UnifyOutbound teardownsRamp

Ramp built an AI SDR that drove about 30% of pipeline for years. Then it shut it down.

Company
Ramp sells corporate cards and spend management software to businesses.
Visit Ramp ↗
Sales team
5 SDRs in 2021, about 500 people in sales by 2025
Headcount
Team size and hiring trends
Ramp headcount ↗
CEO
Eric Glyman
CEO profile ↗

TL;DR

By late 2025 competitors had the same automation, buyers were tired of cold outbound, and with about 500 salespeople selling several products, the sale had grown more complex. Ramp shut its AI SDR down and handed the data behind it to its reps: ranked accounts, the reason why, and a draft to review.

The lesson

Automated outbound stops paying when competitors run the same tools, buyers tire of it, and your sale gets more complex. Then the advantage moves to what competitors can’t buy, like your own data and your reps.

For your team

You can run the same kind of system without building it: ranked accounts, the signal behind each one, and drafts reps approve.

What Ramp did

The original solution

“We had a team of five SDRs that needed to reach out to millions of people,” co-founder Gene Lee wrote later of 2021. Nothing on the market could cover that, so Ramp built its own AI SDR: software that does a sales development rep’s prospecting and first outreach. It called the system OATs, the Outbound Automation Team.1

Ramp’s data stack brought product, CRM, support, marketing and sales data together in one warehouse. OATs used that data to pick prospects and personalize outbound at scale, with an early version of ChatGPT drafting the emails and sorting the replies. By 2023 a growth engineering team owned the channel and ran it like a product.2,3,4

  1. Market data, in one warehouse
  2. Target prospects identifiedScoring criteria not described
  3. Email drafted by an early ChatGPT
  4. Sent automatically, at scale
  5. Replies classified and ranked for reps
Source: dbt Labs case study, 2022; Gene Lee, Ramp co-founder, Dec 2, 2025; Ramp Builders Blog, Nov 2023What it shows: Reconstructed from public sources. OATs ran on Ramp’s own data, with AI drafting and reply triage. Ramp hasn’t described how it chose accounts. Companies, emails and replies are example data.

A 2022 case study put OATs at 25% of pipeline, and Lee later said it accounted for roughly 30% “for years.”1,2

The problem changed

In the years before the shutdown, the returns flattened. Lee gave three reasons. The tools had become common: “there are dozens of automated outbound solutions, all operating on similar data which our competitors also have access to.” Buyers had seen too much of it: “the world is experiencing cold-outbound fatigue.” And Ramp had changed. It now had around 500 people in sales and several products to sell, so the sale had grown more complex.1

2021

  • Sales teamFive SDRs
  • Automated outboundRare. ChatGPT hadn’t launched
  • DataRamp’s own view of the market
  • BuyersNot yet saturated
  • The saleOne market to cover

2025

  • Sales teamAbout 500 people in sales
  • Automated outboundDozens of vendors
  • DataSimilar data available to competitors
  • BuyersCold-outbound fatigue
  • The saleSeveral flagship products
Source: Gene Lee, Ramp co-founder, Dec 2, 2025What it shows: the environment around OATs changed, in Ramp’s own account.

The decision

December 2, 2025

Ramp shut down a system that had driven about 30% of pipeline for years and pointed its data at its sales reps.

The shift had been under way for two years. Lee called the data the company’s long-term edge, and Ramp stopped using it to send email on its own.1

The redesign

Ramp Revenue puts the same data in front of each rep every morning.

Ramp built an internal tool called Ramp Revenue. A rep opens it in the morning and sees, in the words of growth product lead Keyan Sarrafzadeh, “all the prospects that you should be focusing on, an explanation of why.” Each account shows its signals, such as visits to the pricing page or a new hire, and the tool writes a first draft. In the screenshots Ramp shared, the rep reviews each draft before the sequence starts.5,6

What the rep sees now

Home · A ranked list of who to focus on today
Ramp Revenue, Home screen, full size
Ramp Revenue, posted publicly by Parth Gujare (Ramp), Feb 2, 2026, with names masked by Ramp. Screenshots used here to illustrate the sales rep workflow described in the post.

The learning system

One central team can’t anticipate every workflow, so reps now build their own AI tools on the same data, through what Ramp calls its Growth MCP, a permissioned layer that connects AI tools to Ramp’s CRM and outreach data. There were 79 by mid-2026. Every time one runs, it has to log why. Ramp groups those reasons to see what reps need, and builds the most useful ideas back into Ramp Revenue. Agents now do much of the routine work, from inbox triage to CRM updates, and people still approve what goes out.6,7

The result

Ramp says more than 80% of sales workflows and more than 90% of GTM teams now use the new system. Both figures measure how widely the system is used, and neither measures pipeline.7,8

  1. Feb 202680%+of sales workflows run on Ramp Revenue
  2. Jun 202679rep-built tools, with 300 to 400 reps using the layer
  3. Jul 202690%+of GTM teams use the system
Source: Ramp, Feb 4, 2026; Keyan Sarrafzadeh, Ramp, Jun 2026; Parth Gujare, Ramp, Jul 24, 2026What it shows: adoption inside Ramp over time. The 300 to 400 figure is from the event recap of Keyan Sarrafzadeh’s talk.

