprise renewal already has a “why not just use Fabric” slide in the procurement deck.

4. eMarketer cut its 2026 digital ad forecast. Tariff exposure is the cited driver. 94% of US advertisers say tariffs will pressure budgets; social is forecast to take the heaviest pullback. The Q2 and Q3 advertiser-sentiment data is going to be the most consequential signal of the year for any sales-platform org pricing AI features against advertiser revenue.

What I’d ship in your app this week

Feature one: the pacing-delta explainer, on the advertiser dashboard. An advertiser opens their campaign view, sees a 23% pacing delta on a flighted campaign, hovers over the number. They were about to file a support ticket. Instead a small panel renders the three most likely causes — creative fatigue on the top-spending placement, a bid-floor change in the geo, audience overlap with a paused campaign — each cited back to the row in the data it’s drawn from.

Shape. Retrieval over campaign metadata, the last 14 days of bidding logs, creative inventory, and account-team notes. Single grounded LLM call.
Latency budget. 4 seconds, hover-to-paint.
Cost ceiling. $0.02 per call. Hard cap. If the cost rises above the support ticket it's replacing, the feature dies.
Eval. 200 golden examples lifted from past escalations to campaign managers — what the manager would have said, with citations. Run weekly.
Two weeks in. A 40% reduction in "why is my campaign pacing low" tickets. If you don't see it, the retrieval is broken before the generation is.

Feature two: the pre-meeting brief, on the sales rep’s calendar. Fifteen minutes before any advertiser meeting, an agent assembles a four-bullet brief — spend trend, creative changes since last meeting, open support tickets, account-team notes from the last 30 days. The rep sees it on their phone walking to the conference room.

Shape. Async agent, fires on a calendar webhook. Not a chat surface — a card.
Latency budget. 60 seconds. Generous. The constraint is freshness, not paint time.
Cost ceiling. $0.15 per brief. One brief per meeting per rep per day, hard-capped.
Eval. 80 senior-AE-reviewed briefs, scored on "did this surface something the rep didn't already know."
Two weeks in. 30% of reps reference at least one fact from the brief in the first five minutes of the call, measured on recordings.

Both features ship in two to three weeks with the team you already have. Both have a kill switch wired to the cost-per-call ceiling. Neither needs a new model contract.

Sources

Send me an email and we will talk. If something here landed close to what you're working on, the door is open. No calendar funnel, no pitch deck — I read every note that comes in.

Email me Get next week's issue

← All Signal issues · Drafted with Claude · Edited by Paul Brown