Sarah Simionescu, Member of Technical Staff at Composio, argues that the traditional dashboard is becoming obsolete for AI agents. She traces the progression from five tools with five query languages, to the AI 'sparkle button', to MCP, and explains limitations of MCP alone: agents don't learn, too many tools overwhelm the model, and each app is isolated. The talk presents an interface designed for agents using tool search and execution plans, with live demos of a Slack bug report to fix PR workflow and a cross-app analysis across PostHog and Metabase without loading full results into context. Topics covered: why dashboards and query languages acted as translation layers, three problems with wiring up many MCP servers, tool search plus execution plans for cross-app work, and designing products for agents as users. Recorded at the AI Engineer World's Fair 2026; the description's framing of the dashboard as 'dead' is the speaker's argument rather than an established fact. # Video Identity Observations — "Dashboards Are Dead — Sarah Simionescu, Composio" ## Speaker & Affiliation Evidence - **Named presenter:** Sarah Simionescu (title). - **Self-introduction:** <blockquote>"Hello everyone. My name is Sarah and I have a confession to make."</blockquote> - **Role at Composio:** <blockquote>"I am my team is responsible for the Composio dashboard."</blockquote> - **Talk format:** Conference-style talk with a live audience ("great news for everyone in this room", "come visit us at the Composio booth"). ## Intended Audience - **Developers and product builders working with AI agents / MCP:** <blockquote>"If you've ever actually wired up a dozen MCP servers, you will know the reality is a mess."</blockquote> - **Startups building software products:** <blockquote>"So this is a lesson for everyone building anything right now."</blockquote> - **Practitioners using dev tooling:** references to Slack, Datadog, Sentry, VS Code, GitHub, Jira, PostHog, Metabase assume a software-engineering audience. ## Claimed Scope - **Argumentative thesis talk:** the creator claims dashboards are obsolete: <blockquote>"The dashboard is dead. And I think that's great news for everyone in this room."</blockquote> - **Historical framing:** traces an evolution timeline — pre-2023 multi-tool workflows, 2023 "sparkle button" LLM queries, November 2024 Anthropic's MCP announcement, and a projected 2026 workflow. - **Critique of MCP's limits:** agents don't learn, context-window overload ("our GitHub like toolkit alone has over 200 tools"), and app isolation. - **Product positioning:** Composio described as a new agent-first interface layer: <blockquote>"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."</blockquote> - **Industry-level claim:** <blockquote>"For a decade, we've built products to be easy for humans to use. And the next era belongs to those that are easy for agents to use."</blockquote> ## Demonstrated (vs. Claimed) Presentation Patterns - **Demonstrated:** mockup demos of Claude + Composio MCP workflows (bug triage across Slack/Sentry/Datadog; cross-source PostHog/Metabase analysis), and mention of "early unreleased results comparing Composio against each app's own native MCP." - **Creator claims, not independently shown in transcript:** the measurable benchmark improvement ("we see a clear difference" — no figures given verbally) and the 2026 workflow vision, which is explicitly a mockup. - **Self-aware marketing posture:** presenter acknowledges conflict of interest ("except probably for me because... my team is responsible for the Composio dashboard") and promotes a booth visit — evidence this functions partly as a product talk within a conference setting. ## Channel-Level Identity Signals (this video only) - Positions the channel/speaker as a **thought leader at the intersection of developer tooling and AI agents**, using narrative storytelling (personal anecdote, historical "postmortem") to deliver a product-adjacent thesis. - No channel self-description metadata beyond this talk was supplied; no broader channel-wide conclusions are drawn from a single video. # Delivery Observations — "Dashboards Are Dead" (Sarah Simionescu, Composio) ## Overall Communication Style The delivery is **conversational narrative-driven technical storytelling**. The talk is structured as a personal anecdote repurposed as a thesis, with a clear arc: confession → argument → history → problem → demo → broader lesson → callback. ## Recurring Rhetorical Patterns - **Rhetorical question as a pivot device.** Used repeatedly to transition between sections: - "Problem solved, right? Well, not quite." - "But why? Why is this experience so much better than native MCPs?" - "Alas, how did we get here? I believe a postmortem is in order. Shall we?" - **"So" as a connective opener** — appears throughout as a transition or softener: "So, I've been using Data Dog…", "So, we can see the most popular toolkits…", "So, picture me again." - **Self-deprecating parenthetical asides** that hedge product claims: - "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." - "an LLM would write a *sometimes correct* query" - **Callback structure.** The opening Data Dog freeze ("I just froze. I had no idea where anything was") is explicitly revisited at the close: "So, picture me again. I'm frozen in front of that data dog dashboard." ## Teaching Habits - **Historical framing via period markers.