The dashboard just stopped being the front door

For fifteen years the way into your product data has been a dashboard. Somebody builds it, everyone else squints at it, and the real question goes into a ticket queue.

That front door is moving. Last week I got to say so to a room full of the people running Burger King’s digital products.

What was the event?

Restaurant Brands International runs an annual Digital Masters meetup for the digital teams behind Burger King. I was there representing Amplitude, on a panel with my peers from Braze and AppsFlyer.

That line-up was the point. Analytics, customer engagement and attribution on one stage. What people do in the app, what you send them, and where they came from. Three platforms that every large consumer brand runs, and that almost never talk to each other without a data team in the middle.

The panel covered four things: how we use AI day to day, how AI is changing the way customers use our platforms, what it takes to run AI responsibly at scale, and the best applications we’re seeing.

How is AI changing the way people use analytics?

Here’s the shift. You stop building charts to answer questions. You ask the question.

Amplitude now runs an MCP server. MCP (Model Context Protocol) is the open standard Anthropic created that lets an AI assistant like Claude use other software as a tool. Plug Amplitude into Claude as a connector and Claude can read your funnels, cohorts, experiments and dashboards directly.

So instead of “can someone build me a chart”, a product manager types:

Which step of checkout lost the most people after last week’s release, and was it worse for new or returning customers?

Claude pulls the funnel, splits it by cohort, compares it to the week before and tells you where to look. Then you ask the follow-up straight away, which is the bit dashboards never let you do.

Setting it up takes about two minutes. In Claude, go to Settings, Connectors, Add custom connector, point it at https://mcp.amplitude.com/mcp (or the EU URL if your data lives there) and sign in. That’s it.

What capability should teams already be using?

Asking questions of their own data in plain English.

Most teams I work with still export to CSV, paste into a spreadsheet and build a pivot table. The tools to skip all of that exist today. The gap isn’t technology. It’s that nobody has shown the team it’s allowed.

The bigger prize is joining the stack up. As Braze, AppsFlyer and Amplitude each become tools the same assistant can call, a question like “did Tuesday’s push campaign bring back customers who came in from paid social, and did they order?” stops needing three teams and a week. It starts needing one prompt.

Is it safe to plug AI into enterprise data?

This is where large companies rightly slow down, and where most of the good questions came from. My answer has four parts.

Permissions don’t change. The Amplitude connector signs in with OAuth and inherits the user’s own access. Claude sees exactly what that person could already see. Nothing more.

Start read-only. Let people ask anything. Make anything that writes (creating a cohort, launching an experiment, sending a message) need a human yes. Read freely, write with sign-off.

Your taxonomy is the real bottleneck. AI is only as good as your event names. If half your events are called button_click_v2_final, Claude will give you a confident answer to the wrong question. Clean tracking plans were always good practice. Now they’re the difference between insight and nonsense.

Keep a trail. Log what the assistant ran and what it returned. When a number ends up in a board deck, you want to know where it came from.

None of this is exotic. It’s the same governance you’d apply to a new analyst, applied to a very fast one.

What comes next?

The coolest uses I’m seeing aren’t chatbots. They’re agents that watch a metric, notice when it moves, work out why across analytics, engagement and attribution data, and hand a person a short brief with a suggested fix.

That’s where this goes. The dashboard doesn’t die. It just stops being the front door and becomes one of the rooms.

Thank you

Huge thanks to the RBI team for having us, and to my fellow panellists from Braze and AppsFlyer for a sharp, honest conversation. Rooms like that, where the people running the products push back on the hype, are where the useful ideas come from.

The dashboard had a good run. The next front door is a question.