Dashboards Are Dead — Sarah Simionescu, Composio
https://www.youtube.com/watch?v=YiFqcu9YA38
AI Engineer · 10 min · 139 sentences · sentence-level transcript from YouTube's caption track; each line begins at the time shown

[00:12] Hello everyone.
[00:14] My name is Sarah and I have a confession to make.
[00:19] So, I've been using Data Dog every day for the last six months.
[00:22] And the other day my co-orker comes up to me and he's like, "Hey Sarah, can I look at this alert?"
[00:27] I'm like, "Okay."
[00:28] So, I open up Data Dog and I'm logged out.
[00:31] It's kind of awkward, so I have to log in and he's like looking over my shoulder and I'm like overly conscious about everything that I'm doing on my computer.
[00:40] And I open up the data the data dog dashboard and I just froze.
[00:45] I had no idea where anything was.
[00:48] And this is actually not a knock on data dog.
[00:50] The problem was I had genuinely never opened the dashboard.
[00:55] I've been using it every day without ever looking at it.
[00:59] And so today, I want to make an argument that sounds kind of insane and then show you that it's obvious.
[01:06] The dashboard is dead.
[01:07] And I think that's great news for everyone in this room except probably for me because I am my team is responsible for the Composio dashboard.
[01:18] Alas, how did we get here?
[01:20] I believe a postmortem is in order.
[01:22] Shall we?
[01:23] The year is 2022.
[01:26] It is the dark ages.
[01:28] Someone pings me on Slack.
[01:29] They say, "Hey, I get this error when I try and search for prod."
[01:34] I would have instantly spawned five different windows.
[01:37] Read the context in Slack.
[01:38] Write a query in data dog.
[01:39] Check post for this session.
[01:41] Uh, fix the bug in VS Code.
[01:44] Open the PR in GitHub.
[01:45] That is five tools and five UIs I have to learn and relearn every time they ship a new redesign.
[01:51] But frankly, redesigns are the least of my problems because every single tool has its own query language.
[01:57] So, data dog has its own query syntax.
[01:59] Jira has JQL, search, Slack has its search modifiers.
[02:04] And it gets worse because even when these tools claim to speak the same language, let's say SQL, they don't even agree on SQL.
[02:15] Every dashboard you've ever used, every weird obscure query language ever written was a translation device between you and your data because the machines on the other end could not understand what you actually wanted.
[02:27] You never wanted a dashboard or it's cursed query language.
[02:30] You wanted the answer.
[02:32] And in 2023, everything changed.
[02:37] The sparkle button.
[02:40] The sparkle button was born.
[02:42] And with a click of this magical button, an LLM would write a sometimes correct query to get you the answer you need.
[02:48] That is if your question does not require more than two database joins.
[02:55] And soon enough, these sparkling buttons were everywhere.
[02:58] So dashboards became AI native.
[03:00] Problem solved, right?
[03:02] Well, not quite.
[03:04] In November of 2024, Enthropic announced the MCP protocol with a promise to create an open standard for connecting AI systems with its data sources and that it did.
[03:14] Instead of using a sparkle button, your own agent, Claude, whom you've already been using every day, could generate the query for you, execute it on your behalf, and give you the answer directly.
[03:24] And you might be thinking, problem solved, right?
[03:27] This is the end of the story.
[03:29] Well, far from it, because MCP is a protocol.
[03:32] It's a channel for communication between agents and your service.
[03:35] And it's up to you, the service, to choose how to communicate with the agent.
[03:40] And it turns out that makes all the difference.
[03:43] If you've ever actually wired up a dozen MCP servers, you will know the reality is a mess for three reasons.
[03:52] Agents don't learn.
[03:53] Every conversation starts from zero.
[03:55] It has no memory of how it fumbled formatting links properly in Slack yesterday, and so it's just going to fumble again today.
[04:01] And we patched this with skills, but skills is just a band-aid because agents become dumber with the more context you provide.
[04:09] And loading more tools, more skills takes more context.
[04:12] And if you connect enough servers, you're dumping thousands of tool definitions straight into the context window.
[04:17] I mean, our GitHub like toolkit alone has over 200 tools.
[04:22] The model just drowns.
[04:23] It grabs the wrong tool or struggles to resolve dependencies and it can't figure out which one it needs to call first.
[04:29] And third, every app is isolated.
[04:32] Each MCP server knows about itself and nothing else.
[04:36] So the moment a task spans two apps and like they always do, it is your job to piece that together.
[04:44] And so MCP gave agents a door into every app.
[04:48] But it left them standing in thousands of separate rooms with no map and no memory of ever being there.
[04:54] I'm going to show you a mockup demo here of how I would solve this bug report uh from the very beginning how I would solve it in 2026.
[05:03] So imagine in this mockup I have my claude connected to the composio MCP.