What it took

Source: Gene Lee, Ramp co-founder, Dec 2, 2025; Ramp Builders Blog, Apr 2025; Ramp, Feb 4, 2026; Keyan Sarrafzadeh, Ramp, Jun 2026What it shows: Ramp’s build history, from Ramp’s own posts and talks. The start of the Ramp Revenue shift is approximate, since Lee says only that it took “the last two years.”

Ramp built two systems, and the first one became something anyone could buy.

What Ramp spent years building is now, in Lee’s words, one of “dozens of automated outbound solutions, all operating on similar data.”1

Rank your accounts in Unify.

See which accounts your reps should work today, and why.

What to borrow

  1. Track what your automation adds to pipeline over time. Pipeline contribution, conversion and reply quality show when it starts to flatten.
  2. When automated outbound stops paying, put your own data in front of reps. Ramp kept the warehouse that fed OATs and gave it to reps as ranked accounts with a reason.
  3. Let software draft, and have the rep approve before anything sends. In Ramp Revenue each draft waits for the rep’s review.

How to build a version of this in Unify

Start where Ramp ended up

You can run a version of this without building it.

Ramp built a warehouse to get one view of its market. In Unify, your CRM, warehouse and product usage data come together with enrichment from 40+ data sources, so you start with the data layer Ramp had to build. A version of Ramp’s system takes three steps:

  1. Define the accounts. Build an Audience from your CRM and Unify’s data: the companies that fit, weighted by what they’re worth.
  2. Add the why-now. Turn on Signals like website visits, new hires and champion job changes, so every account carries a reason.
  3. Route the work. A Play sends lower-priority accounts into automated sequences and puts the rest in each rep’s Book of Business, with research and a draft to review.

One account, from signal to next step

Account
Halden SystemsIn your Audience: fits on size, industry and the tools it already uses.
Signal
Three pricing page visits this week, and a new VP of FinanceFrom Signals: website visits and new hires.
Why it matters
New finance leaders often review spend tools early, and someone there is comparing prices nowAI research writes the reason on the account, so the rep doesn’t start from zero.
Next step
Call task for the account owner, due today. An email to the new VP waits for the rep’s approval.From the Play: owner assigned, contacts found, Sequence started.
Example data: the company and people are fictional. Features are Unify’s as documented, October 2026.
outbound motion
Do it in
One view of the marketA warehouse joined to the CRM
Data and CRM syncCombines your CRM, warehouse and product usage data with enrichment from 40+ sources
Why-now signalsOn every account
SignalsWebsite visits, new hires, champion job changes
Ranked accountsWith the reason why
Book of BusinessEach rep’s accounts by fit and timing, with the latest signal
Drafts the rep reviewsBefore anything sends
SequencesManual steps the rep approves before sending
Agents for the busyworkResearch and follow-up
AI research and AgentsOn every account, without a tools team
Automation where it paysFor the lower tier
PlaysLower-priority accounts go into automated sequences
Left column: Ramp’s system as described in the sources above. Right column: Unify features as documented, October 2026. Unify offers comparable capability and doesn’t reproduce Ramp’s own implementation.

Each rep’s Book of Business shows their accounts, ranked by fit and timing, with the latest signal, where each conversation stands and the last sales activity.

Source: Unify product, animated recreation by Unify’s design team, with fictional companiesWhat it shows: each rep’s accounts ranked by fit and timing, with the latest signal, status and last sales activity.

Run a version of Ramp’s system in Unify.

Ranked accounts, the reason behind each one, and drafts your reps approve.

Frequently asked questions

What was Ramp’s AI SDR?

Ramp’s AI SDR was OATs, the Outbound Automation Team, which Ramp built in 2021 when five SDRs had to reach millions of businesses. It used Ramp’s own market data, drafted emails with an early version of ChatGPT, sent them automatically and ranked replies so reps answered the most interested companies first.

Why did Ramp shut down its AI SDR?

Automated outbound became something anyone could buy, running on similar data. Buyers were tired of cold automated email, and Ramp’s sale had grown more complex, with around 500 people in sales and several products. Gene Lee said the returns had plateaued.

What replaced OATs?

Ramp Revenue, an internal tool that gives each sales rep ranked accounts, the reason behind each one and a draft to review. On top of it, Growth MCP lets reps build their own AI tools on Ramp’s data; there were 79 by mid-2026.

Should my team build an AI SDR in-house?

Ramp could, because it had a growth engineering team, its own market data and about 500 people in sales to spread the cost across. It still built two systems, and the first became something anyone could buy.

How do I run a version of this in Unify?

Build an Audience of accounts that fit, add Signals such as website visits, new hires and champion job changes, then use a Play to send lower-priority accounts into automated sequences and put the rest in each rep’s Book of Business with research and a draft to review.