US Software: Seat Compression, Data Moat Durability, Regulatory Entrenchment Determine AI Displacement Risk Across Select High-Yield Credits

• The AI displacement threat is not uniform across software credits. How vulnerable a company is depends on (1) whether its revenue is tied to human head count, (2) how proprietary the underlying data is or (3) how deeply entrenched its product is within regulated workflows, among other traits.
• Seat-based vendors face a structural pricing model risk that is independent of product obsolescence. The relevant question is not whether AI replaces the platform but whether AI-driven enterprise head count reductions compress the unit of monetization, a dynamic that is most imminent for ZoomInfo, given sales head count pressure, and most gradual for RingCentral, where united-communications-as-a-service, or UCaaS, seats are provisioned across broadly distributed knowledge worker populations.
• For data-as-a-service providers, the displacement threat concentrates above the data layer and at the analytics and workflow intelligence layer, where foundation models are enabling customers to replicate premium-tier capabilities natively. Clarivate‘s core patent and citation assets are more insulated given their legal provenance and nonreconstructible curation, while ZoomInfo‘s contact data relies more heavily on publicly available sources and carries greater replication exposure compounded by seat compression.
• Credits with regulatory and institutional entrenchment, such as Fair Isaac and Consensus Cloud, exhibit a distinct risk category where AI can replicate the technical function but cannot replicate the regulatory approval, legal authority and institutional adoption that make these standards the required reference point. The relevant monitoring risks within the bond maturity horizon are competitive and secular rather than technological.
• RingCentral‘s 2030 unsecured notes, indicated at an approximately 320-bps spread, represent the most straightforward relative value opportunity across the credits discussed in this report. With only 1.7x net leverage, retention above 99% and low product displacement risk, RingCentral’s fundamentals do not support pricing roughly 140-bps wide of the BB index, in our view. The market appears to be conflating pricing model risk with product obsolescence risk in a way that overstates the near-term credit threat.
• ZoomInfo’s 2029 unsecured notes, indicated at a 700-bps spread and 2.4x net leverage, appear to compensate for relatively higher exposure to displacement risk, though there remains risk of further widening given its relative exposure to AI-related pressures, which could pressure pricing power and margins. Its near-term fundamentals are supported by considerable free cash flow generation and resilient upmarket retention, though investors would require conviction on capital allocation priorities given the risk that AI displacement in the upmarket accelerates ahead of the 2029 refinancing window.
• Among the lower displacement risk credits, Consensus Cloud’s 2028 notes have largely recovered recent spread widening and now trade inside the single-B index, limiting further tightening potential, while Fair Isaac’s 2033 unsecured notes at an approximately 230-bps spread represent the most straightforward example of regulatory and institutional moat defensibility for investors seeking software credit exposure with minimal AI displacement risk.

• ZoomInfo (High): Dual exposure across seat compression and data replicability, with an additional architectural vulnerability in the intent product that cannot be addressed through incremental investment.
• Clarivate (Medium): Core data is defensible, but the workflow and intelligence layer above it faces growing replication risk that could compress EBITDA within its bond maturity horizon.
• RingCentral (Medium): Seat-based model is structurally exposed in principle, but enterprise head count compression is gradual and current retention metrics suggest limited near-term materialization.
• Twilio (Low): Consumption model insulates against seat compression, AI agent proliferation is a direct volume tailwind and per-unit economics are expanding rather than compressing as AI voice orchestration commands a meaningful premium to standard programmable voice.
• Fair Isaac (Low): Moat is institutional and regulatory rather than product-driven; near-term risk is VantageScore competition in conforming mortgage, not AI.
• Consensus Cloud (Low): Compliance entrenchment creates switching costs independent of available alternatives; relevant risk is secular fax volume decline, not displacement.

Historical net revenue retention, or NRR, or the most applicable similarly reported metrics, are shown below:

The impact of rapidly evolving artificial intelligence capabilities on software companies has been widely publicized and debated frequently over the past few months, with fears of widespread displacement causing a substantial selloff across software equity and bond instruments. Octus addressed the initial selloff in February and published a follow-up report shortly afterward on how select companies may or may not be exposed to displacement.

Since then, although overall multiple compression has slowed, software equity and debt instruments have remained highly volatile, with entire subsectors within the broader category swinging up or down materially on just about every new feature announcement or leaks of a new model by the major large language models. Year to date, the iShares Tech-Software ETF (IGV) is down over 21%, versus the broader Nasdaq 100 index being up almost 4% year to date.

