← Back to the run

Independent work sample by Jerry Nettles. Not commissioned by, affiliated with, or endorsed by Ramp. Built from public information only; every material claim carries a label (Publicly verified · Reasoned hypothesis · Internal input required · Illustrative output). Figures marked illustrative are placeholders, not Ramp's numbers.

Company Proof — ICP & Buying-Committee Model + Positioning & Message System: Ramp / AI Router

Phase: 3 — Company-Specific Proof. Subject: Jerry's PMM operating system, running its Module 02 mechanism (01_OS_Architecture/02_icp-and-positioning.md) against a live test case. Demonstration company: Ramp, specifically the AI Router launch (Router.com, 2026-08-20) and the broader AI-native finance-ops narrative it sits inside. Approved by the Founder 2026-09-04, canon/DECISIONS.md D-024.

Company facts used here are inputs, not the deliverable. Per D-024's Jerry-add test, every section below is split into OBSERVED (what's merely true about Ramp, sourced) and OS CONTRIBUTES (what Jerry's system does with that input — the decision, the confidence-rated map, the missing-input it names, the validation it runs, the messaging logic). A reader should be able to tell, without being told, what came from Ramp's own public materials versus what this operating system added. Claim-status labels follow CHARTER.md's fixed vocabulary: Publicly verified · Reasoned hypothesis · Internal input required · Illustrative output · Workflow tested · Live in operating system · Prototype · Requires integration.

Sourcing note (methodology, stated once): Facts carried over from Phase 2 cite 02_Company_Selection/scorecard.md directly and are not re-derived. Fresh research for this document was gathered via web search on 2026-09-04; direct WebFetch of the underlying source domains (techcrunch.com, ramp.com, g2.com, prnewswire.com, wikipedia.org, unite.ai, aiindustrytoday.com) returned EGRESS_BLOCKED for every domain tried in this session — an organization-level egress policy denial, not a transient failure, confirmed against a control fetch of Wikipedia. Facts below sourced this way are cited as [Publicly verified — via WebSearch snippet, <domain>, accessed 2026-09-04; direct fetch unavailable this session] rather than pulled from full page text. (Superseded 2026-09-06: the Material Claim Verification Gate re-ran in a local session with direct web access; every entry citing "fetched directly 2026-09-06" is a primary-document read, not search synthesis.) This is a real, disclosed capability gap for this session, stated once here per CLAUDE.md's standing-behavior rule rather than re-litigated per citation below.

Part A — ICP & Buying-Committee Model

A1 — OBSERVED: the public evidence bundle

This populates Module 02 Area 1's "Known evidence" categories with Ramp/Router-specific facts. Nothing in this subsection is OS output — it is the raw input the mechanism reasons over in A2.

Company baseline (carried from Phase 2 scorecard, not re-derived): - $750M Series F, closed 2026-06-04, $44B valuation, co-led by ICONIQ Capital, GIC, Ontario Teachers' Pension Plan. [Publicly verified — scorecard, citing TechCrunch] - Ramp Stack (AI accounting-automation) launched alongside the raise. [Publicly verified — scorecard, citing Payments Dive] - Router launched 2026-08-20/08-19. [Publicly verified — scorecard, citing TechCrunch/Benzinga] - G2 rating 4.8/5 across 600–2,452 reviews depending on which G2 aggregate page is cited. [Publicly verified — scorecard] - Public pricing: free core product; paid tiers from ~$15/user/month; enterprise negotiated. [Publicly verified — scorecard] - Competitive set: Brex, Airbase, Navan, Expensify. [Publicly verified — scorecard] - ARR estimated $1.4B+ (third-party estimate, lower-confidence figure). [Reasoned hypothesis — scorecard, per getlatka.com]

Reviewer/segment evidence (review-site language + firmographic pattern, fresh): - G2's Ramp product page shows the review base skewing toward Small-Business (~57%) and Mid-Market (~38%), with Enterprise a distinctly smaller share. [Publicly verified — via WebSearch snippet, g2.com, accessed 2026-09-04; not independently re-fetched by the 2026-09-06 gate (G2 blocks direct fetches) — single-snippet, load-bearing for A2, flagged] - (Removed 2026-09-06 by the Material Claim Verification Gate: a single secondary digest's claim that reviewers "reported issues with enterprise-level visibility and permissions" could not be corroborated on re-check (2026-09-06, search-snippet pass over the G2 and Capterra Ramp review pages, which block direct fetch — a negative search, not a read of every review); a negative claim about a company's product does not ship on one uncorroborated source.)

