Aug 1, 2024 · 49m · mad
AI at Ramp: Making Every Team Radically More Productive | Eric Glyman, CEO, Ramp
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast hosted by Matt Turck, Ramp CEO Eric Glyman discusses how the fintech unicorn leverages real-time data architecture, generative AI, and tight cross-functional teams to automate corporate finance and drive rapid organizational velocity.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 15.6% of the talking time here. How this is scored →
Matt as informed peer 2.7 Guest teaching 2.1 Guest disagreement 0.3 Matt pushing back 0.4
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:48 Mocking 2023 chatbot trends
The guest forcefully rejects the conventional 2023 hype around conversational AI, stating he has never met a single person who wished they could chat with their bank account.
Hardest push from Matt ▶ 19:51 Challenging underwriting ML valueThe host explicitly challenges the guest's thesis on machine learning in underwriting, noting that traditional credit metrics like FICO scores are already effective and asking whether ML actually provides real juice.
Biggest teaching moment ▶ 20:24 Reframing corporate vs consumer credit lossesThe guest corrects the host's skeptical framing by breaking down the structural difference between consumer credit (5-10% loss rates) and corporate credit (0.1% loss rates), explaining how ML safely expands approval rates.
Matt holds his own ▶ 16:13 Demonstrating data stack fluencyThe host demonstrates deep domain familiarity by citing specific backend data infrastructure tools including Snowflake, dbt, and Metaflow.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Ramp Overview and 30-Second Elevator Pitch | 1 | 1 | 0 | 0 | The host opens with standard podcast setup and asks for Ramp's 30-second elevator pitch. The conversation is friendly and collaborative, with the guest sharing the company's pitch and origin story behind counting company age in days. | |
| Origin Story: Paribus to Founding Ramp | 1 | 2 | 0 | 0 | The host asks about the guest's previous company Paribus and how it led to founding Ramp. The guest explains how analyzing card reward misalignments showed that users prefer saving money over getting points. | |
| Ramp as an Automation and Workflow Platform | 3 | 1 | 0 | 0 | The host connects Paribus's pattern recognition DNA to Ramp's identity as a workflow and automation platform. The guest details how Ramp automates expense management directly into card issuing. | |
| Data Architecture, Integrations, and Unfair Advantages | 2 | 2 | 0 | 0 | The host asks about the underlying data integrations and plumbing required to run Ramp. The guest details the hundreds of integrations across financial institutions, HRIS, single sign-on, and ERP systems. | |
| Data Moats, Network Effects, and Price Benchmarking | 4 | 1 | 0 | 0 | The host posits that Ramp possesses a strong data moat and flywheel where each new product leverages existing transactional data. The guest agrees and cites vendor price benchmarking and bill pay network effects as key examples. | |
| Safeguarding Customer Data Privacy and Building Trust | 5 | 1 | 0 | 1 | The host raises the inevitable data privacy concerns before diving into technical data infrastructure questions. The host names specific backend tools such as Snowflake, dbt, and Metaflow to prompt guest details. | |
| AI Spectrum: Internal Operations to Credit Risk | 4 | 2 | 0 | 1 | The host frames the evolution from classical machine learning to generative AI before interrupting briefly to press on credit underwriting. The guest categorizes Ramp's historical ML focus across fraud, underwriting, and receipts. | |
| Machine Learning in Corporate Credit Underwriting | 5 | 4 | 1 | 4 | The host directly challenges whether machine learning yields meaningful gains in underwriting over traditional credit scores like FICO. The guest pushes back with industry data, contrasting consumer subprime loss rates with corporate credit loss realities. | |
| Boosting Internal Team Productivity with Generative AI | 2 | 3 | 0 | 0 | The guest outlines how Ramp uses internal generative AI tools to make sales development representatives three to four times more productive. He emphasizes human workflow augmentation rather than total job replacement. | |
| In-House AI Development and Cross-Functional Pods | 3 | 2 | 1 | 0 | The host asks whether internal AI tooling is built in-house and how humans interact with these systems. The guest explains how cross-functional pods combine growth engineering with domain experts, criticizing standard corporate siloing. | |
| Centralized Data Architecture and Applied AI Integration | 3 | 3 | 0 | 0 | The host asks how Ramp structures its data and AI teams organizationally. The guest explains their horizontal Applied AI strike team that embeds into operational units to automate manual tasks and slash underwriting turnaround times. | |
| Ramp Intelligence: AI-Powered Financial Capabilities | 2 | 3 | 1 | 0 | The host introduces Ramp Intelligence, and the guest dismisses early 2023 trends of adding useless chatbots to bank accounts. He showcases practical applications like multimodal agentic UI navigation and query engines for sales calls. | |
| Sustaining High Product Velocity as Companies Scale | 2 | 2 | 1 | 0 | The host asks how Ramp maintains rapid product shipping velocity as the team expands. The guest explains that keeping single-threaded engineering teams under 15 people avoids corporate consensus traps. | |
| Reaching Product-Market Fit and Go-to-Market Strategy | 2 | 3 | 0 | 0 | The host asks about reaching product-market fit and early go-to-market execution. The guest reveals that Ramp lost early deals by refusing to integrate with Concur, betting successfully on a unified card and expense platform. | |
| Building and Scaling a Technology Giant in NYC | 2 | 1 | 0 | 0 | The host invites reflections on building a tech company in NYC compared to Silicon Valley. The guest praises NYC's talent density, lower mercenary turnover, and proximity to finance and design. |
Statements from this episode (26)
Glyman: Ramp customers save an average of 5% on expenses
“And the upshot is the average company using, using RAMP saves about five percent. On their expenses.”