Below, we show the historical yield for an Octus-constructed weighted average index of 50 of the largest software high-yield bonds relative to the broader ICE BofA high-yield index. Software spreads have widened significantly since late January, while the broader high-yield index has widened only moderately during the same period, reflecting both idiosyncratic credit deterioration and a market-wide repricing of software names that, in our view, has been applied with insufficient precision across names with meaningfully different competitive profiles.

Further, we show more detailed pricing moves across the 50 largest high-yield software bonds compared with the broader ICE BofA high-yield indexes over a few different time horizons, below. As shown, there has been a moderate recovery in software spreads in the last month, though over the past three months, the group has significantly underperformed the broader indexes.

In the sections below, we examine the AI displacement moat, or lack thereof, for six high-yield issuers organized around the mechanisms that actually determine defensibility: seat compression exposure and pricing model transition for ZoomInfo, RingCentral and Twilio; proprietary versus publicly available data as a determinant of pricing power for ZoomInfo and Clarivate; and regulatory and institutional entrenchment as a displacement buffer for Fair Isaac and Consensus Cloud. Where current spread levels appear to diverge from that underlying analysis, we draw relative value conclusions towards the end of the analysis.

Against a backdrop where software spreads have widened significantly relative to the broader high-yield index, with insufficient precision applied across names with meaningfully different competitive profiles, the credits discussed in this piece present a range of relative value outcomes – from names where current spread levels appear well-anchored to fundamentals to those where the market’s implied impairment thesis appears to overstate near-term credit risk.

Twilio‘s 2029s and 2031 notes indicate spreads at 120 bps and 127 bps, only slightly wide to the BBB index despite a BB+ rating and well tight of any of the other credits included in the table above. We view this as justified given the consumption model’s structural insulation, AI volume tailwind and 0.3x net leverage. High free cash flow generation that totaled almost $900 million in fiscal year 2025 and almost $600 million in fiscal year 2024 further support the company maintaining sub-1x net leverage. Twilio’s notes are relatively protected on the downside due to the company’s strong credit profile and relative insulation from AI displacement trends, though at spreads indicating near-investment-grade levels, upside from further spread tightening is relatively limited.

RingCentral‘s 2030 unsecured notes indicating a spread of about 320 bps, approximately 140-bps wide of the BB index, are the most straightforward spread-tightening opportunity in the table. Net leverage of 1.7x, retention above 99% and low product displacement risk make the roughly 200-bps spread differential to Twilio appear excessive. The seat compression risk is real but gradual, and current fundamentals do not support pricing this wide of the index.

ZoomInfo‘s 2029 unsecured notes indicated at a spread of 698-bps price the credit as structurally impaired. Net leverage of 2.4x and $300 million to $400 million of annual free cash flow suggest the near-term credit picture is more resilient than that discount implies, and upmarket NRR near 100% provides a credible stabilization narrative.

The notes offer compelling compensation for the uncertainty but would require conviction that upmarket retention holds as the notes approach refinancing, and the capital allocation debate between buybacks and debt repayment remains a key watch item given that refi interest could be limited as longer-term, AI-related churn risk comes into closer view. There also remains further downside risk given its relatively higher exposure to AI displacement, which could introduce incremental pricing pressures and margin deterioration.

Although Clarivate’s core data offering is relatively more defensible than ZoomInfo’s, it is significantly more levered at 4.1x and carries its own moderate level of displacement risk. The secured notes due 2028 indicate about 80-bps wide of the BB index, though at 3.2x net leverage, we feel the risk-reward dynamic is more favorable for ZoomInfo’s 2029 notes, which reflect a sizable premium to compensate for risk and are less levered, which provides some downside protection in a scenario where the displacement thesis begins to play out ahead of the maturity date.

Although Clarivate’s unsecured notes due 2029 offer a compelling premium to the secured tranche despite only an additional turn of leverage, ZoomInfo’s notes still offer more than an 80-bps spread premium, with almost 2x fewer turns of leverage, which we believe more than compensates for the incremental displacement risk.

Fair Isaac and Consensus Cloud both stand out as low displacement risk credits with spread levels that appropriately reflect that view. Consensus Cloud’s unsecured notes due 2028 have more than recovered any spread widening that took place over the past few months, tightening by almost 40 bps over the past month.

