Ramp Glass Playbook: Internal AI Workspaces for Company-Wide AI Adoption

Ramp had already hit 99 percent adoption of AI tools across the company when it noticed the actual problem: most people were stuck anyway. Adoption was never the bottleneck. As the Ramp team that built the fix put it, "Terminal windows, npm installs, and MCP configurations were too much for most people to grok, and the few who pushed through had wildly different setups, with no way to share what they'd learned." So Ramp built Glass, an internal AI productivity suite that hands every employee a workspace that is already configured. The lesson is not a clever model choice -- Ramp's own section heading is "The Models are Good Enough, The Harness Isn't." This article distills that playbook into something other companies can borrow.

Key Takeaways

• 99 percent AI adoption came before Glass; Glass fixed what happened after people adopted
• Models are good enough; setup friction is the actual bottleneck
• One Okta SSO sign-in connects every internal tool, including Ramp's home-grown ones
• A skills marketplace turns one person's win into the team's baseline
• Persistent memory from Slack, Notion, Linear, and Calendar means agents do not start cold
• Cron-style automations let agents work overnight and post to Slack
• Ramp Research, the one Glass-connected tool Ramp documented itself, reports a 10-20x rise in questions asked -- the harness thesis as a measurement, not a claim
• Build vs buy: owning the tool is what lets you ship a fix the same day someone gets stuck

What Is Ramp Glass?

Glass is Ramp's internal AI productivity suite for its own employees. In the words of the team that built it: "So we decided to build our own AI productivity suite to make every employee an AI power-user without the pain of having to configure their environment. We've called it Glass."

Concretely, Glass bundles five things into one pre-configured environment:

Two things Glass is not. It is not a product you can buy: Ramp has published the architecture and the lessons, but Glass appears on no Ramp product, pricing, or announcements page as of July 29, 2026, and Ramp has announced no external release. And it is not related to Google Glass, the discontinued eyewear -- the shared name is a coincidence.

The write-up that describes it, "We Built Every Employee at Ramp Their Own AI Coworker", was published on April 9, 2026 by Sebastien Goddijn with Shane Buchan, Cameron Leavenworth, Calvin Kipperman, Jay Sobel, and Caroline Horn. Everything below is drawn from it.

The Harness vs Models Thesis

Ramp titled this section of its own write-up "The Models are Good Enough, The Harness Isn't." That is the part most internal AI strategy decks get wrong.

The diagnosis, in the Ramp team's own words: "It wasn't that the models weren't good enough or that people lacked ambition, they just had no idea how to improve their set up." Their sharper image for it: most people were "driving a Ferrari with the handbrake on." A frontier model is impressive on a benchmark and useless if the salesperson cannot connect it to Gong, the support agent cannot read a Zendesk ticket, and the finance lead cannot pull yesterday's spend. Setup friction, not capability, kept AI usage from turning into AI leverage.

Note the ordering, because it is the detail most summaries of Glass get backwards: the 99 percent number came first, and Glass was the response to what Ramp found underneath it. Nearly everyone had adopted an AI tool. Almost nobody had a good setup, the people who did had built wildly different ones, and none of that work was shareable. Glass exists to close that second gap, not the adoption gap.

The shift is conceptual: stop treating AI as an app you launch and start treating it as a workspace your employees already live inside. Leverage follows the path of least resistance, and that path is a workspace that opens already configured.

The harness includes the model, but it also includes authentication, integrations, memory, the catalog of skills the team has authored, the scheduler that runs jobs overnight, and the multi-pane interface that lets an analyst work three threads at once. Most of that work is unsexy. All of it compounds.

Weekly operating playbooks

Get the next AI workspace playbook before it turns into a backlog.

One practical breakdown each week on AI tools, internal workflows, and the systems that make agents useful at work.

Send me the next playbook

SSO: Making Setup One Click Instead of a Quest

The first piece of the harness is single sign-on wired through Okta SSO. When a new Ramp employee opens Glass, they do not chase API keys, install CLIs, or read a setup guide. In Ramp's description, people "sign in once via their Okta SSO, and all Ramp's tools become available to them with a one-click setup." That includes Ramp's own home-grown products -- Ramp Research, Ramp Inspect, and the Ramp CLI -- alongside the third-party systems the work actually lives in: Slack, Notion, Linear, Gong, Salesforce, Zendesk, and Snowflake. The agent reads and writes against those systems immediately, scoped to that employee's permissions.