Ramp's own role taxonomy (a direct, load-bearing public source — Ramp's own help center, not third-party commentary): - Ramp's product defines a fixed base-role set (Business Owner/Admin, Employee, Accounting, IT Admin, Guest, View-Only Admin) plus add-on roles (Manager, Finance Admin, Accounts Payable, Assistant, Custom Roles). [Publicly verified — via WebSearch snippet, support.ramp.com "User roles overview," accessed 2026-09-04] - The Finance Admin add-on role is explicitly described by Ramp as designed for CFOs, Controllers, and Finance leads, with full financial-product access and read-only user-management visibility. [Publicly verified — via WebSearch snippet, support.ramp.com, accessed 2026-09-04] - This is a genuinely strong public source for the core Ramp platform's buying-committee shape: Ramp itself names its intended finance-buyer titles in its own documentation, which is stronger evidence than the job-posting/case-study inference Module 02 treats as the default-strength public signal.

Router-specific targeting language (pricing-page gating + competitor/category marketing, fresh): - Router's own framing, consistent across TechCrunch, Unite.AI, and PR Newswire coverage: "no Ramp card or company account required," a one-line base-URL swap for existing OpenAI/Anthropic SDK users, free routing through 2026, a $26 signup credit, and "enterprise features and more countries coming soon." [Publicly verified — via WebSearch snippet, techcrunch.com / prnewswire.com / unite.ai, accessed 2026-09-04] - Router runs on its own domain, router.com, distinct from ramp.com. [Publicly verified — via WebSearch snippet, prnewswire.com headline "Ramp Launches Router.com," accessed 2026-09-04] - Router's product mechanics target a developer/engineering audience specifically: single API endpoint, OpenAI/Anthropic-compatible, automatic fallback on provider failure, "100+ optimizations" across model selection/caching/compression/timing. [Publicly verified — via WebSearch snippet, multiple outlets, accessed 2026-09-04] - Ramp's own leadership framing at the time of its co-CEO announcement (Karim Atiyeh, formerly CTO, elevated to co-CEO alongside founder Eric Glyman, June 2026): "Technology is not a distinct part of the company — it is the entirety of it," and "every part of the company must be positioned to leverage the continued explosion in model intelligence." [Publicly verified — via WebSearch snippet, bankingdive.com/paymentsdive.com, accessed 2026-09-04] This is a why-now signal for treating Router as a strategic company bet, not a side skunkworks project.

Competitive/category convergence signal (competitor targeting patterns, fresh — the assignment's explicit ask): - Ramp's own vs.-Brex positioning frames Ramp as built for cost-conscious teams automating manual finance work (automation, policy enforcement, 99% accurate invoice extraction), while Brex is framed around funded, global-facing startups with multi-entity/global-payments needs. [Publicly verified — via WebSearch snippet, ramp.com/versus/brex and third-party comparison sites, accessed 2026-09-04] This is Ramp's own self-positioning language, a directly comparable public artifact per Module 02's "competitor positioning teardown" method. - Router did not launch into open territory. The same day (2026-08-19), Stripe announced its acquisition of OpenRouter — Stripe's own announcement states no price; press reports range from "over $7 billion" (Bloomberg, 2026-08-16) to more than $8 billion (Semafor, 2026-08-21; Axios) — with the two announcements landing roughly 70 minutes apart according to independent tech press. [Publicly verified — via WebSearch snippet, thenewstack.io "Forget the model wars, Stripe and Ramp just started the router wars," accessed 2026-09-04] Separately, Cursor and Meta are also building their own model routers. [Publicly verified — via WebSearch snippet, thenewstack.io, accessed 2026-09-04] A distinguishing fact reported by that same outlet: Stripe and Ramp do not sell their own models, while Cursor and Meta do — a real, sourced, third-party-drawn distinction between "neutral router" and "router with a model to sell." [Publicly verified — via WebSearch snippet, thenewstack.io, accessed 2026-09-04] - Stripe's own integration choice for OpenRouter is public and directly comparable: OpenRouter's own announcement stated it will "continue to operate as it is: same mission, same name, same product" post-acquisition — i.e., Stripe chose brand separation, not folding OpenRouter into stripe.com. [Publicly verified — via WebSearch snippet, OpenRouter's own blog language as reported by multiple outlets, accessed 2026-09-04] - A crowded field of comparably-positioned "cost-optimized model router" products exists independent of Ramp and Stripe: Azure AI Foundry's Model Router, Bifrost, Orq.ai Router, Inworld Router, and others all publicly claim cost-aware or business-metric-aware routing as their core value proposition. [Publicly verified — via WebSearch snippet, orq.ai/braintrust.dev/digitalocean.com comparison articles, accessed 2026-09-04]