Glyman: Approximately 25,000 businesses use Ramp
“About 25,000 businesses use it from, you know, early-stage startups to publicly-traded companies like Shopify to Boys and Girls Club of America as a non-profit to, you know, everything in between.”
Paribus reached about 1 million customers within one year of launch
“Within a year we had about a million customers”
Glyman: Ramp is a productivity and data company disguised as a card
“And so it's I think really a productivity or a data driven company kind of in disguise of it's your friendly neighborhood credit card.”
Glyman: Ramp has hundreds of bi-directional and thousands of one-way integrations
“I mean, I think in terms of You know, bi-directional, fairly deep integrations. I mean, we're talking you know, low hundreds, probably when you start expanding into, you know, one way, getting, you know, low thousands.”
Glyman: Ramp has unfair advantages over expense software by powering transactions
“We have a lot of unfair advantages over pure play expense management software because we're powering the card transaction.”
Glyman: Ramp is the only platform indicating whether companies overpay vendors
“Today ramp is the only place where you cannot just spend money on a vendor, but we can tell you if we're, if you're paying too much.”
Glyman: Ramp operates a horizontal Applied AI team across all departments
“We do have an applied AI team, which, you know, is very horizontal. There's almost no part of, you know, how we operate that's off limits”
Glyman: Ramp approves nearly 100% of prospective clients
“For us you know, if a sales team is talking to you know, a prospect, it's almost a hundred percent of companies. That we're able to approve.”
Glyman: Ramp's corporate credit losses are lower than Amex's historical rates
“The corporate, if you look at Amex's disclosures you know, historically, they lose .1% you know, to .2% per year, and we're below that I would argue I think below.”
Glyman: Ramp SDRs book 3 to 4 times more meetings than competitors
“The average SDR sales development representative at RAMP we believe books three to four times as many meetings as their next closest competitor.”
Glyman: Every function at Ramp uses AI in its work
“I don't think there's a function at RAMP that doesn't use AI in some part of their job.”
Glyman: AI human augmentation is currently more valuable than full automation
“People tend to jump right to full automation when really what is incredibly valuable in the short run is augmentation of people on, on aspects of what they do.”
Glyman: Ramp built its internal AI tools predominantly in-house
“Predominantly homegrown. I mean, we've been doing this for almost as long.”
Glyman: All Ramp employees have access to company financial records
“At RAMP is, like, we report, you know, everyone has access to financials.”
Glyman: Ramp AI sprint cut manual underwriting time from days to hours
“Underwriting times for manual cases that we couldn't use machine learning to write, it could take an underwriter two days to do that. They sat side by side with the underwriter, and I think within a couple, I think like a two week sprint that the average under…”
Glyman: Customers do not want to chat with their bank accounts
“We'd never met a single person who's like, I just wish I could chat with my bank account.”
Glyman: Half of non-automatable Ramp support requests stem from user confusion
“About half of support requests that, you know, aren't fully automatable now fall under the question of user confusion.”
Glyman: Ramp's experimental AI agent achieves 60% to 90% task accuracy
“I think right now, and the agent probably works for you know, depending on the nature of the task 60 to 90% of the task of when it's controlled and inside of the RAMP experience. Obviously, when you're dealing with financial products, it's not high enough to g…”
Ramp has analyzed over 100,000 Gong sales calls for AI training
“We've now recorded for training you know, help salespeople improve, give feedback from managers, you know, over a 100,000 gong calls”
Glyman: Expanding small product teams into 30-person consensus groups kills velocity
“I think one of the worst things you can do to people building products is go from, hey, it's you and four people trying to decide what a well-run product is, a great experience could look like, ah, to go do a good job, and you say, great, now get 30 other peop…”
Glyman: Ramp's core spend management team is 13 to 14 people
“The core spend management team is, I think now largest has ever been. I think it's 13 people, 14 people, something like that.”
Glyman: Ramp teams cannot ship new features if existing product SLAs break
“You're getting the tickets, and if SLAs are broken, you can't shift new product. You need to fix the existing one”