The Hidden Layer of Innovation: How Shadow Integrations Between Corporates and Startups Reshape Business Models, Tech Stacks, and UX

Abstract

Corporate–startup collaboration is often narrated through visible partnerships, acquisitions, or corporate venture capital. Yet an increasingly important share of innovation now happens in a hidden layer: shadow integrations. In this paper, shadow integrations refer to intentional but low‑visibility, technology‑driven arrangements where corporates embed startup capabilities—APIs, scoring engines, AI models, logistics and fintech infrastructure—without presenting them as a public alliance to end users. This hidden layer intersects with another phenomenon: shadow IT and shadow AI, where teams adopt unapproved tools that bypass central governance, increasing both innovation and risk. The rise of such hidden connections is reshaping how value is created and captured, how tech stacks evolve, and how user experience (UX) is perceived, even when interfaces and brands look unchanged.

Drawing on recent evidence on shadow AI, shadow IT governance, and regulatory pressures around data sharing and transparency [1][2][3][4], this paper develops a three‑layer comparative framework—business model, technology/architecture, and UX—and applies it across fintech, retail/e‑commerce, and digital health. It then proposes a typology of shadow integrations, analyzes power dynamics and strategic risks, and concludes with actionable recommendations for both corporates and startups. Ultimately, the boundary that matters most is no longer “corporate vs. startup,” but “visible vs. invisible” innovation.

Background

Over the past decade, corporate–startup relations have professionalized through accelerators, innovation labs, and corporate venture capital. Yet much of the public discourse still treats incumbents and startups as opposing camps: the former as slow, regulated, burdened by legacy systems; the latter as fast, experimental, “disruptive.” This framing obscures a critical reality: a large share of startup technology now reaches markets through incumbents’ channels, often without visible branding or public partnership announcements.

At the same time, enterprises are grappling with an explosion of shadow IT and shadow AI—tools, apps, and models adopted outside formal procurement or governance. In some sectors, the prevalence of shadow AI has surged by 250%, with nearly half of agents admitting to using unauthorized AI tools [1]. These tools promise productivity gains but introduce security, compliance, and data governance risks that can erode trust if not managed. Research indicates that 84% of consumers already assume brands use their data to train AI systems and prefer honesty about it [2]. When organizations lack visibility into what is actually running in their environment, they risk misinforming customers and regulators.

Shadow integrations, as defined in this paper, sit at the intersection of formal and informal innovation. They are formal in the sense that they rely on contracts, APIs, SLAs, and support agreements between corporates and startups. They are informal at the UX and brand level because the relationship is either invisible to users or disclosed only minimally. This differs from purely unauthorized shadow IT, but many of the same governance and risk issues apply. Case evidence from enterprises such as Steelcase—which discovered numerous unsanctioned cloud applications and responded by auditing, classifying, and in some cases blocking high‑risk services—shows the scale of hidden technology usage inside large organizations [3].

Regulation complicates this picture. Mandatory disclosure rules and complex data‑sharing regimes can simultaneously incentivize opacity (to protect proprietary techniques) and transparency (to comply with regulators and reassure customers). Smaller firms often struggle more with these requirements; they lack the internal compliance capacity that larger organizations enjoy, and can be disproportionately exposed when detailed operational information must be disclosed [4]. For corporates and startups building hidden integrations, the challenge is to achieve speed and differentiation while staying inside tightening legal and ethical boundaries.

Methods

This paper synthesizes insights from three bodies of evidence: empirical reporting on shadow AI and shadow IT, case‑oriented documentation on enterprise IT governance, and conceptual and regulatory analyses related to data sharing and startup collaboration.

First, recent reporting on customer experience and AI adoption quantifies the growth of unapproved AI tools in sensitive industries such as finance and healthcare. For example, industry analysis points to a 250% surge in shadow AI usage and notes that almost half of frontline agents admit using unauthorized AI tools [1]. Complementary surveys indicate that 84% of consumers believe brands already use their data to train AI and would prefer transparent communication about this usage [2]. These data points ground the discussion of risk, trust, and user perception.