A2 — OS CONTRIBUTES: applying Module 02's mechanism

This is the actual deliverable for Area 1 — not a description of Ramp, but the operating system's decision logic run against the evidence in A1.

Step 1 of the Decision (Module 02 §Decision.1) — firmographic + behavioral fit hypothesis. The first substantive judgment the mechanism makes, and one Ramp's own marketing does not state anywhere directly: Router and the core Ramp platform are almost certainly two different buying motions wearing one brand, not one ICP with a new feature. The evidence for this is inferential, not a quoted Ramp claim — it is built by reading the shape of the public evidence bundle against itself: the core platform is gated behind a named, documented buyer role (Finance Admin — CFO/Controller/Finance lead, per Ramp's own help center) and a pricing model with negotiated enterprise tiers, while Router is explicitly marketed as needing "no Ramp card or company account," lives on a separate domain, and is priced as a free self-serve developer product with enterprise terms still pending. Two artifacts pointing the same direction from independent angles (role-gating language + domain/brand separation) is the kind of convergence Module 02's validation logic treats as confidence-raising without being confirming — this is exactly that pattern, applied live. Confidence: Medium that the two are materially different committees; Low on the specific shape of either committee's roles, per Step 2 below.

Step 2 of the Decision — map to the canonical role set, confidence-rated. Because this is a bifurcated-motion hypothesis, the mechanism builds two role maps rather than one and treats the Router map as primary (it is the assigned test case), with the core-platform map included only as the contrast case the bifurcation argument depends on.

Core Ramp platform — buying committee (contrast case, not this doc's primary target):

Canonical role Candidate Confidence Evidence source
Economic buyer / champion CFO, Controller, or "Finance lead" (Ramp's own Finance Admin role definition) Medium Ramp's own help-center role documentation directly names these titles — stronger than inferred job-posting evidence, but still self-reported by the vendor, not confirmed by a deal record
Technical evaluator IT Admin (a named base role in Ramp's own taxonomy) Low-Medium Role exists and is named by Ramp; no public evidence on how much actual technical-evaluation authority it carries in a real deal
End user Employee base role; Accounting/AP staff for expense and bill-pay workflows Medium Named directly in Ramp's own role taxonomy
Procurement/legal/security blocker Unnamed in Ramp's public role taxonomy Low Absence of a named role is itself the signal — Ramp's documentation is UI-permission-oriented, not deal-process-oriented, so it doesn't surface this role even if one exists in real enterprise deals

Router — buying committee (this doc's primary target):