** The evolution is staged chronologically with vivid era labels: "The year is 2022. It is the dark ages." → "And in 2023, everything changed. The sparkle button." → "In November of 2024, Anthropic announced the MCP protocol" → "how I would solve it in 2026." - **Numbered problem enumeration** for clarity: "the reality is a mess for three reasons. Agents don't learn… And we patched this with skills… And third, every app is isolated." - **Concrete step-by-step walkthrough of demos.** The demo narration narrates the agent's actions in sequence: "the first thing it does is it calls composio search… It wants to fetch slack messages. It wants to search sentry issues. It wants to search data dog logs." - **Clarification asides** for audience context: "Toolkits is another word for apps we like to use at Composio." ## Vocabulary and Phrasing - **Vivid metaphor density**, often extended across a passage: - "MCP gave agents a door into every app. But it left them standing in thousands of separate rooms with no map and no memory." - "The model just drowns." - "skills is just a band-aid" - "You are now serving a new species of user. They don't have eyes." - **Anthropomorphizing of agents**: agents "fumble," become "dumber," get "armed with all the context," and "judge you." - **Informal register mixed with product claims**: "It's kind of awkward," "I mean, our GitHub like toolkit alone has over 200 tools," "let's take this up a notch" (as "What if I wanted to take this up a notch?"). - **Humanizing typography**: refers to the agent by name — "Claude" — as a coworker ("I can simply just ask Claude"). ## Pronoun and Audience Positioning - Frequent direct audience address: "You never wanted a dashboard or it's cursed query language. You wanted the answer." / "It is your job to piece that together." - Inclusive framing in the closing lesson: "So this is a lesson for everyone building anything right now." - First-person confession frame establishes vulnerability before persuasion: "Hello everyone. My name is Sarah and I have a confession to make." ## Structural Signature The pattern **thesis → proof via lived experience → history → technical diagnosis → demo → market claim → callback** appears once here, so its stability as a personal signature is **uncertain**, but the callback and rhetorical-question pivots within it are consistent throughout the single talk. ## Cautions - Only one talk is available, so I cannot distinguish stable personal habits from one-off choices. - Some phrasings appear to be transcription artifacts (e.g., "Data Dog" vs "Datadog," "Enthropic," "clawed marketplace," "reax string" likely "regex string"); I report them as transcribed without correcting them. # Observation: "Dashboards Are Dead," Sarah Simionescu (Composio) ## Source and Attribution - **Speaker:** Sarah Simionescu, identified in the video title as affiliated with Composio. The transcript itself opens with "My name is Sarah" and refers to "my team" being responsible for the Composio dashboard. - **Format:** A conference-style talk. Most content is the speaker's own framing, with one mockup demo and one live-style example. - **Transcript quality:** Several words appear to be mis-transcribed (e.g., "Enthropic" for Anthropic, "Cloud" and "Clawed" for Claude, "quering," "reax" for regex). These are recorded here as the most likely intended terms, not as quoted speech. ## Core Thesis > **Stated position:** "The dashboard is dead." The speaker presents this as an argument that "sounds kind of insane and then show[s] you that it's obvious." The speaker frames the shift as good news for "everyone in this room," with the caveat that it is "probably" not good for the speaker's own team, which maintains the Composio dashboard. ## Reasoning Structure The talk moves through a historical "postmortem" in four stages. ### 1. The pre-AI era (2022) - Debugging a single Slack report required five tools and five UIs (Slack, Datadog, a session tool, VS Code, GitHub). - Each tool had its own query language (Datadog syntax, Jira JQL, Slack search modifiers). - Even tools that claim to use SQL "don't even agree on SQL." - **Stated interpretation:** Dashboards and query languages were "translation device[s]" that existed "because the machines on the other end could not understand what you actually wanted." ### 2. The "sparkle button" era (2023) - LLMs were embedded into dashboards to generate queries from a click. - **Qualification stated by the speaker:** the approach works "sometimes correct" and only "if your question does not require more than two database joins." - The speaker's conclusion: dashboards became "AI native," but the problem was not fully solved. ### 3. The MCP era (late 2024 onward) - The speaker credits Anthropic with announcing MCP in November 2024 "with a promise to create an open standard." - The speaker describes MCP as "a protocol," not a complete solution: "it's up to you, the service, to choose how to communicate with the agent. And it turns out that makes all the difference." ## Claimed Problems with Naive MCP Deployments The speaker lists three problems, stated as the result of "actually wired up a dozen MCP servers": - **Agents don't learn.** Each conversation starts from zero, with no memory of prior failures. - *Partial remedy mentioned:* "skills," which the speaker calls "just a band-aid." - **Context overload.** Agents "become dumber with the more context you provide." The speaker cites a GitHub toolkit with "over 200 tools" and says "the model just drowns," picking wrong tools or failing to resolve dependencies. - **Isolated apps.