GenAI PMRamp
A fintech company described here as using multiple AI agents across product discovery, UX fixes, reviews, and testing. This is relevant to PMs as an example of AI-enabled product development at scale.
Key Highlights
- Ramp is described as a fintech company using AI agents across product discovery, coding, UX fixes, reviews, and testing.
- Newsletter mentions say Ramp shipped 500+ features with 25 PMs and made AI usage a company-wide expectation.
- Ramp’s Inspect agent was reported to return deployed previews in under 5 seconds and to have handled 1 million sessions.
- A key lesson from Ramp is that context files, memory, MCPs, and workflows were necessary before AI adoption created real value.
- Ramp’s L0–L3 AI proficiency ladder offers PMs a concrete model for scaling AI capability across teams.
Ramp
Overview
Ramp is a fintech company that has become notable in AI product circles for how aggressively it applies AI agents across the product development lifecycle. In newsletter coverage, Ramp is described as using Claude-powered coding workflows, custom agents like Inspect, and a company-wide proficiency model to help employees move from basic AI usage to building reusable AI systems.For AI Product Managers, Ramp matters less as a payments or finance brand and more as a case study in AI-enabled product development at scale. Mentions highlight a setup where PM prompts can become production-ready features, where non-engineers submit pull requests, and where AI adoption is treated as an organizational capability with workflows, context files, memory, and internal infrastructure—not just access to chatbots.
Key Developments
- 2026-03-05: Peter Yang highlighted Ramp as one of three AI-native companies, describing it as a company that drives performance by mandating Claude Code usage.
- 2026-03-06: Ramp was described as having shipped 500+ features in the prior year with just 25 PMs by requiring every employee—from engineering to finance—to onboard to and use Claude Code AI agents.
- 2026-03-07: Tyler Folkman noted that Ramp mandates AI agents for every role and tracks a four-level proficiency framework from L0 to L3, ranging from occasional ChatGPT use to codified, reusable AI skills.
- 2026-03-14: Peter Yang outlined Ramp’s four-stage AI proficiency ladder, from L0 “Disengaged” ChatGPT dabblers to L3 “Systems builders” creating team-wide AI infrastructure.
- 2026-03-15: Peter Yang reported that Ramp shipped 500+ features with 25 PMs using Claude Code’s three-phase workflow: problem framing with pushback questions, parallel research agents scanning sources like competitors, Gong calls, Zendesk tickets, and code, and iterative conversion into a concise spec.
- 2026-03-16: Ramp was described as using Claude-powered Cloud Code and the Inspect AI agent to turn PM prompts into production-ready front-end and back-end features with pull requests in under five minutes. Coverage also said 50% of Ramp’s code was built by AI, up from 30% in December, with a projection to reach 80% by March.
- 2026-05-18: Marc Baselga shared Sebastien Goddijn’s view that Ramp’s AI adoption only delivered real value after engineers built scaffolding such as context files, MCPs, memory, and workflows; otherwise non-technical users of Claude, ChatGPT, or Cursor pay a hidden setup tax.
- 2026-07-11: Harrison Chase said Ramp, Stripe, and Coinbase already offer Sierra-like tools, placing Ramp among companies building advanced internal AI experiences.
- 2026-09-18: Cognition recapped an experiment where Devin was given access to a Ramp card and told to make money; Devin ran outreach, built payment portals, tested a business plan, and reportedly made $75.
- 2026-09-27: Lenny’s Podcast recapped Geoff Charles’s overview of Ramp’s AI agents spanning product discovery through UX fixes. Inspect was described as running in under 5 seconds, returning a deployed preview, handling 1 million sessions, building 75% of Ramp’s pull requests, and submitting 1,000 pull requests from non-engineers in the prior month.
Relevance to AI PMs
1. A practical model for AI-native product workflows: Ramp shows how PM work can be restructured around AI, from problem framing and research synthesis to spec writing and code generation. For AI PMs, the tactical lesson is to design repeatable multi-step workflows rather than treating AI as a generic assistant.2. Proof that internal scaffolding matters more than raw model access: Ramp’s reported gains came after building context files, MCPs, memory, and workflows. PMs evaluating AI adoption should plan for operational infrastructure, permissions, context management, and reusable prompts or skills—not just tool licenses.
3. An example of capability-building as product strategy: Ramp’s L0–L3 proficiency ladder suggests that successful AI adoption can be managed like a product rollout. PMs can apply this by defining maturity levels, training paths, and measurable behaviors that move teams from ad hoc experimentation to reusable systems.
Related
- Geoff Charles: Ramp’s CPO and a central voice in explaining the company’s AI-agent operating model.
- Peter Yang: Frequently surfaced Ramp as a case study in AI-native product development and PM workflows.
- Claude Code / Claude: Core to Ramp’s reported coding and product-development workflows.