Second, enterprise case studies on shadow IT, including Steelcase’s experience with cloud services sprawl, illustrate how organizations discover and govern hidden technologies. By conducting audits, classifying services by risk, and blocking or integrating them through official channels, Steelcase and similar organizations show practical governance responses [3]. These cases inform our discussion of how corporates can evolve toward managed, strategic shadow integrations instead of unmanaged tool sprawl.

Third, legal and academic sources examine the regulatory environment for startup collaboration, mandatory disclosure, data sharing, and privacy. They highlight how smaller firms can be disadvantaged by information‑intensive regulations and how technical approaches such as federated learning can mitigate privacy and IP concerns when collaborating on AI models across organizational boundaries [4][5]. These sources underpin the analysis of power dynamics, contractual design, and data governance in corporate–startup integration.

Taken together, these materials are interpreted through a comparative, three‑layer framework—business model, technology/architecture, and UX—and applied across several industries. Where specific named examples are not public, the paper uses realistic composite scenarios grounded in the documented patterns above.

Key Findings

1. Business Model Layer: From Product Silos to Capability Markets

At the business model level, shadow integrations are accelerating a shift from end‑to‑end product ownership toward modular “capability markets.” Instead of owning every component—from risk models to logistics optimization—corporates increasingly rent critical capabilities from specialized startups. This approach reshapes cost structures: fixed R&D and capex are partially replaced by variable, usage‑based fees, revenue‑sharing arrangements, or subscription models.

In fintech, for example, a bank that once invested heavily in building an internal credit scoring engine may now license a startup’s machine‑learning risk model via API. The bank keeps control of pricing, branding, and distribution, but a portion of value creation migrates upstream to the startup, which monetizes the same engine across multiple institutions. This changes bargaining power: if several banks rely on the same scoring API, the startup becomes a quasi‑utility with potential network effects, yet it is still at risk of being treated as a commoditized supplier. Meanwhile, the bank accelerates its time‑to‑market and can pilot new lending products quickly, but may become strategically dependent on a third party.

Shadow integrations also influence how revenue is shared and how value is captured over time. In retail, a large omnichannel retailer may pay a startup for AI‑driven recommendation services that quietly raise conversion rates by a few percentage points. Even a 1–2% uplift in conversion at scale can translate into millions of dollars annually, justifying a revenue share or performance‑based pricing model. However, when such economic impact is invisible to users and even to some internal stakeholders, it can be underappreciated in corporate strategy debates, leading to underinvestment or overly aggressive price negotiations that threaten the startup’s sustainability.

2. Technology and Architecture Layer: From Monoliths to Managed Heterogeneity

Technologically, shadow integrations formalize what shadow IT has already signaled: organizations are moving from monolithic systems to a patchwork of microservices, APIs, and cloud platforms. The difference is that these integrations are now increasingly strategic, even when they remain invisible to customers. Yet many of the operational challenges observed in unmanaged shadow IT—data silos, incompatible schemas, security gaps—remain relevant.

The Steelcase case provides a warning. The company discovered numerous unsanctioned cloud services in use across departments, reflecting employees’ desire for better tools but also increasing risk exposure [3]. By auditing and categorizing services, Steelcase was able to block high‑risk applications and integrate others into its official stack. Shadow integrations between corporates and startups face a similar architectural tension: they must be modular enough to plug into legacy cores, yet governed enough to avoid fragmentation and uncontrolled risk.

Hybrid stacks—where a legacy core coexists with startup‑provided microservices—are becoming standard in banking, retail, and healthcare. APIs turn once‑rigid systems into platforms, but they also increase the attack surface. In sectors handling sensitive data, the rapid 250% growth in shadow AI [1] has already stressed governance frameworks. Corporates must ensure that embedded startup capabilities meet or exceed internal security standards and that data flows (training data, inference logs, telemetry) are governed to prevent unintended data leakage or IP loss.