** "Each MCP server knows about itself and nothing else," so cross-app tasks fall to the user to assemble. ## Mockup Demo: Bug Triage - **Explicitly labeled:** "a mockup demo." - **Scenario:** Paste a Slack message link into Claude connected to Composio's MCP, with the instruction to find the root cause and draft a PR. - **Claimed sequence:** 1. Claude calls Composio search to declare three sub-tasks: fetch Slack messages, search Sentry issues, search Datadog logs. 2. Composio returns tools plus "a plan of how to use them" (e.g., a Slack channel ID must be resolved first). 3. Data is pulled from Slack, then Datadog and Sentry in parallel, then the codebase is scanned. 4. A PR with a fix is produced "in less than five minutes." - **Speaker's claim:** No workflow or skill was built for this; it is "literally just claude using Composeio's MCP." ## Second Example: Analytics Across Two Tools - **Setup:** Find the distribution of user "verticals" from onboarding in PostHog, then identify which toolkits users in the e-commerce segment use, pulling user IDs from PostHog and data from Metabase. - **Claimed mechanics:** - Results are saved without loading full output into context. - Metabase schema is inspected and sampled. - A "Composio Remote Workbench" generates SQL with a regex string to filter user IDs. - **Claimed outcome:** Popular toolkits identified "in less than a few minutes." - **Note:** "Toolkits" is defined by the speaker as "another word for apps." ## Evidence Offered for Performance Claims - The speaker presents "some early unreleased uh results comparing Composio against each app's own native MCP," described as showing "a clear difference" on "the same tasks" with "the same model." - **Limitation:** No figures, task definitions, sample size, or metrics are shown or stated in the transcript. The claim is recorded as the speaker's assertion only. ## Stated Positioning of Composio - Composio is "developing a brand new interface specifically designed for agents." - It "translate[s] messy, sparsely documented, everchanging APIs" into forms agents can use. - It is built "on top of MCP CLI and native tools." - The speaker reports that Composio "began as a tool to help developers" and that startups are now asking for agent access to their services. ## Forward-Looking Claims - **Users are changing:** "You are now serving a new species of user. They don't have eyes... 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." - **Design priority:** "For a decade, we've built products to be easy for humans to use. And the next era belongs to those that are easy for agents to use." - **Adoption signal (self-described):** The speaker's own habits have shifted; opening Claude "has just become like muscle memory," and "the dashboards I used to open every day are becoming increasingly unfamiliar." - **Contrast noted by speaker:** Many startups optimize for "GEO" (AI-friendly landing pages) but "so few have really prepared their applications to be used by agents." ## Tensions, Qualifications, and Unresolved Points - **Self-interest acknowledged:** The speaker notes the thesis is "probably" bad for their own team's dashboard work. The talk is also promotional for Composio, which the speaker explicitly invites audiences to try at its booth. - **Hedged opening:** The argument is introduced as sounding "kind of insane," then presented as "obvious." The speaker does not reconcile these framings. - **Dashboard critique vs. dashboard use:** The opening anecdote attributes freezing to never having opened the dashboard, and the speaker states "this is actually not a knock on data dog." The broader claim that dashboards are dead is not tied to that anecdote beyond the framing. - **Demo status:** The headline bug-fix demo is a mockup. The analytics example is described as real but is not shown with verifiable output in the transcript. - **Performance comparison:** Asserted as "clear" but unquantified. - **Sparkle-button assessment:** The speaker's critique of 2023-era tools is qualified ("sometimes correct," "two database joins") and not tested against any named product. - **Unaddressed risks:** The transcript does not discuss security, permissions, or reliability of agent-executed queries, PR creation, or data access, beyond the speaker's remark that agents must be able to "get the job done." - **Ambiguous remedy:** Skills are called a "band-aid," but the speaker does not state whether Composio's approach replaces skills entirely or complements them. # Speaker and Provenance Observation: "Dashboards Are Dead" ## Video Identification - **Title:** Dashboards Are Dead — Sarah Simionescu, Composio - **Video ID:** 01a15ca7-1b66-7f54-95d3-c7c4b111fef3 - **Format:** Single-speaker presentation (transcript; no timestamps provided) - **Supplied metadata:** Title only. No curated speaker list, cited-source list, or comment data was supplied. ## Speakers and Roles ### Sarah Simionescu - **Role:** Presenter and primary speaker (the only speaker in the transcript) - **Affiliation (stated in transcript):** Says her team is responsible for "the Composio dashboard" and refers to "Composio" as the company she works for. The title also names Composio. - **Self-described experience:** Uses Datadog daily for "the last six months" (anecdotal, first-person). - **Speaker-attributed statements:** Everything in the transcript is spoken by Sarah. No other voices are identified. The transcript ends with a `[music]` marker, which is not attributed to any speaker. ### Other Persons Referenced (not speaking) - **Unnamed co-worker:** Appears in an anecdote about being asked to look at an alert. Not a speaker. - **Unnamed startup clients/startups:** Described second-hand as requesting agent access to services. Attributed to the speaker's account only. ## Referenced Sources and Named Entities These are named in