- Inspect / Inspect AI: Ramp’s custom AI agent for turning prompts into previews and pull requests quickly.
- ChatGPT and Cursor: Mentioned as tools that non-technical staff can use, though Ramp’s experience suggests they need organizational scaffolding to create durable value.
- Linear: Another AI-native company referenced alongside Ramp in discussions of new ways of working.
- Factory / Factory AI: Mentioned with Ramp as an example of packaging reusable AI skills into workflows.
- Stripe and Coinbase: Peer companies cited as already offering Sierra-like tools, placing Ramp in a broader set of advanced adopters.
- Devin / Cognition: Connected through the experiment where Devin used a Ramp card to attempt autonomous money-making.
- Review Buddy and Testo: Related to the broader ecosystem of AI agents involved in review and testing workflows discussed around Ramp.
Frequently asked questions
What is Ramp?
A fintech company described here as using multiple AI agents across product discovery, UX fixes, reviews, and testing. This is relevant to PMs as an example of AI-enabled product development at scale.
What should AI product managers know about Ramp?
Ramp is described as a fintech company using AI agents across product discovery, coding, UX fixes, reviews, and testing. Newsletter mentions say Ramp shipped 500+ features with 25 PMs and made AI usage a company-wide expectation. Ramp’s Inspect agent was reported to return deployed previews in under 5 seconds and to have handled 1 million sessions.
Explore this topic
Newsletter Mentions (11)
“Lennys Podcast recapped Geoff Charles’s overview of Ramp’s AI agents, which span product discovery through UX fixes; Inspect runs in under 5 seconds, returns a deployed preview, has handled 1 million sessions, built 75% of Ramp’s pull requests, and submitted 1,000 pull requests from non-engineers in the previous month.”
#2 ▶️ What product looks like when coding is solved | Geoff Charles (Ramp CPO) Lennys Podcast Lennys Podcast recapped Geoff Charles’s overview of Ramp’s AI agents, which span product discovery through UX fixes; Inspect runs in under 5 seconds, returns a deployed preview, has handled 1 million sessions, built 75% of Ramp’s pull requests, and submitted 1,000 pull requests from non-engineers in the previous month.
“Cognition recapped an experiment in which Devin was given access to a Ramp card and instructed to make money.”
GenAI PM Daily September 18, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 20 insights for PM Builders, ranked by relevance from Blogs, X, and YouTube. Cognition recapped an experiment in which Devin was given access to a Ramp card and instructed to make money. Devin conducted cold outreach, built payment portals, experimented with a business plan, and eventually made $75.
“Harrison Chase notes that Ramp, Stripe, and Coinbase already offer Sierra-like tools.”
#25 𝕏 Harrison Chase notes that Ramp, Stripe, and Coinbase already offer Sierra-like tools. He recommends OpenSWE, a fully open-source, model-agnostic repo with seamless LangSmith integration that they use internally for coding.
“#2 in Marc Baselga shares Sebastien Goddijn’s insight that Ramp’s AI adoption only drove real value after engineers built context files, MCPs, memory and workflows.”
#2 in Marc Baselga shares Sebastien Goddijn’s insight that Ramp’s AI adoption only drove real value after engineers built context files, MCPs, memory and workflows. Without this scaffolding, non-technical staff using Claude, ChatGPT or Cursor foot the hidden “setup tax.”
“Ramp uses Claude-powered Cloud Code and the Inspect AI agent to convert PM prompts into production-ready front-end and back-end features complete with pull requests in under five minutes.”
#7 in Peter Yang unveils a new episode with Ramp CPO Geoff where he breaks down an AI-native playbook—using Claude Code, custom AI agents for research, data & coding, plus an L0-L3 framework to get every employee shipping production code. #8 ▶️ Inside Ramp, the $32B Company Where AI Agents Run Everything | Geoff Charles Peter Yang Ramp uses Claude-powered Cloud Code and the Inspect AI agent to convert PM prompts into production-ready front-end and back-end features complete with pull requests in under five minutes. 50% of Ramp’s code is built by AI (up from 30% in December), with a projection to reach 80% by March.
“Ramp Ships 500+ Features Using Claude Code #1 𝕏 Peter Yang : Ramp shipped 500+ features last year with just 25 PMs using Claude Code’s 3-phase skill—phase 1 frames the problem with defendable pushback questions, phase 2 spins up 6–10 parallel agents to scan competitors, Gong calls, Zendesk tickets and code, and phase 3 conv...”