Canonical role Candidate Confidence Evidence source What would raise confidence
Champion Individual developer / AI-platform engineer performing the one-line SDK swap Medium Converges across independent coverage (TechCrunch, Unite.AI, PR Newswire) on dev-first, self-serve, no-account-required framing Internal: actual signup-flow data — self-reported title/role at signup, if Ramp collects it
Technical evaluator A distinct AI/platform infrastructure lead at organizations large enough to formally evaluate before scaling usage past a free trial; same person as champion at smaller orgs Low-Medium "Automatic fallback," "100+ optimizations," reliability-and-quality-bar language is written for a technical evaluator's checklist, not a finance buyer's Internal: real evaluation criteria used by actual trial accounts; whether a distinct eval stage exists at all before paid use
End user Same as champion — the API is consumed programmatically; there is no separate "user" persona the public evidence supports Medium Product mechanics (single API endpoint, SDK-level swap) are consistently reported n/a for this role — the mechanics are relatively unambiguous
Economic buyer Genuinely unresolved, and stated as such rather than forced to a single answer — two live hypotheses: (a) the same engineering leader who champions/evaluates it, once usage crosses from free trial into paid inference spend, in a pure self-serve motion; or (b) the existing Ramp economic buyer (CFO/Finance Admin) once "enterprise features" ship and Router becomes a negotiated line item inside an existing Ramp contract Low Evidence points both ways and is itself in tension: "no Ramp account required" argues against (b); the shared brand, the co-CEO "technology is the entirety of it" framing, and the fact Router is a Ramp property at all argues for eventual convergence toward (b) Internal: the actual Router GTM motion — is a Ramp AE/CSM ever involved, or is this run as a pure product-led-growth funnel with no sales touch, even for large accounts? This is the single highest-value internal input named in this document
Procurement/legal/security blocker A security/compliance reviewer, most likely triggered only once a company's usage moves from an individual's free-tier trial to sanctioned, budgeted, production spend Low Inferred, not stated: "U.S.-only" availability and "enterprise features... coming soon" language implies missing SSO, data-residency, or compliance documentation that would already stall a security-mature buyer's evaluation today — no public source confirms this gap directly, it is read from an absence Internal: whether SOC2/data-handling documentation already exists for Router today and simply isn't public-marketed, versus genuinely not existing yet
Bridge / expansion role (non-canonical addition, OS-flagged) The existing core-platform economic buyer (CFO/Finance Admin), encountering Router as a cross-sell inside the broader "Ramp is an AI-native finance platform" narrative Low Inferred from the co-CEO framing and shared brand equity, directly contradicted by the "no Ramp account required" self-serve framing — flagged here specifically because the public evidence is internally inconsistent on this point, which is itself useful output: it tells Jerry's system exactly where the ambiguity lives rather than resolving it falsely Internal: same as Economic buyer row above — GTM motion data resolves both at once

Per Module 02's Decision step 4, every row above is published as Reasoned hypothesis, internal validation required — none is claimed as settled, and the "Bridge/expansion" row is a genuine judgment call this operating system is adding on top of the canonical five-role set defined in 01_OS_Architecture/02_icp-and-positioning.md, because the evidence itself surfaced a role (an existing-customer expansion buyer) the fixed role set doesn't name. That addition, and the decision to flag it rather than force-fit it into "economic buyer," is itself an OS output, not an observation about Ramp.

A3 — OS CONTRIBUTES: missing internal inputs, sharpened for this case

Module 02's generic internal-input list (CRM win-rate data, deal-room composition, etc.) is a template. Applied to Router specifically, the mechanism names these as the highest-value gaps, in priority order:

  1. Router GTM motion data — is any Router account, at any size, ever touched by a Ramp AE/CSM, or is it 100% self-serve today? This single fact resolves the Economic-buyer and Bridge-role ambiguity flagged in A2 more than any other input.
  2. Signup-funnel title/role data — if Router's signup flow captures self-reported title or company, that data would convert the Champion/Technical-evaluator rows from Medium/Low confidence to something closer to ground truth without needing a single sales call.
  3. Base-population overlap — what share of Router trial or paid accounts are already core-Ramp customers vs. net-new logos never previously in a Ramp relationship? This is the direct test of the bifurcated-motion hypothesis in A2 Step 1.
  4. Actual enterprise security/compliance posture — whether SOC2, data-residency, or SSO documentation exists today for Router (even if not yet public-marketed), needed to confirm or kill the Blocker-role hypothesis.
  5. Real deal-room composition for the first enterprise Router contracts, once "enterprise features" ship — the only ground truth for whether Finance/procurement actually re-enters the process as Router scales past self-serve.

A4 — OS CONTRIBUTES: validation run now, without internal access

Module 02's Area 1 validation names one check available pre-Phase-3: whether the hypothesized ICP pattern converges with what's visible in competitors' own published material. Run live against real evidence gathered above:

Part B — Positioning & Message System

B1 — OBSERVED: Ramp's actual public messaging for Router

Raw input only — this subsection restates what Ramp and independent press already say, which is exactly the material the Jerry-add test warns against treating as the deliverable. It is included because Area 2's mechanism requires it as an input; B2 is where the OS does something with it.