the transcript. No links or citations were given. **Tools and platforms mentioned as used in the workflow:** - Datadog - Slack - Jira (JQL query language) - GitHub (including a "GitHub toolkit" with "over 200 tools," per speaker) - VS Code - Sentry - PostHog - Metabase - Claude **Protocols and technical components:** - MCP (Model Context Protocol) - Composio search, Composio MCP, Composio Remote Workbench, and MCP CLI (Composio-named components) **Third-party claims stated by the speaker (not independently sourced in the transcript):** - In November 2024, Anthropic announced the MCP protocol with a promise of an open standard. The speaker states this; no citation is given. - The speaker says the Claude marketplace lists native MCP servers for each app. The transcript reads "clawed uh marketplace," so the exact name is uncertain. - The speaker says a GitHub toolkit has over 200 tools. No source is given. - The speaker describes a "sparkle button" trend beginning in 2023. This is presented as the speaker's historical account. ## Promotional and Company-Affiliated Segments The following segments are promotional or company-affiliated. They should be treated as the presenting company's own claims. - **Overall framing:** The talk presents Composio's product direction as the answer to the problem it describes ("we are developing a brand new interface specifically designed for agents"). - **Mockup demo of the bug-fix workflow:** Described as "a mockup demo" showing Claude connected to the Composio MCP. The speaker says the PR was produced "in less than five minutes." The transcript does not provide a recording of the run, and the speaker labels it a mockup. - **Post Hog / Metabase example:** Shows Composio-connected workflow results. The transcript shows no data output; it describes results verbally. - **Early unreleased results:** The speaker says these compare Composio against each app's native MCP and show "a clear difference." The transcript provides no data, metrics, methodology, or sample size. Labeled "early" and "unreleased" by the speaker. - **Booth invitation:** "If you want to play with the live version, come visit us at the Composio booth." - **Company self-description:** Statements about what Composio "translates" and "builds," and the claim that Composio "began as a tool to help developers." - **Concluding call to action:** The closing statements frame the next era as belonging to products "easy for agents to use," in line with Composio's offering. ## Availability and Limits - **Comments:** No comment data was supplied. Comment availability is **unknown**. - **Curated metadata:** Only the title was supplied. No speaker bio, description, chapter markers, or source list accompanies the video. - **Unavailable evidence channels:** - No slides, screen recording, or live demo output is present in the transcript; the demos are described verbally. - The "early unreleased results" are not presented in the transcript and cannot be checked from it. - Named third-party claims (Anthropic announcement date, GitHub tool count, marketplace listings) have no supplied citations. - **Transcription uncertainty:** Some names appear in garbled form in the transcript (for example, "Data Dog," "Enthropic," "clawed," "Composeio," "reax string"). The record uses the most likely standard names and flags the marketplace name as uncertain. Hello everyone. My name is Sarah and I have a confession to make. So, I've been using Data Dog every day for the last six months. And the other day my co-orker comes up to me and he's like, "Hey Sarah, can I look at this alert?" I'm like, "Okay." So, I open up Data Dog and I'm logged out. 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. And I open up the data the data dog dashboard and I just froze. I had no idea where anything was. And this is actually not a knock on data dog. The problem was I had genuinely never opened the dashboard. I've been using it every day without ever looking at it. And so today, I want to make an argument that sounds kind of insane and then show you that it's obvious. The dashboard is dead. 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. Alas, how did we get here? I believe a postmortem is in order. Shall we? The year is 2022. It is the dark ages. Someone pings me on Slack. They say, "Hey, I get this error when I try and search for prod." I would have instantly spawned five different windows. Read the context in Slack. Write a query in data dog. Check post for this session. Uh, fix the bug in VS Code. Open the PR in GitHub. That is five tools and five UIs I have to learn and relearn every time they ship a new redesign. But frankly, redesigns are the least of my problems because every single tool has its own query language. So, data dog has its own query syntax. Jira has JQL, search, Slack has its search modifiers. 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. 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. You never wanted a dashboard or it's cursed query language. You wanted the answer. And in 2023, everything changed. The sparkle button. The sparkle button was born. And with a click of this magical button, an LLM would write a sometimes correct query to get you the answer you need. That is if your question does not require more than two database joins. And soon enough, these sparkling buttons were everywhere. So dashboards became AI native. Problem solved, right? Well, not quite. 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. 