Today's top 12 insights for PM Builders, ranked by relevance from X, LinkedIn, and Blogs. Ramp Ships 500+ Features Using Claude Code #1 𝕏 Peter Yang : Ramp shipped 500+ features last year with just 25 PMs using Claude Code’s 3-phase skill—phase 1 frames the problem with defendable pushback questions, phase 2 spins up 6–10 parallel agents to scan competitors, Gong calls, Zendesk tickets and code, and phase 3 conv... #2 𝕏 Santiago processed PDFs with Claude Code by copying them into a folder and asking it to read them. The tool then auto-installed poppler and pdftoppm, enabling seamless opening and processing of the files. #5 in Dharmesh Shah says the new 1M-token context window for agentic coding isn’t just about handling more code—it frees him from context anxiety so he can steamroll through tasks without ever hitting the limit. #12 in Peter Yang shows how Ramp’s 25 PMs shipped 500+ features last year by using Claude Code’s three-phase workflow—framing the problem with targeted Q&A, launching 6–10 parallel research agents, and iteratively shaping a concise 2-minute spec.
“in Peter Yang unveils Ramp’s four-stage AI proficiency ladder—from L0 “Disengaged” ChatGPT dabblers to L3 “Systems builders” creating team-wide AI infrastructure—and shows how Ramp is methodically elevating every employee’s AI-native skills.”
in Peter Yang unveils Ramp’s four-stage AI proficiency ladder—from L0 “Disengaged” ChatGPT dabblers to L3 “Systems builders” creating team-wide AI infrastructure—and shows how Ramp is methodically elevating every employee’s AI-native skills.
“#24 in Tyler Folkman : Ramp shipped 500+ features last year with just 25 PMs by mandating AI agents for every role—using tools like Claude Code—and tracking a 4-level proficiency framework from L0 (occasional ChatGPT use) to L3 (codified, reusable AI skills).”
GenAI PM Daily March 07, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 25 insights for PM Builders, ranked by relevance from LinkedIn, YouTube, X, and Blogs. #24 in Tyler Folkman : Ramp shipped 500+ features last year with just 25 PMs by mandating AI agents for every role—using tools like Claude Code—and tracking a 4-level proficiency framework from L0 (occasional ChatGPT use) to L3 (codified, reusable AI skills). #25 in Saharsh Agrawal built a weekend-in-a-peak custom CRM with Claude—complete with contact records, pipeline stages, and deal tracking—only to learn in two weeks that without a dedicated owner it constantly broke and onboarding new sales or marketing hires (all used to HubSpot/Sa...
“Ramp shipped 500+ features last year with just 25 PMs by mandating every employee—from engineering to finance—onboard and use Claude Code AI agents.”
GenAI PM Daily March 06, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 25 insights for PM Builders, ranked by relevance from Blogs, X LinkedIn, and YouTube. OpenAI Introduces GPT-5.4 Model #1 📝 OpenAI News Introducing GPT-5.4 - Announcement of GPT-5.4 as a new product release, highlighting improvements and new capabilities over prior models. The post introduces features and potential applications of GPT-5.4. Also covered by: @There's An AI For That , @Kevin Weil 🇺🇸 #15 in Tyler Folkman : Ramp shipped 500+ features last year with just 25 PMs by mandating every employee—from engineering to finance—onboard and use Claude Code AI agents.
“Peter Yang unveils how three AI-native companies—Linear assigns tasks to AI “team members” via natural language, Ramp drives performance by mandating Claude Code usage, and Factory AI packages product management, UI, and data analysis into reusable AI skills—offering concrete...”
#10 𝕏 Peter Yang unveils how three AI-native companies—Linear assigns tasks to AI “team members” via natural language, Ramp drives performance by mandating Claude Code usage, and Factory AI packages product management, UI, and data analysis into reusable AI skills—offering concrete...
Related
Anthropic’s coding agent/environment used here to control physical tools in a long-running home experiment.
Anthropic’s AI assistant, referenced here through new dashboard, motion, docs, slides, and design capabilities.
An AI coding tool shown here supporting follow-up questions and a /visualize command that generates a new chart in-chat.
A product-focused creator who demonstrates AI builds and critiques product UX/packaging. In this issue he showcases Tabi and discusses simplifying product surfaces.
An AI company associated with Devin, mentioned for allowing users to connect personal ChatGPT plans to the agent. Relevant to AI PMs thinking about consumption-based product design.
OpenAI’s chat product, referenced here as the user-owned subscription plan connected to Devin. For PMs, this highlights quota-linked integration patterns.
An AI coding agent from Cognition that can be connected to a personal ChatGPT subscription quota. Relevant to PMs exploring agent productivity and cost-sharing models.
A payments company referenced for internal agent use and operational safeguards around agentic workflows.
A project management and issue-tracking tool used in agentic software workflows. In this newsletter it is used by Warp’s Wilson to open issues from Slack requests.
A company mentioned as already offering Sierra-like tools. It matters to PMs as another example of a large platform using AI assistant capabilities at scale.
An AI-native startup mentioned as delegating tasks to AI agents across multiple functions. Relevant to PMs as an example of an AI-first operating model.
Operator or commentator discussing enterprise adoption of AI agents. He highlights Ramp's use of Claude Code and a small PM team shipping many features.
Stay updated on Ramp
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free