Its notes indicate about 30-bps tight of the single-B index, and although its low displacement risk and modest leverage profile protect against significant spread widening from these levels, further spread-tightening potential appears relatively limited, given that the notes already indicate inside of the single-B index and the company’s recent comments that it will shift its capital-allocation bias toward share buybacks over additional debt retirement after deleveraging about a turn over the past two years.

Still, with relatively protected downside, consistent positive free cash flow generation and manageable leverage, the notes may warrant consideration given the steeper risk characteristics within the broader software space.

Fair Isaac’s unsecured notes due 2028, 2033 and 2034 indicate spreads of 174 bps, 230 bps, and 240 bps, respectively, with the two longer-dated notes indicating 50 bps to 60 bps premiums to the BB index. Relatively tighter spreads than some of the more exposed credits above to AI displacement risks correctly reflect its institutional and regulatory moat that places it in a different risk category than most software credits.

The Federal Housing Finance Agency’s 2025 decision permitting VantageScore as an alternative in the conforming mortgage market is the most credible near-term headwind, though the operational burden of transitioning underwriting models limits the pace of adoption meaningfully. The notes due 2033 represent the cleanest low displacement risk expression in the table for investors seeking defensive software credit exposure, with mortgage origination pricing power the more relevant monitoring item than any AI-displacement scenario.

Although much of the concern driving the broader software selloff has centered on AI directly replicating software functionality, whether through AI-native startups capturing market share or enterprises building capable alternatives in-house, a consequential secondary impact warrants equal consideration. The potential head count reductions that follow successful AI deployment across the enterprise represent a structural headwind to any software vendor with seat-based pricing, quietly compressing addressable markets without requiring enterprises to make any active platform replacement decision. If AI productivity gains translate into leaner workforces, seats would simply go unrenewed at contract expiration, placing compounding pressure on top lines.

The risk applies to any vendor that charges by the seat. The speed at which that pressure materializes depends on how quickly AI can replicate the specific workflows those seats were purchased to support. The magnitude of the ultimate impact remains uncertain, though incumbent vendors are not standing still.

The most common response has been a deliberate pivot away from per-seat pricing toward consumption-based, outcome-based or value-based models that decouple revenue from head count entirely – an acknowledgment by management teams that the per-seat model faces structural pressure but also a credible mechanism through which durable revenue can be preserved if the transition is executed successfully.

As the slide below from a Redpoint software presentation shows, software companies are broadly expected to shift to a more usage-based approach to pricing over the next few years.

ZoomInfo

Go-to-market intelligence platform ZoomInfo has historically been priced on the number of sales and marketing seats accessing the platform, with its top-line performance directly correlated to head count growth from existing enterprise customers and the acquisition of new enterprise customers.

In response to mounting seat compression pressure, the company has taken steps to diversify away from pure seat-based pricing through three mechanisms: a data-as-a-service offering that bills customers on a consumption basis via API and cloud data integrations rather than seats, enterprise license agreements that provide enterprise-wide access and decouple contract value from individual user counts, and the Copilot platform, its AI-powered prospecting assistant, which now represents more than 20% of total average contract value, or ACV, having more than doubled during 2025. However, the majority of contracts remain tied to seat volume, leaving the company materially exposed as contracts reach renewal dates.

ZoomInfo’s NRR declined to 87% in 2023 and 2024 as technology companies, its most concentrated customer segment, reduced sales and business development head count following COVID-era overhiring, before recovering toward 90% in 2025, as the company increased its exposure to the upmarket.

The upmarket, which consists of large enterprises embedded across Salesforce, marketing automation and sales development representative tooling on multiyear agreements, has maintained NRR above 100%, supported by deep integration switching costs and differentiated products such as intent data that lack credible low-cost substitutes.

The downmarket, by contrast, used ZoomInfo primarily as a contact lookup tool, a use case heavily commoditized by AI-native competitors such as Clay and Apollo, driving churn to 20% to 30%.

Upmarket now represents 74% of total ACV, with a stated target of 80% by 2027, and the company sees a path to 105% upmarket NRR driven by consumption-based AI products. Even so, at 90% consolidated NRR, ZoomInfo’s existing customer base is shrinking by ten cents on every dollar annually, requiring continuous new logo acquisition to hold revenue flat at a moment when the addressable pool of go-to-market seats faces structural pressure for the first time.

RingCentral

Unified-communications-as-a-service, or UCaaS, provider RingCentral sells cloud-based voice, video and messaging subscriptions to enterprise customers on per-user contracts, making its revenue directly tied to its customer head count. Management has pushed back on more extreme displacement scenarios, noting at a December conference that it views the demise of human workers as “largely overexaggerated” though conceding that certain roles are being eliminated and that a number of use cases can be offloaded to purely AI.