B2 — OS CONTRIBUTES: the positioning hypothesis

Applying Module 02 Area 2's structured framework (competitive alternatives → unique attributes → value delivered → target customer → market category) to the evidence bundle, not restating Ramp's own claims:

The category claim itself is not open territory, and the OS names this as the central risk rather than passing over it. "Cost-optimized model routing via a single API" is claimed, with materially similar language, by Azure AI Foundry's Model Router, Bifrost, Orq.ai Router, Inworld Router, and — most directly — by OpenRouter itself, now backed by Stripe's payments infrastructure (acquisition announced 2026-08-19; price undisclosed by Stripe, reported between over $7B and more than $8B). Per Module 02's own validation logic ("does the differentiation claim survive a side-by-side check against every competitor's own public claims — is it actually not already claimed by three other vendors in the space?"), Router's headline claim fails this check: it is claimed by at least five other vendors, not zero. This means the 40%-cheaper / lowest-cost-model claim cannot be the durable differentiator this messaging system leans on, regardless of whether it is Ramp's own headline in launch coverage today. That is an OS-derived conclusion, not something Ramp's own site says about itself.

A more durable differentiator exists in the evidence. Ramp already states both halves of it in its own launch material; the OS's contribution is choosing which half goes to which buyer, how far it can be pushed, and whether it survives a durability check. (1) Neutrality. Ramp's launch post (2026-08-19, section "Independent by design") states: "We don't build models, so we have no reason to favor any particular one." The New Stack independently drew the same line between routers with a model to sell (Cursor, Meta) and routers without one (Stripe, Ramp). OpenRouter sells no models either, so this claim separates Router from Cursor- and Meta-style routers, not from the incumbent, and it is not unique. (2) Spend position. Ramp's press release states that Router "connects model selection decisions to Ramp's broader AI spend visibility and controls," and router.com frames the product as "Built for CTOs. Loved by CFOs." No pure-play router (OpenRouter, Bifrost, Orq.ai, Inworld) is the system of record for the rest of a company's vendor spend. The OS's positioning decision: Router's durable differentiation is the spend position, not "cheapest routing" (contested, unverified, matched by five-plus competitors). The sharpest form of it, "the only model router that already sees and reconciles the rest of your company's spend," goes one step past Ramp's own release, which stops at "visibility and controls." This is a Reasoned hypothesis about which existing claim to lead with and how far to push it, not a message the OS invented. Ramp's headline still leads with the cost claim, and its "no Ramp account required" mechanic runs counter to the spend angle, which the OS flags as a live tension, not a settled recommendation (see B4). Primary sources, fetched directly 2026-09-06 and snapshotted in sources/ramp-router-launch-2026-08-19/: the neutrality sentence appears in the blog post only, the spend sentence in the press release only. The first version of this paragraph said neither claim was one "Ramp's own launch messaging currently makes"; that was false against Ramp's own words and is corrected here on the record (JOS-31, D-067).

Why-now narrative, OS-derived version: not "Ramp built a router" (the vendor framing) but "the same three years of internally-tuned AI-spend-routing data that already cuts Ramp's own bill is now externally available at the exact moment routing infrastructure is consolidating (Stripe/OpenRouter) and could plausibly be positioned as spend-management expertise applied to a new spend category, rather than as a new, undifferentiated infrastructure product." This reframes the why-now around Ramp's actual institutional advantage (spend data, financial-ops credibility) rather than the commodity claim (cost savings), which is the kind of internally-generated-vs.-buyer-evidence distinction Module 02 names as the single most common PMM failure mode to avoid.

B3 — OS CONTRIBUTES: the messaging matrix

Keyed directly to the Router buying-committee map built in A2 — per Module 02's explicit handoff rule, this matrix cannot be built without that map, and every low-confidence role in A2 propagates a corresponding low-confidence cell below.