Compare that to the standard internal AI rollout. An enthusiastic engineer connects three tools in 20 minutes; the rest of the company hits a permission wall, files a ticket, and never comes back. By centralizing integrations behind SSO, Ramp turned a setup quest into a sign-in. The bar to first useful action drops from hours to seconds.

The architectural detail worth borrowing: integrations are owned by a platform team, not by individual employees. New tools onboard centrally, get audited once, and become available to everyone. That is the only model that scales beyond a handful of power users.

Ramp Research: The One Home-Grown Tool Ramp Documented Itself

Of the three internal products Glass connects, Ramp Research is the only one Ramp has written up on its own engineering blog, and that post lands roughly seven months before the Glass write-up. It is the closest thing there is to a first-party account of what a Glass-connected tool actually does, so it is worth reading alongside the Glass architecture rather than as a footnote to it.

In its builders' words, "Ramp Research is our in-house AI analyst agent, able to answer data questions directly in Slack, 24/7, in minutes -- not hours." The post is credited to Faiz Hilaly, Cesar Duran, and Jay Sobel; Jay Sobel is also one of the six authors of the Glass write-up, which is the direct link between the two documents.

The numbers Ramp published are the strongest evidence on this page for the harness thesis, because they measure demand rather than capability:

Read that last row carefully. A 10-20x rise in questions asked is not a model getting better at answering. It is the price of asking falling far enough that people ask the questions they were previously swallowing. That is the harness thesis stated as a measurement, and it is why this section belongs in the playbook: the tool did not make analysts smarter, it made the queue disappear.

The architecture is borrowable, and it is deliberately not just retrieval. Ramp aggregated and indexed metadata from dbt, Looker, and Snowflake, then had domain owners contribute technical documentation into "a file system that Ramp Research can access." Instead of leaning on retrieval alone -- "Rather than rely exclusively on generic compression methods...we gave the agent tools to inspect column values, branch, and backtrack" -- the agent interrogates the data the way an analyst would. Threads are stateful, which Ramp credits with letting "users clarify intent, collaborate in threads, and reason through complex problems," and answers come back with in-thread CSV previews so the asker can sanity-check the numbers without leaving Slack. On scope, the post is explicit: "Ramp Research does not have access to any personally identifiable information (PII)."

Two boundaries, stated so this section is not read as more than it is. The post never mentions Glass, Dojo, or Sensei -- it describes a standalone Slack agent, and the connection to Glass comes from the Glass write-up naming Ramp Research as one of the tools an SSO sign-in wires up, not from Ramp claiming continuity. And the post dates its launch only as "early August"; the year here is inferred from its publication date of September 18, 2025, not asserted by Ramp. What that ordering does support is a useful reading of the whole playbook: Ramp proved the harness argument on one workflow first, then generalized it into Glass.

The Skills Marketplace Model

The second piece is a marketplace called Dojo. Dojo holds 350+ skills, each a markdown file describing how to do one task well. In Ramp's words they are "Git-backed, versioned, and reviewed like code," and they are discoverable inside Glass. A recommender called Sensei surfaces relevant skills to each employee -- it "looks at which tools you've connected, what role you're in, and what you've been working on, and recommends the skills most likely to be useful to you" -- so a new sales rep does not have to know Dojo exists to get value from it.

Ramp's framing for why this matters: "One person's breakthrough should become everyone's baseline." Without a marketplace, every employee invents their own version of the same prompt, makes the same mistakes the previous person already solved, and the company pays the discovery cost a hundred times. With a marketplace, the cost is paid once and amortized across everyone who installs the skill.

The structural choice is the giveaway. Skills are markdown, not closed prompts buried in a vendor UI. They live in Git so they can be reviewed like code, rolled back if a regression lands, and forked when a team needs a variant. (For a complementary pattern, see how Cursor rules can be auto-generated from chat history -- skills emerge from real usage rather than being designed in a vacuum.) The marketplace is not a content library. It is a code repository with a discovery layer on top.

The cultural payoff is the second-order effect Ramp names: "We don't believe in lowering the ceiling. We believe in raising the floor." A skills marketplace does not cap what your strongest users can do. It pulls the median user up to where the strong users already are.