[05:09] What I would do is I would literally just copy and paste the link to the message from Slack and say please use century data dog find the root cause a draft PR make no mistakes.
[05:19] And the first thing it does is it calls composio search to state the task it wants to accomplish.
[05:24] So in this case it stated three.
[05:26] It wants to fetch slack messages.
[05:28] It wants to search sentry issues.
[05:29] It wants to search data dog logs.
[05:31] And for each of these three, Composeio returns not only the correct tools it needs, but actually a plan of how to use them.
[05:39] So for example, for Slack, you actually have to find the Slack channel ID before you can start quering for messages.
[05:46] Now that it's armed with all the context it needs, it has a plan.
[05:50] It starts pulling the message from Slack to see what the user complaint was.
[05:54] Then it begins pulling from data sources in parallel.
[05:56] Data dog sentry and it begins scanning the codebase.
[06:00] Once it identifies the root cause, PR is up with a fix in less than five minutes.
[06:06] I didn't have to build a workflow or write a skill to teach it how to do this.
[06:10] This is literally just claude using Composeio's MCP.
[06:14] And it has become so much easier to do things this way that opening up Cloud has just become like muscle memory for me to do literally anything.
[06:21] And the dashboards I used to open every day are becoming increasingly unfamiliar to me.
[06:27] And the impact of designing for agents is measurable.
[06:30] So these are some early unreleased uh results comparing Composio against each app's own native MCP that's listed in the clawed uh marketplace.
[06:38] So it's the same tasks, it's the same model, and we see a clear difference.
[06:44] But why?
[06:45] Why is this experience so much better than native MCPs?
[06:49] It's because at Composeio, we are developing a brand new interface specifically designed for agents.
[06:56] We translate messy, sparsely documented, everchanging APIs from all of the apps that you live in every day into something that agents really love to use.
[07:05] So, we build on top of MCP CLI and native tools to deliver a cohesive, unified, and excellent experience for your agents that allows them to perform complex operations across these apps seamlessly.
[07:19] I'm going to show you a more classic example of like what I usually use it for every day.
[07:23] So, let's say I want to dig into some user data.
[07:26] I want to see the distribution of what vertical my users select during onboarding.
[07:30] I can simply just ask Claude.
[07:32] Claude will again run Composio search.
[07:35] It will write a query in Post Hog, execute it nice and simple.
[07:38] There's my results.
[07:40] But what if I wanted to take this up a notch?
[07:43] So I want to look at that e-commerce section right there.
[07:46] And I want to see what tool kits they like to use.
[07:50] Toolkits is another word for apps we like to use at Composio.
[07:54] So let's see if cloud can figure out how it can pull this data from metabase uh given the user ids in post hog.
[08:02] So once again it uses composio search.
[08:06] It queries the user ids of those who selected e-commerce from post hog and it actually saves it without loading the full results into its context.
[08:16] Then it makes some queries uh in metabase to get the database schema sample some data get a feel for how things are structured.
[08:22] And once it's confident, it actually uses a very cool tool called Composio Remote Workbench to dynamically generate an SQL query with a reax string to search for all those user IDs
[08:34] again without ever loading the entire thing into its context window.
[08:39] And there we go.
[08:42] So we can see the most popular toolkits amongst this user persona in less than a few minutes.
[08:48] Uh and that's data pulled from both sources.
[08:51] If you want to play with the live version, come visit us at the Composio booth.
[08:54] Uh, I'd love to have you play with this live.
[08:59] And so, the dashboard has died.
[09:01] And I'm thinking more and more about what this means for the future.
[09:04] Many startups have begun making their landing pages and their websites AI friendly for GEO purposes.
[09:11] But so few have really prepared their applications to be used by agents.
[09:16] Compose began as a tool to help developers.
[09:19] We were taking these popular apps that users like and turn them into these tools that agents could use.
[09:28] But the humans are tired of dashboards and their agents are tired of poorly designed MCP servers.
[09:33] And now we're getting requests from startups saying their clients are begging for a way to use their services through their agents.
[09:41] So this is a lesson for everyone building anything right now.
[09:44] You are now serving a new species of user.
[09:47] They don't have eyes.
[09:48] They are not going to click your sparkle button.
[09:51] It shows up with a goal and a set of tools and it judges you on exactly one thing, whether it can get the job done.
[09:59] For a decade, we've built pro uh we've built products to be easy for humans to use.
[10:05] And the next era belongs to those that are easy for agents to use.
[10:11] So, picture me again.
[10:12] I'm frozen in front of that data dog dashboard.
[10:14] And I was thinking to myself in that moment that I'm falling behind.
[10:17] Uh, but now I think of that moment as a preview of what's to come.
[10:22] Thank you.
[10:29] [music]
[10:37] [music]