Role (from A2) Funnel stage Message Proof point Evidence source Confidence
Champion (developer) Awareness "Swap one line, keep your code" One-line base-URL change for existing OpenAI/Anthropic SDK integrations Publicly verified (launch coverage) [Restates Ramp's own current copy — not an OS addition] High — this message is Ramp's own current claim and is well-evidenced
Champion (developer) Consideration "Built by the team that's already run this in production for three years" Internal-use claim, "three years" of internal routing before public release Publicly verified as a claim; unverified as a fact Medium — claim exists but independent verification doesn't
Technical evaluator Consideration "Neutral routing — we don't sell models, so we have no reason to route you to our own" Ramp's own launch post, 2026-08-19: "We don't build models, so we have no reason to favor any particular one"; The New Stack's independent Stripe/Ramp vs. Cursor/Meta framing Publicly verified (Ramp's own launch copy, corroborated by third-party press) Medium — observed in Ramp's launch post; selected by the OS for this role. Ramp leads with the cost claim, not this line. Shared with OpenRouter, so it separates Router from Cursor- and Meta-style routers only
Technical evaluator Decision "Automatic fallback, 100+ optimizations, quality-bar routing" — reliability/quality checklist items Publicly reported product mechanics Publicly verified [Restates Ramp's own current copy — not an OS addition] Medium — real claims, but not competitively differentiated per B2's durability check
Economic buyer — self-serve engineering-leader hypothesis (A2) Decision Message untestable from public evidence alone — no public pricing exists past the free period, so no economic-buyer-facing value message can be confidently drafted n/a n/a Low — Internal input required before this cell can be filled at all
Economic buyer — existing CFO/Finance-Admin bridge hypothesis (A2) Decision (future / cross-sell) "The AI spend you're already worried about is now inside the platform that already manages the rest of your company's spend" The B2 spend position applied to this role. Ramp's release already connects Router to "broader AI spend visibility and controls" and router.com says "Loved by CFOs"; the OS's addition is aiming that claim at the existing-customer finance buyer as a cross-sell line, phrased as the buyer's worry rather than the product's feature, to test the bridge-role hypothesis from A2 Reasoned hypothesis, built on Ramp's own release framing Low — depends entirely on the unresolved GTM-motion question named in A3
Security/compliance blocker (A2) Decision No public message exists targeting this role at all — Router's current public material has nothing addressed to a security reviewer n/a Absence is itself the finding Low — Internal input required; this is a named gap in Ramp's current public messaging, not a gap in this research

Two cells in this matrix are OS-selected lines (the "neutral routing" line for the technical evaluator, and the spend-data-bridge line for the CFO/Finance-Admin economic buyer). Both draw on claims Ramp already makes in its own launch material: the neutrality sentence in the blog post, the spend connection in the press release. Neither is the line Ramp leads with. router.com already names both audiences ("Built for CTOs. Loved by CFOs."); what the matrix adds is the funnel stage, the proof point and the confidence for each role, and the cross-sell framing for the existing-customer finance buyer. The first version of this document called them "OS-authored candidate messages that Ramp is not currently running"; that was false against Ramp's own launch post (primary source fetched 2026-09-06) and is corrected here on the record. The mechanism's job is to say which existing claim goes to which buyer, how far it can be pushed, and whether it survives a durability check, not to invent a message nobody has said.

B4 — OS CONTRIBUTES: validation before any of this ships

Per Module 02 Area 2's validation logic, applied to this specific matrix:

Human owner, status, and update trigger (per Universal Module Contract)

Human owner: This document is PMM-function output (this build) demonstrating Module 02's mechanism; Sales/RevOps-equivalent internal validation is structurally unavailable for a company with no engagement relationship, which is the point being demonstrated, not a gap being hidden. Per Module 02, any resulting claim used externally about Ramp is Founder-approval-gated and would additionally require Ramp's own involvement — nothing here is a claim about to be published as fact.

Current status: Prototype — the mechanism (Module 02's Decision logic) has now run against one real company's real public evidence, which is a step beyond the "Reasoned hypothesis, mechanism designed, not yet run" status Module 02 carried at the architecture stage. Every specific role/confidence/message conclusion above remains Reasoned hypothesis, internal validation required — running the mechanism once does not validate its outputs; only internal data (named in A3) does.

Update trigger: Per Module 02's linked cadence — if Router's actual GTM motion becomes publicly discoverable (a Ramp sales-assisted enterprise Router deal gets case-studied, or Router pricing/enterprise terms are announced), this document's A2/A3 should be re-run immediately, since that is precisely the input this document identifies as the single highest-value missing piece. If a competitor claims "neutral, disinterested routing" or "spend-data-aware routing" publicly, B2's differentiation-durability conclusion should be re-checked before being relied on further.

[op:claude-interactive | JOS-25 Phase 3 | 2026-09-04 | authority: Notion ticket JOS-25 Current Contract v2 + canon/DECISIONS.md D-024]