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. And you might be thinking, problem solved, right? This is the end of the story. Well, far from it, because MCP is a protocol. It's a channel for communication between agents and your service. And it's up to you, the service, to choose how to communicate with the agent. And it turns out that makes all the difference. If you've ever actually wired up a dozen MCP servers, you will know the reality is a mess for three reasons. Agents don't learn. Every conversation starts from zero. It has no memory of how it fumbled formatting links properly in Slack yesterday, and so it's just going to fumble again today. And we patched this with skills, but skills is just a band-aid because agents become dumber with the more context you provide. And loading more tools, more skills takes more context. And if you connect enough servers, you're dumping thousands of tool definitions straight into the context window. I mean, our GitHub like toolkit alone has over 200 tools. The model just drowns. It grabs the wrong tool or struggles to resolve dependencies and it can't figure out which one it needs to call first. And third, every app is isolated. Each MCP server knows about itself and nothing else. So the moment a task spans two apps and like they always do, it is your job to piece that together. And so MCP gave agents a door into every app. But it left them standing in thousands of separate rooms with no map and no memory of ever being there. 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. So imagine in this mockup I have my claude connected to the composio MCP. 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. And the first thing it does is it calls composio search to state the task it wants to accomplish. So in this case it stated three. It wants to fetch slack messages. It wants to search sentry issues. It wants to search data dog logs. And for each of these three, Composeio returns not only the correct tools it needs, but actually a plan of how to use them. So for example, for Slack, you actually have to find the Slack channel ID before you can start quering for messages. Now that it's armed with all the context it needs, it has a plan. It starts pulling the message from Slack to see what the user complaint was. Then it begins pulling from data sources in parallel. Data dog sentry and it begins scanning the codebase. Once it identifies the root cause, PR is up with a fix in less than five minutes. I didn't have to build a workflow or write a skill to teach it how to do this. This is literally just claude using Composeio's MCP. 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. And the dashboards I used to open every day are becoming increasingly unfamiliar to me. And the impact of designing for agents is measurable. 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. So it's the same tasks, it's the same model, and we see a clear difference. But why? Why is this experience so much better than native MCPs? It's because at Composeio, we are developing a brand new interface specifically designed for agents. 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. 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. I'm going to show you a more classic example of like what I usually use it for every day. So, let's say I want to dig into some user data. I want to see the distribution of what vertical my users select during onboarding. I can simply just ask Claude. Claude will again run Composio search. It will write a query in Post Hog, execute it nice and simple. There's my results. But what if I wanted to take this up a notch? So I want to look at that e-commerce section right there. And I want to see what tool kits they like to use. Toolkits is another word for apps we like to use at Composio. 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. So once again it uses composio search. 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. Then it makes some queries uh in metabase to get the database schema sample some data get a feel for how things are structured. 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 again without ever loading the entire thing into its context window. And there we go. So we can see the most popular toolkits amongst this user persona in less than a few minutes. Uh and that's data pulled from both sources. If you want to play with the live version, come visit us at the Composio booth. Uh, I'd love to have you play with this live. And so, the dashboard has died. And I'm thinking more and more about what this means for the future. Many startups have begun making their landing pages and their websites AI friendly for GEO purposes. But so few have really prepared their applications to be used by agents. Compose began as a tool to help developers. We were taking these popular apps that users like and turn them into these tools that agents could use. But the humans are tired of dashboards and their agents are tired of poorly designed MCP servers. And now we're getting requests from startups saying their clients are begging for a way to use their services through their agents. So this is a lesson for everyone building anything right now. You are now serving a new species of user. They don't have eyes. They are not going to click your sparkle button. 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. For a decade, we've built pro uh we've built products to be easy for humans to use. And the next era belongs to those that are easy for agents to use. So, picture me again. I'm frozen in front of that data dog dashboard. And I was thinking to myself in that moment that I'm falling behind. Uh, but now I think of that moment as a preview of what's to come. Thank you. \>\> \[music\]