Persistent Memory From Existing Systems

The third piece is persistent memory that the agent does not have to build from a cold start every session. Glass auto-builds context from where employees already do work, carrying "references to relevant Slack channels, Notion documents, Linear tickets, and more." Ramp keeps it current by "mining users' previous sessions and connected tools like Slack, Notion, and Calendar for updates" -- a synthesis and cleanup pipeline that runs every 24 hours.

The contrast with cold-start AI is stark. A vanilla chat session starts knowing nothing about your team, your projects, or your last quarter. You spend the first 10 minutes of every conversation pasting context. Persistent memory eliminates that loop. The agent walks in already knowing what your week looks like.

The architectural decision worth noting is the source of the memory. Ramp did not ask employees to maintain a separate "memory file." The memory is derived from systems employees use anyway, so the agent reads from systems of record directly. That is the only design that survives contact with reality. A separate memory store decays the moment people stop updating it.

Scheduled Automations and Headless Mode

The fourth piece is the move from chat to scheduled automation. In Ramp's words, "You can schedule automations that run daily, weekly, or on custom cron, and post results directly to Slack." The example Ramp shared: "A finance team lead pulls yesterday's spend anomalies every morning at 8am and posts a summary." The report lands in the channel before the team starts work. This is the enterprise version of the always-on pattern also showing up in Perplexity Personal Computer and other computer-use agents.

This is the headless mode. You set up the task once, walk away, and the agent does the work overnight. If a step needs human approval -- a permission scope to widen, an irreversible action to confirm -- the request comes to your phone. The interaction model shifts from "I sit at the keyboard while the agent works" to "the agent works and I check in when asked."

The leverage compounds. Headless mode also unlocks workflows humans never do because they are too tedious to remember: weekly competitive scans, daily dashboard checks, end-of-day pipeline summaries. For a public-product analog, Claude Code Routines lets individual developers schedule the same kind of recurring work without building a Glass-grade harness in-house. The pattern is the same: cron the agent, ship the results to where the team already looks.

Skill Examples by Role (Sales, Support, Finance, Engineering)

The skills in Dojo are not abstract templates. They are role-specific workflows authored by the people who do the work. A representative cross-section of what the marketplace looks like in practice:

Sales. A Gong call analysis skill takes a recorded call, extracts objections and competitive mentions, and produces a structured summary. A competitive battlecard generator pulls public information about a named competitor, formats it against Ramp's positioning, and hands the rep a one-page document before the next call.

Support. A Zendesk investigation skill takes a ticket, auto-pulls the customer's full ticket history, checks account health signals, and surfaces the relevant context the support agent needs without manual digging. The agent does the lookup; the human does the judgment.

Finance. The spend-anomaly skill mentioned earlier runs as a scheduled job, but variants of it are also available on demand for ad hoc investigation. A finance lead can ask "show me anomalous spend in the last week" and get the same analysis the cron job produces, scoped to whatever window matters.

Engineering. PR-review workflows that read a diff, run static checks, and surface likely issues for the human reviewer. Repo-specific debugging skills that load the right context for a given service before the engineer starts asking questions. For teams in the Google ecosystem, the same skill-marketplace idea maps to Gemini tool definitions, Chrome Skills, and CLI workflows. A different orchestration surface for some of this work is Warp's agentic development environment, which gives engineers a terminal-native version of the multi-pane workspace pattern.

The unifying property: every skill encodes a real workflow that one employee figured out, then turned into a reusable artifact for the rest of the team. The marketplace is an institutional memory of how to do things well.

The Six-Step Playbook for Your Company

Most of what Ramp built is replicable without their engineering org. The playbook in six steps:

1. Audit which configuration steps stop non-technical employees from using AI. Sit with three employees from non-engineering roles and watch them try to use the AI tool of your choice. Every time they hit a permission wall, an API key prompt, or a "connect to your account" detour, write it down. That list is your harness backlog.

2. Centralize tool integrations behind SSO so connection is one click, not a setup quest. A platform team owns the integrations. Individual employees never paste tokens. New tools onboard once, audit once, and become available to everyone. The bar to first useful action should be a sign-in, not a guide.

3. Build an internal skills/prompts marketplace, code-reviewed and versioned. Skills live in Git as markdown, get reviewed like code, and surface in the workspace through a discovery layer. Sensei-style recommendations are nice but optional; the non-negotiable is that the marketplace exists and the review loop closes.