The CEO separately acknowledged that the contact-center-as-a-service, or CCaaS, market, which it has some exposure to, is slowing as AI displaces human contact center agents but argued that RingCentral is a net beneficiary gaining share as enterprises migrate toward AI-native contact center solutions and that its revenue mix is heavily weighted toward UCaaS rather than CCaaS, where seat compression is a more gradual dynamic tied to broader enterprise head count trends rather than acute agent displacement.

Over the past few years, it has maintained net monthly subscription dollar retention rates consistently above 99%, signaling strong retention and limited seat compression from its customer base. Although its core UCaaS product still features a seat-based pricing model, the company has rolled out new AI products that are being layered on top of per-seat subscriptions with pricing based on usage, which have been supportive of both average revenue per user, or ARPU, and net retention rates.

RingCentral’s defensibility against AI displacement is concentrated around whether the consumption revenue generated by its AI add-on products can offset seat erosion on a net basis over a full renewal cycle. Management notes that dedicated AI product ARR is approximately $100 million, or roughly 4% of its $2.5 billion estimated ARR base (using reported annual revenue as a proxy), meaning the offset math has a long way to go before it functions as a structural cushion against seat compression.

Compounding this, some of its AI products are priced per user, which also exposes these products to seat compression. The near-term CCaaS share gain story from enterprises re-platforming onto AI-native contact center infrastructure provides a partial offset, but that tailwind is finite and concentrated in the migration cycle rather than a durable structural dynamic.

The longer-term question is whether agentic AI reaches a capability threshold that converts knowledge worker displacement from a slow attrition-driven dynamic into something more acute, at which point the UCaaS seat base, which today represents the more gradual and manageable compression risk, becomes a more material headwind to revenue and free cash flow.

Twilio

Twilio occupies a structurally distinct position in this discussion. The majority of its revenue is already consumption-based rather than seat-based, charging per message sent, per call minute and per API call, rather than per employee, providing genuine insulation from the head count-driven churn mechanism that threatens pure per-seat vendors.

Management has acknowledged the dynamic directly, noting that if seat counts decline, value previously generated from license-based arrangements transitions to AI-oriented workload consumption. Twilio also cited that 75% of its incoming support tickets are now deflected by AI, illustrating that productivity gains from AI deployment can increase the volume of work serviceable without adding head count, a dynamic that supports consumption revenue rather than compressing it.

Active customer accounts grew to approximately 402,000 from approximately 325,000 through 2025, while the dollar-based net expansion rate improved to 108% in fourth-quarter 2025 from 104% in the prior-year period, consistent with AI agent deployment acting as a consumption tailwind rather than a headwind.

Enterprise customers typically operate under annual or multiyear committed use agreements that provide some renewal friction, though the structure differs meaningfully from seat-based contracts. Twilio’s customers can reduce usage or redirect traffic to a competing provider incrementally without a formal vendor replacement decision, making switching optionality greater than in a traditional software-as-a-service renewal dynamic.

The more relevant AI risk is not displacement of existing customers but longer-term pricing pressure at the infrastructure layer. AI is reducing the cost and complexity of building competing communications APIs, and the emergence of vertical AI agent platforms creates a potential disintermediation risk as application-layer vendors consider building natively rather than routing through Twilio.

Management’s core counterargument is that its super network of nearly 5,000 unique carrier connections across over 180 countries, built over nearly two decades, is not replicable in software, and that as AI startups scale, they increasingly choose to buy rather than build given the complexity of telecom infrastructure, compliance and fraud prevention.

Critically, AI voice orchestration also appears to be expanding rather than compressing Twilio’s per-unit economics, with its ConversationRelay product priced at a meaningful premium to standard programmable voice, reflecting the greater infrastructure complexity involved in real-time AI voice interactions. These factors provide meaningful protection against acute market-share erosion, but the volume-over-price dynamic that has supported revenue growth to date requires monitoring as the competitive infrastructure layer evolves.

Companies that provide data as a service face materially different levels of AI displacement exposure depending on the proprietary nature of their underlying data. Where a service is built on the aggregation of widely available public information, AI tools are increasingly capable of replicating the core value proposition at a fraction of the cost, placing significant pricing pressure on those businesses as contracts come up for renewal. The competitive moat in these cases was never the data itself but the operational excellence of maintaining it, a capability that is rapidly being commoditized.

