The Defensibility Diagnostic · Step 1 of 3

Abridge
moat scorecard.

A 0–3 score across the 8 moats, anchored to comparable companies. It surfaces where Abridge's defensibility is real, where it's thin, and where the story outruns the structure.

STACKED · 10 / 24 · 3 moats at 2+
Scored August 18, 2026 · Framework: 8 Moats for the AI Era according to Gokul Rajaram, 20VC Interview
Headline Finding

Abridge is Stacked. Three real moats, and not one of them is the model.

Abridge clears the three-moat threshold on Workflow, Distribution, and Data: 10 points out of 24, with three moats at 2 or higher. That places it above nearly every AI-native challenger and below the scaled incumbents it now competes against. Nothing on this board scores a 3. Every moat Abridge holds is one that a better-capitalized competitor is actively contesting, which is the difference between defensible and safe. The useful finding is not the total. It is where the durability actually sits, because it is not where the company points and not where the critics look.

Defensibility state: STACKED · 10 / 24 · 3 moats at score 2 or higher
⚠ The Moat Narrative Trap

The model is the marketed wall. The deployment is the load-bearing one.

Abridge's public story now runs through the model layer: an AI-native platform, a Contextual Reasoning Engine, the first foundation model purpose-built for clinical conversations. The scoring puts the model at the most contested surface on the board. It is fine-tuned on an open-weight family, the company still routes work to third-party labs by its CEO's own account, and Microsoft and Epic are building the same capability with more capital and closer proximity to the record. What actually holds is duller: 300+ health systems live, clinicians who now document by talking, and coding and billing entangled downstream of the note. Worth noting that the loudest critique of this category makes the same mistake in reverse. Calling Abridge a wrapper and calling it a foundation model company both locate the moat in the model. The structure says the moat is in the deployment.

The Scorecard

Eight moats, scored at a glance

Each moat is scored 0–3 for structural durability — not for how useful or popular it is. The tier is set by how many moats reach 2+, not by the total.

MoatMeterScoreOne-line take
01Data2A real cross-customer corpus, with a comparable one sitting at Microsoft.
02Workflow2The deepest 2 in the category, and still a 2.
03Regulatory1Compliance infrastructure done well, not a legal barrier.
04Distribution2Owned installed base, contested channel.
05Ecosystem1An integrator, not a platform. Dependency points outward.
06Network Effects1A training loop, not a network.
07Physical / Infra.0Structurally unavailable, not missing.
08Scale1Category leadership without demonstrated cost or pricing power.
Total10 / 24
How the tier is assigned

Durability comes from stacking real moats

The tier is set by counting how many moats score 2 or higher — not by the total out of 24. One strong moat isn't a stack.

Exposed0 moats at 2+No defenses; open to attack from every direction.
Partial1–2 moats at 2+Some defense, but below the stack threshold.
Stacked← This company3 moats at 2+Stack threshold met; defensibility begins here.
Fortified4+ moats at 2+Multiple moats compound; the strongest state.

Abridge has 3 moats at 2 or higher, which lands it in Stacked.

Moat by moat

Each score, anchored to the competitive set

Scores are anchored against comparable companies in the same category — not isolated opinion.

01Data
2 / 3

Abridge processes more than 100 million clinical conversations a year across 300+ health systems, and trains on de-identified data through pre, mid, and post-training stages. That is genuine cross-customer compounding, not the per-tenant data most B2B software calls a data moat. It stops short of the top of the ladder for two reasons. A comparable corpus exists: Microsoft's Nuance has been collecting clinical audio far longer at roughly the same market share. And the rights are granted by health systems under contract rather than owned outright, which makes the asset durable but not unconditional. Gong scores 3 here because nobody else holds its conversation corpus. Abridge holds a very good version of a corpus that two other well-capitalized parties also hold.

02Workflow
2 / 3

This is the moat doing the most work, and it is the one the wrapper framing misses entirely. Kaiser Permanente rolled out to roughly 24,600 physicians, Northwestern Medicine went enterprise-wide, and the output now feeds notes, flowsheets, billing codes, and orders. Removing it means retraining thousands of clinicians and disturbing the revenue cycle, which is switching cost by any definition. It caps at 2 rather than 3 because Abridge is not the system of record. Epic is, and Epic shipped a native alternative in February 2026, which hands every Epic customer an exit path that Salesforce customers do not have. Salesforce holds the Workflow 3 in this framework precisely because there is no house alternative one menu away. HubSpot sits at 2 for the same structural reason.

03Regulatory
1 / 3

HIPAA posture, BAAs, consent architecture, note traceability back to the source audio, and an AHIMA collaboration on coding auditability all represent real compliance investment. None of it is a license a competitor cannot obtain. Ambience, Suki, Nabla, and Nuance operate inside the same envelope, which makes this table stakes done well rather than a barrier. A 2 would require something structural, closer to Stripe's money transmitter licenses, where the wall is legal rather than operational. If ambient tools push far enough into clinical decision support to trigger device regulation, or if coding accuracy becomes a certified rather than a marketed claim, this score moves.

