Ramp on X

12 people are delighted outright. One says a first week of expenses lived up to hype they were allergic to; another calls it the most delightful business product of their career and says it "makes that thing that used to suck, not suck." A third moved an org onto it and finds it simple for everyone who spends.

8 admire the execution and craft rather than the product alone: scaling from 250 to 1,600+ cited as a lesson in talent density, props to the marketing team, a brand recognizable from a sliver of a testimonial, and an engineering choice — the sandbox as a tool the agent calls, with session state in the database — picked out for praise.

7 praise particular features: the Chrome extension and texts, Gmail receipts, a different card number per transaction, a Starlink filter added to travel, onchain treasury rails. 2 credit the published spend data itself.

5 are critical, over an unprofessional mass email, an incomplete AI token spend feature, data published a week or two late, and one who finds the attention on Ramp "quite odd."

6 people describe AI spend as hard to see and control. One lists the stack — Cursor with and without OpenRouter, Codex, ChatGPT and Claude without enterprise plans — and says keeping one global view is so hard they are "really pissed." Ramp's AI token spend feature doesn't capture non-Anthropic and non-OpenAI usage through OpenRouter, can't properly assign usage to team members, and can't connect to ChatGPT or Claude without enterprise plans on both. Replies pick at the edges: tracking those bills is "a nightmare," and a seat is a flat fee no vendor reports per use, so enterprise still wouldn't give spend per person. One asks why basic usage tracking is still such a pain.

2 were unhappy with a mass email that read as unprofessional; one reply says Ramp trained its "mass email agent on too much spam." 2 find budgets confusing — which budget a physical card charges to, and the mental model when new; a Ramp account agrees complexity crept in as capabilities were added and says simplification is underway. 2 posts from one person complain published data lags a week or two behind other trackers.

3 people weigh Ramp against Mercury, with Airwallex and Slash named alongside. The comparison runs on business models and numbers rather than features: one asks whether a table should be updated with a $1B run rate, another guesses at where the revenue comes from (subscriptions, interchange, possibly a setup or annual fee, a cut on yield), and notes that in fintech fee models multiply and "the pricing page doesn't always cover every case." One wants year-on-year growth rates on the Y axis and a valuation-scaled dot, guessing Ramp may be growing fastest and super-linearly, which would point to cross-selling.

The other two are single voices. 1 builder says they started building their own rails on Column and are skipping Ramp and Brex, keeping the interchange and pushing people toward stablecoin payments; they only learned Column had an API after the prototype. 1 reader says Ramp's new Connectors look a lot like executorsh and wonders how the implementations differ.

3 people question the data behind Ramp's published spend numbers. The doubts are about coverage and bias: how much enterprise AI spend happens outside Ramp's customer base, whether spend billed directly through AWS or GCP is caught, and whether there's "insane selection bias," with one asking honestly whether startups are normalized against other companies.

3 ask about how Ramp sells and hires, prompted by a founder's account of its go-to-market. One wants to know what replaced forecasting calls in the weekly routine; another asks how quota capacity was tied to hiring pace as the company scaled, noting that overpaying reps works until ramp time stretches and the comp plan outruns the pipeline; a third does the math on paying twice for more meetings and wants to know whether meeting quality held up.

2 ask where the revenue comes from given there are no wire fees, with replies pointing to subscriptions, interchange, possible setup or annual fees and a cut on yield. 2 ask what others used before Ramp, wanting to know what it replaced.

Every wish here is a single voice, so none carries more weight than another.

On AI spend, one person wants to see which workflow burned the budget, not just the total per tool: "the total per tool doesn't help me much. I want to see which workflow burned it so I know what to kill."

On cards, someone wants creating virtual cards to be "as easy as one tap," saying Ramp's and Mercury's APIs don't support it yet, and asking to be connected to people at both companies.

On the roadmap, a reader replying to Ramp staff asks simply, "Bookkeeping next?"

On travel, one person suggests predicting wifi quality from airline, tail number and refund data on wifi charges, and wants a "filter by starlink" option until it exists. Two weeks later a user posted that they had added a Starlink filter to Ramp travel, so that wish was met in the meantime.