Services built around genuinely proprietary databases occupy a more defensible position. These datasets cannot be re-created by AI tools trained on public information, and in regulated fields such as pharmaceuticals, patent law and academic research, the provenance and auditability of the underlying source carries legal and compliance value that synthetic alternatives cannot replicate.

However, even companies with strong data moats are not fully insulated. The threat manifests not at the data layer but at the intelligence layer above it – the analytics and workflow products that have historically driven upsell revenue and margin expansion. As foundation models become capable of performing sophisticated analysis on connected datasets, customers can increasingly replicate these capabilities themselves, compressing willingness to pay for premium tiers without abandoning the base subscription.

Data-as-a-service providers are responding by embedding AI-native features into their offerings, though sustainable differentiation ultimately depends less on AI capability, which is accessible to customers and competitors alike, and more on the depth of data integration and switching costs that can be built around it.

ZoomInfo (Continued)

ZoomInfo’s data assets, built over two decades through user contributions, third-party publisher agreements and proprietary collection infrastructure, have historically commanded a quality premium over competitors including Apollo and Seamless, though that gap may be narrowing as competitors invest in AI-powered validation and enrichment, compressing ZoomInfo’s ability to justify its price differential. The company has already suffered significant churn in the downmarket segment in recent years, which management has characterized as partly strategic, though whether that framing reflects deliberate repositioning or competitive displacement is difficult to assess from public disclosures.

The intent product, which identifies in-market buyers based on content consumption across ZoomInfo’s publisher network, is the most defensible portion of the company’s product offering given that much of its core contact data is derived from publicly available sources. Its primary risk is not direct replication by competitors but architectural obsolescence; the product assumes B2B purchase research happens through publisher content, an assumption that becomes less representative as buyers increasingly conduct research through AI tools that generate no signal within ZoomInfo’s network. Unlike the contact database quality gap, which ZoomInfo can address through continued data investment, this risk cannot be engineered away because it reflects a change in buyer behavior rather than a gap in ZoomInfo’s own capabilities.

Overall NRR declined materially to 87% in both 2023 and 2024 before partially recovering to 90% in 2025, driven primarily by contraction in the small and medium-size business and mid-market segments. The upmarket segment has shown relative resilience with retention metrics closer to 100%, consistent with ZoomInfo’s data quality advantage and entrenchment in large enterprise workflows. Workflow orchestration platforms, however, including Clay, list ZoomInfo alongside competing data providers as interchangeable integrations, which, to the extent enterprises adopt these platforms, shifts ZoomInfo from a workflow system with embedded switching costs to a commoditized data feed, a dynamic that may not yet be visible in upmarket retention metrics but has the potential to weaken renewal pricing power at future contract cycles regardless of underlying data quality.

Clarivate

Clarivate’s data assets occupy a stronger competitive position than most data-as-a-service peers, though the degree of protection varies across its portfolio. Its most defensible assets include Derwent patent records and the Web of Science citation index, which represent decades of expert human curation that cannot be reconstructed through AI tools.

Patent records are legal primary source documents by definition, and the Web of Science citation linkages built over 60 years carry academic authority that synthetic alternatives cannot replicate. Customers in patent law and academic research are also often unable to substitute these sources in many cases due to regulatory, compliance and reputational requirements. Management claims 97% of revenue is derived from proprietary solutions, though this figure reflects the company’s own classification across both data assets and the workflow tools built on top of them and likely overstates the irreplaceability of the broader portfolio.

Clarivate is responding to the AI competitive threat by embedding AI capabilities across its product suite. Core AI features including Smart Search and the Web of Science Research Assistant have been bundled into existing subscriptions as defensive retention tools, while separately priced higher-tier products, including Web of Science Research Intelligence and Derwent Patent Monitor, both launched in the second half of 2025, represent the genuine upsell opportunity.

The structural challenge facing both tiers is the same: The AI capabilities Clarivate is building rely on the same foundation models accessible to its customers, meaning the analytics and workflow layer above the core data is increasingly replicable by sophisticated customers connecting foundation models directly to Clarivate’s own data exports. The base data subscription remains protected by genuine proprietary moats, but the intelligence layer above it faces displacement pressure that the bifurcated AI strategy addresses only partially.

The companies in this section are defensible for a different reason than the others discussed in this piece. Their durability comes from being the established standard inside regulated workflows, where switching requires an entire industry to move at once rather than a single customer making a vendor replacement decision. The collective cost of that kind of migration almost always exceeds the benefit of any available alternative, regardless of how good that alternative is.