04Distribution
2 / 3

300+ health systems live with more added weekly, land-and-expand inside each system from physicians to nurses, and two consecutive Best in KLAS rankings that enterprise buying committees use as a procurement filter. That is owned reach. The qualifier is that a meaningful share of the channel runs through EHR partners, and the most important of those partners changed sides: Epic wound down the co-development program, sold its stake in Abridge, and launched a competing product. Hims anchors the Distribution 2 in this framework for a similar reason, an owned brand that does not lock the category. A 3 would require a channel that brings in the next health system without Abridge paying for it.

05Ecosystem
1 / 3

Abridge integrates into Epic, Oracle Health, and athenahealth, and licenses evidence content inward from UpToDate, NEJM, JAMA, the American Heart Association, and others. Smart-room partnerships with Artisight and hellocare.ai point the right direction but are thin. Nobody has built a business on top of Abridge. Salesforce scores 3 here because AppExchange supports an entire labor market; Clay scores 2 because builders make a living on its templates and enrichment. Abridge sits with Apollo at 1: a real API surface, consumed rather than depended on. A 2 would require third parties whose own revenue runs through Abridge.

06Network Effects
1 / 3

More deployments do make the shared models better, which is why this scores above zero. It is not a network effect in the strict sense, because no clinician gets value from another health system's participation directly, only through a retraining cycle already counted under Data. The payer and life sciences work announced in June 2026, with Aetna, Cigna, and an Eli Lilly investment, is the version that could become genuinely two-sided, since providers and payers would each be worth more to the other on shared rails. That is announced, not built, and it is not scored here. Apollo earns the 2 on contributory data users actually exchange. Nothing in this category approaches the G2 or LinkedIn 3.

07Physical / Infrastructure
0 / 3

Software on rented compute. The NVIDIA collaboration runs on Blackwell infrastructure Abridge does not own, and the smart-room hardware belongs to Artisight and hellocare.ai. Salesforce's Hyperforce residency layer is the ceiling a software platform reaches on this moat, and that is a 1. Read this as a road not taken rather than a gap. Scoring an unavailable moat as a failing is a category error.

08Scale
1 / 3

Roughly 30 percent of a $600M ambient scribe market with Microsoft at roughly 33 percent, and contracted ARR reported near $117M in early 2025 against approximately $830M raised. Leadership by mindshare is not leverage. Pricing sits mid-field, above Nabla and below Nuance DAX, which is a competitive position rather than pricing power. Distilling onto open-weight models is a real unit-cost lever at 100 million conversations a year, but it is a plan, not a demonstrated margin. ZoomInfo earns a 2 on demonstrated operating leverage; Calendly and Apollo sit at 1 for the same reason Abridge does. A 3 stays out of reach while Microsoft and Epic can bundle at zero.

Structural availability

Which moats are even on the table

Not every moat fits every business. Three buckets narrow the field — they don't name which available moat to pursue. That's what the Stress Test is for.

Available to deepen

Moats at 2+ where investment compounds

  • Workflow (2) clinician ritual plus downstream coding and billing entanglement across 300+ systems.
  • Distribution (2) an installed base that expands inside each system, filtered through two years of category-leading procurement rankings.
  • Data (2) a cross-customer clinical conversation corpus compounding at 100M+ encounters a year.
Available to build

Moats this model could support

  • Network Effects (1) the provider-payer and life sciences rails are announced but not yet load-bearing.
  • Regulatory (1) compliance depth exists; a certified rather than marketed accuracy claim would change the category.
  • Scale (1) the model-distillation cost lever is real but undemonstrated in margin.
  • Ecosystem (1) an API surface exists; nobody yet builds a business on it.
Structurally unavailable

Moats the model doesn't fit

  • Physical / Infrastructure (0) software on rented compute. Hyperforce-class residency is the ceiling here, and that is a 1.
What this scorecard doesn't tell you yet

The Diagnostic finds the state. The Stress Test finds the move.

You've seen where Abridge stands. Two ways forward: run this same free diagnostic on your own company, or commission the rigorous version.

Free · Done for you · Next business day

Run your own FREE diagnostic

Please enter your name.
Free providers (gmail, yahoo, outlook, icloud, hotmail) aren't accepted.
Please use a business email (not a free provider).
We score what's publicly visible — your homepage, product pages, about page.
Please enter your company.
Please select a stage.
Please pick one company type.
Your scorecard, plus a short email series on what to do with it. Unsubscribe anytime.
Customer Centric Solutions LLC · customercentricllc.com
Request received.

Thanks — your scorecard and a short email series are on the way, and I'll be in touch within one business day. If you want the rigorous version, the Stress Test on the right is the next step.

Post-discovery · Done for you · The rigorous version

Commission the Defensibility Stress Test

The same eight moats, run with the full methodology visible. Where the Diagnostic names the state, the Stress Test names the move:

  • 01Five lenses, applied with attribution · every score traced to the analytical test that produced it
  • 02Trajectory per moat · appreciating, holding, or eroding, because a 2 going up is not a 2 going down
  • 03Both diagnostic flags · contestation risk and disruption exposure, assessed rather than assumed
  • 04Head-to-head competitive scoring · the same eight moats run on 2 to 4 named peers
  • 05The narrative trap, named · which documented trap shape the company has fallen into
  • 06Investment Reality Check · what the market is mispricing about this stack
$1,350 · Fixed scope · Starts with a discovery call
Book a discovery call →