Dashboards Are Dead — Sarah Simionescu, Composio

AI Engineer · 10 min · 139 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

  1. 00:12Hello everyone.
  2. 00:14My name is Sarah and I have a confession to make.
  3. 00:19So, I've been using Data Dog every day for the last six months.
  4. 00:22And the other day my co-orker comes up to me and he's like, "Hey Sarah, can I look at this alert?"
  5. 00:27I'm like, "Okay."
  6. 00:28So, I open up Data Dog and I'm logged out.
  7. 00:31It'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.
  8. 00:40And I open up the data the data dog dashboard and I just froze.
  9. 00:45I had no idea where anything was.
  10. 00:48And this is actually not a knock on data dog.
  11. 00:50The problem was I had genuinely never opened the dashboard.
  12. 00:55I've been using it every day without ever looking at it.
  13. 00:59And so today, I want to make an argument that sounds kind of insane and then show you that it's obvious.
  14. 01:06The dashboard is dead.
  15. 01:07And 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.
  16. 01:18Alas, how did we get here?
  17. 01:20I believe a postmortem is in order.
  18. 01:22Shall we?
  19. 01:23The year is 2022.
  20. 01:26It is the dark ages.
  21. 01:28Someone pings me on Slack.
  22. 01:29They say, "Hey, I get this error when I try and search for prod."
  23. 01:34I would have instantly spawned five different windows.
  24. 01:37Read the context in Slack.
  25. 01:38Write a query in data dog.
  26. 01:39Check post for this session.
  27. 01:41Uh, fix the bug in VS Code.
  28. 01:44Open the PR in GitHub.
  29. 01:45That is five tools and five UIs I have to learn and relearn every time they ship a new redesign.
  30. 01:51But frankly, redesigns are the least of my problems because every single tool has its own query language.
  31. 01:57So, data dog has its own query syntax.
  32. 01:59Jira has JQL, search, Slack has its search modifiers.
  33. 02:04And 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.
  34. 02:15Every 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.
  35. 02:27You never wanted a dashboard or it's cursed query language.
  36. 02:30You wanted the answer.
  37. 02:32And in 2023, everything changed.
  38. 02:37The sparkle button.
  39. 02:40The sparkle button was born.
  40. 02:42And with a click of this magical button, an LLM would write a sometimes correct query to get you the answer you need.
  41. 02:48That is if your question does not require more than two database joins.
  42. 02:55And soon enough, these sparkling buttons were everywhere.
  43. 02:58So dashboards became AI native.
  44. 03:00Problem solved, right?
  45. 03:02Well, not quite.
  46. 03:04In 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.
  47. 03:14Instead 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.
  48. 03:24And you might be thinking, problem solved, right?
  49. 03:27This is the end of the story.
  50. 03:29Well, far from it, because MCP is a protocol.
  51. 03:32It's a channel for communication between agents and your service.
  52. 03:35And it's up to you, the service, to choose how to communicate with the agent.
  53. 03:40And it turns out that makes all the difference.
  54. 03:43If you've ever actually wired up a dozen MCP servers, you will know the reality is a mess for three reasons.
  55. 03:52Agents don't learn.
  56. 03:53Every conversation starts from zero.
  57. 03:55It has no memory of how it fumbled formatting links properly in Slack yesterday, and so it's just going to fumble again today.
  58. 04:01And we patched this with skills, but skills is just a band-aid because agents become dumber with the more context you provide.
  59. 04:09And loading more tools, more skills takes more context.
  60. 04:12And if you connect enough servers, you're dumping thousands of tool definitions straight into the context window.
  61. 04:17I mean, our GitHub like toolkit alone has over 200 tools.
  62. 04:22The model just drowns.
  63. 04:23It grabs the wrong tool or struggles to resolve dependencies and it can't figure out which one it needs to call first.
  64. 04:29And third, every app is isolated.
  65. 04:32Each MCP server knows about itself and nothing else.
  66. 04:36So the moment a task spans two apps and like they always do, it is your job to piece that together.
  67. 04:44And so MCP gave agents a door into every app.
  68. 04:48But it left them standing in thousands of separate rooms with no map and no memory of ever being there.
  69. 04:54I'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.
  70. 05:03So imagine in this mockup I have my claude connected to the composio MCP.
  71. 05:09What 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.