4. Wire persistent memory from your existing systems (chat, docs, tickets) so the agent does not start cold each session. Read from systems of record. Do not ask employees to maintain a separate memory store. A daily synthesis pipeline keeps context fresh without a custom event bus.

5. Bias to learn-by-doing: ship installable wins instead of training sessions. This is the one Ramp calls out as its biggest lesson, verbatim: "The single most important thing we learned building Glass: the people who got the most value weren't the ones who attended our training sessions. They were the ones who installed a skill on day one and immediately got a result." Optimize for the moment a new user runs their first useful skill, not the moment they finish onboarding documentation.

6. Build vs buy: own it if AI productivity is strategic. Ramp's stated case is that "Internal productivity is a moat," and that ownership buys a feedback loop you cannot rent: "When you own the tool, you see exactly where people get stuck. You can ship fixes the same day someone reports a problem." Off-the-shelf tools cover the easy 80 percent. The remaining 20 percent is where your competitive edge lives, and the build cost pays back in iteration speed.

The playbook does not require a Ramp-sized engineering team. The smallest viable version is one engineer running an SSO-backed proxy in front of an off-the-shelf agent, a Git repo of skills, and a daily cron that posts a report to Slack. Start there. Add Sensei later.

Where the Market Is Going (Anthropic Cowork, Notion Custom Agents)

Ramp's pattern is not isolated. The same architecture is showing up in shipping products from major vendors over the last few quarters.

Anthropic Cowork gives individual users a desktop agent with file access, scheduled tasks, and an open-source library of role-specific plugins, and Anthropic has since been extending it to mobile and web in beta. Organizing that library by role is the same instinct Glass applies to Dojo. The convergence is not coincidence. Roles are the natural unit of organization once you accept that AI productivity is workflow-shaped, not chat-shaped.

Notion 3.3 Custom Agents, released February 24, 2026, lets teams define agents that run on a trigger or a schedule, reach other systems through MCP, and operate inside the same workspace where their docs and databases live. The bet is that a Notion-shaped agent works because it is already inside the system of record -- the same persistent-memory thesis Glass operationalizes. Notion's own launch page quotes an ops lead at Ramp on the result.

Across all three -- Glass, Cowork, Notion -- the pattern is identical: SSO-connected tools, role-specific skills or plugins, persistent context from existing systems, scheduled jobs that run without a human at the keyboard. The same trend is visible in the consumer and developer market: Perplexity, Codex, and Claude computer-use agents are all trying to turn local computers into supervised workspaces. The category is consolidating around this shape because it is what works.

The implication for any team picking a 2026 strategy: the question is no longer about model choice. The question is what your harness looks like. The model is a swappable component. The harness is the moat.

Sources

• Sebastien Goddijn, Shane Buchan, Cameron Leavenworth, Calvin Kipperman, Jay Sobel, and Caroline Horn, "We Built Every Employee at Ramp Their Own AI Coworker", April 9, 2026. Every Glass detail on this page -- the 99 percent figure and its ordering, Okta SSO, the naming of Ramp Research / Ramp Inspect / Ramp CLI, Dojo, Sensei, the 24-hour synthesis pipeline, the 8am spend-anomaly job, headless mode, and all Glass quotations -- comes from this write-up. Re-verified against it on August 10, 2026.
• Faiz Hilaly, Cesar Duran, and Jay Sobel, "Meet Ramp Research: Our Agentic Data Analyst", Ramp Builders blog, September 18, 2025. Sole source for everything in the Ramp Research section: the definition quote, the usage figures, the dbt / Looker / Snowflake metadata index, the inspect-branch-backtrack design, thread state, CSV previews, and the PII exclusion. Fetched and read on August 10, 2026.
• Notion, "Notion 3.3: Custom Agents", February 24, 2026.
• Ramp has published no product, pricing, or announcements page for Glass. Checked July 29, 2026. Ramp's engineering blog carries no post on Glass, Dojo, or Sensei either; its full feed was read on August 10, 2026 and Ramp Research is the only Glass-connected tool covered there.

Provenance note (August 10, 2026). This page previously drew every Glass-related claim from a single write-up. It now carries a second, Ramp-owned source for one of the tools Glass connects, and the two are kept separate above so a reader can see which claims rest on which document. The same revision corrected a one-word misquotation: Ramp's sentence reads "You can ship fixes the same day someone reports a problem," and this page had rendered it as "reported" in two places, including inside the FAQ. Verified character-exact against the source with two independently-framed reads.