AI can replicate what these companies do technically, and in some cases already has, but replicating the regulatory approval, legal authority and institutional trust that make these standards the required reference point is a problem technology alone cannot solve. For credit investors, this means the displacement timeline is long enough that the more relevant risks within the bond maturity horizon are competitive and regulatory rather than technological.

Fair Isaac

Fair Isaac has a considerable moat in that its FICO score is an entrenched industry standard, affecting the large majority of U.S. consumer-level credit decisions by most major banks, credit card issuers and mortgage lenders. There have been decades of lender underwriting models, regulatory frameworks and investment decisions built around the FICO score as the common reference point, creating significant switching costs.

The score’s resistance to AI displacement is institutional and regulatory rather than technical. The argument is not that the scoring methodology is irreplicable, as machine learning can and already does produce comparably predictive credit models. The moat operates through mechanisms AI cannot shortcut.

Consumer credit laws require lenders to provide specific explainable denial reasons, a requirement FICO satisfies through decades of legally validated reason codes that AI models cannot replicate without extensive regulatory reengineering.

Regulatory approval of a new scoring model for conforming mortgage use requires multiyear historical validation across economic cycles, a process that cannot be compressed regardless of a model’s technical quality. Secondary market investors also have decades of institutional experience pricing mortgage-backed security pools against FICO score distributions.Together, these mechanisms mean that AI’s ability to replicate FICO’s scoring methodology does not translate into a near-term displacement threat, as the switching costs are structural rather than product-driven.

The more credible near-term competitive threat is VantageScore rather than AI directly. The FHFA’s July 2025 decision to allow lenders to choose between VantageScore 4.0 and classic FICO when selling loans to Fannie Mae and Freddie Mac ended decades of mandatory FICO use in the conforming mortgage market, implementing a legislative mandate that had been in process since the Credit Score Competition Act was passed in 2018.

FICO’s pattern of aggressive royalty increases in the years surrounding this transition, including a 41% mortgage origination price increase for 2025, compounded lender frustration and reinforced the political case for introducing competition, while simultaneously driving strong near-term revenue growth.

VantageScore benefits from a structural distribution advantage through its joint ownership by the three major credit bureaus reporting agencies, Equifax, Experian, and Transunion, which control both the underlying consumer credit data and the distribution channels through which both scoring models reach lenders.

Equifax has further reduced the per-score cost barrier by offering VantageScore at no cost through 2026 to customers who also purchase FICO scores, though lenders still bear the more significant operational and compliance costs of transitioning underwriting models and revalidating loan pricing against a new scoring standard. That operational burden, combined with the depth of institutional entrenchment across the broader lending market outside the government sponsored enterprise conforming segment, limits the pace at which VantageScore adoption can realistically accelerate and preserves meaningful defensibility over the near-to-medium term.

Consensus Cloud

Consensus Cloud, which is a healthcare-focused digital document exchange company best known for its eFax brand, is similarly entrenched within the industries it operates in driven by regulatory and compliance requirements. Healthcare organizations transmitting prior authorizations, clinical documentation, and referrals are subject to strict security and auditability requirements for protected health information that digital cloud fax satisfies through decades of established regulatory acceptance.

Replacing fax with a more capable alternative requires re-validating the compliance posture of the replacement technology across an organization’s entire clinical workflow, a process sufficiently expensive and slow that the switching cost persists independent of how technically capable alternatives become.

The more credible medium-term risk is secular volume decline rather than direct AI displacement. Fast healthcare interoperability resource-based data exchange platforms and broader electronic health record interoperability investments are gradually enabling healthcare organizations to route clinical data through structured digital channels, reducing the fax workflows that sustain Consensus Cloud’s volumes over time. The SoHo segment, which serves small office and home office customers, faces a more immediate version of this pressure and is already in structural decline.

The company’s primary defensive response is Clarity, an AI-powered data extraction product that converts unstructured fax documents into structured healthcare data formats, including HL7 and FHIR, repositioning Consensus Cloud from a transmission infrastructure provider toward a data interoperability platform. The Veterans Affairs contract, which is expanding across facilities, with management guiding the opportunity toward $10 million to $20 million over two to three years, provides the clearest proof of concept for this model. Overall, its corporate revenue retention has shown resilience to any secondary pressures, with retention at approximately 101% in fourth-quarter 2025, improving slightly versus the prior-year quarter.