  72. 05:19And the first thing it does is it calls composio search to state the task it wants to accomplish.
  73. 05:24So in this case it stated three.
  74. 05:26It wants to fetch slack messages.
  75. 05:28It wants to search sentry issues.
  76. 05:29It wants to search data dog logs.
  77. 05:31And for each of these three, Composeio returns not only the correct tools it needs, but actually a plan of how to use them.
  78. 05:39So for example, for Slack, you actually have to find the Slack channel ID before you can start quering for messages.
  79. 05:46Now that it's armed with all the context it needs, it has a plan.
  80. 05:50It starts pulling the message from Slack to see what the user complaint was.
  81. 05:54Then it begins pulling from data sources in parallel.
  82. 05:56Data dog sentry and it begins scanning the codebase.
  83. 06:00Once it identifies the root cause, PR is up with a fix in less than five minutes.
  84. 06:06I didn't have to build a workflow or write a skill to teach it how to do this.
  85. 06:10This is literally just claude using Composeio's MCP.
  86. 06:14And 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.
  87. 06:21And the dashboards I used to open every day are becoming increasingly unfamiliar to me.
  88. 06:27And the impact of designing for agents is measurable.
  89. 06:30So these are some early unreleased uh results comparing Composio against each app's own native MCP that's listed in the clawed uh marketplace.
  90. 06:38So it's the same tasks, it's the same model, and we see a clear difference.
  91. 06:44But why?
  92. 06:45Why is this experience so much better than native MCPs?
  93. 06:49It's because at Composeio, we are developing a brand new interface specifically designed for agents.
  94. 06:56We 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.
  95. 07:05So, 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.
  96. 07:19I'm going to show you a more classic example of like what I usually use it for every day.
  97. 07:23So, let's say I want to dig into some user data.
  98. 07:26I want to see the distribution of what vertical my users select during onboarding.
  99. 07:30I can simply just ask Claude.
  100. 07:32Claude will again run Composio search.
  101. 07:35It will write a query in Post Hog, execute it nice and simple.
  102. 07:38There's my results.
  103. 07:40But what if I wanted to take this up a notch?
  104. 07:43So I want to look at that e-commerce section right there.
  105. 07:46And I want to see what tool kits they like to use.
  106. 07:50Toolkits is another word for apps we like to use at Composio.
  107. 07:54So let's see if cloud can figure out how it can pull this data from metabase uh given the user ids in post hog.
  108. 08:02So once again it uses composio search.
  109. 08:06It 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.
  110. 08:16Then it makes some queries uh in metabase to get the database schema sample some data get a feel for how things are structured.
  111. 08:22And 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
  112. 08:34again without ever loading the entire thing into its context window.
  113. 08:39And there we go.
  114. 08:42So we can see the most popular toolkits amongst this user persona in less than a few minutes.
  115. 08:48Uh and that's data pulled from both sources.
  116. 08:51If you want to play with the live version, come visit us at the Composio booth.
  117. 08:54Uh, I'd love to have you play with this live.
  118. 08:59And so, the dashboard has died.
  119. 09:01And I'm thinking more and more about what this means for the future.
  120. 09:04Many startups have begun making their landing pages and their websites AI friendly for GEO purposes.
  121. 09:11But so few have really prepared their applications to be used by agents.
  122. 09:16Compose began as a tool to help developers.
  123. 09:19We were taking these popular apps that users like and turn them into these tools that agents could use.
  124. 09:28But the humans are tired of dashboards and their agents are tired of poorly designed MCP servers.
  125. 09:33And now we're getting requests from startups saying their clients are begging for a way to use their services through their agents.
  126. 09:41So this is a lesson for everyone building anything right now.
  127. 09:44You are now serving a new species of user.
  128. 09:47They don't have eyes.
  129. 09:48They are not going to click your sparkle button.
  130. 09:51It 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.
  131. 09:59For a decade, we've built pro uh we've built products to be easy for humans to use.
  132. 10:05And the next era belongs to those that are easy for agents to use.
  133. 10:11So, picture me again.
  134. 10:12I'm frozen in front of that data dog dashboard.
  135. 10:14And I was thinking to myself in that moment that I'm falling behind.
  136. 10:17Uh, but now I think of that moment as a preview of what's to come.
  137. 10:22Thank you.
  138. 10:29[music]
  139. 10:37[music]