3. User Experience Layer: Perceived Continuity vs. Hidden Discontinuity

At the UX layer, the paradox of shadow integrations is that they are designed to be invisible. From the user’s point of view, a bank app that suddenly offers instant credit decisions or a hospital portal with seamless telemedicine appointments may appear simply as “the brand improving.” Yet underneath, different engines—scoring algorithms, triage bots, recommendation systems—may be operated by specialized startups.

User perception is crucial here. Surveys suggest that 84% of consumers already believe brands use their data to train AI and prefer honesty about it [2]. If users later discover that their data has flowed through multiple unseen vendors, trust can erode quickly, particularly if there are incidents such as data breaches or biased algorithmic outcomes. Organizations that downplay or do not fully understand their own integration landscape risk making inaccurate statements to customers about where and how data is processed.

The UX gains from shadow integrations are real: faster onboarding, more relevant recommendations, smoother logistics, and lower perceived friction. Yet these gains sit alongside hidden discontinuities. A retailer’s website may feel modern and responsive, but customer support may still rely on older systems that are not integrated with the startup‑powered recommendation engine, creating inconsistent experiences. Similarly, in healthcare, patients may enjoy convenient online booking powered by a startup, while medical records remain locked in legacy EHR systems. When disconnects occur—conflicting appointment information, lost messages, or inconsistent personalization—the invisible nature of the integration can make root causes hard to diagnose for both users and internal teams.

4. Sector Lens: Fintech, Retail, and Digital Health

Fintech and Traditional Banking

In banking, the contrast between legacy cores and fintech agility is well established, but the most significant changes often occur behind the scenes. Traditional banks are increasingly buying credit scoring, fraud detection, KYC, and payment routing as services from startups. While the public narrative may speak of a “new proprietary engine,” the underlying capability is often licensed.

At the business model level, this enables banks to shift more of their IT spend from capital expenditure to operational expenditure. Instead of multi‑year projects to build systems that may be obsolete on delivery, banks can pay per API call or per account. This also allows them to experiment with new products—such as buy‑now‑pay‑later or contextual credit—without overhauling their entire infrastructure. However, the strategic dependency created by such integrations is non‑trivial. If a core lending product depends on a single external scoring provider, negotiating power can flip over time, and exit costs—migrating models, revalidating risk frameworks—can be high.

Architecturally, fintech shadow integrations embody “managed heterogeneity”: legacy mainframes or core banking platforms wrapped with API layers that route requests to startup services. This cohabitation introduces resilience benefits, as specialized providers often maintain state‑of‑the‑art security, but it also creates integration complexity and compliance questions. If the startup’s AI models learn from customer data, issues of model explainability, auditability, and data residency arise, particularly under regimes modeled on HIPAA/GDPR‑style constraints [5]. UX‑wise, customers experience faster approvals, richer mobile apps, and contextual offers, but still largely attribute these to their bank, not the underlying fintech suppliers.

Retail and E‑commerce

Retailers, especially those with large physical footprints, increasingly rely on AI‑driven recommendation engines, demand forecasting, inventory optimization, and last‑mile logistics platforms provided by startups. These capabilities are often integrated invisibly into online storefronts and order management systems.

The business model impact is multidimensional. Shadow integrations allow retailers to expand into new offerings such as subscriptions, memberships, or same‑day delivery without building full in‑house capabilities. A third‑party logistics startup can manage last‑mile routing and capacity utilization, while the retailer keeps the direct customer relationship. Margins can improve not only through cost efficiencies but also through higher basket sizes driven by more relevant recommendations. However, reliance on external algorithms can entrench opaque dependencies; if the recommendation vendor changes pricing or is acquired by a competitor, re‑platforming can be disruptive.

Technologically, retailers are moving from monolithic commerce platforms to composable architectures built around microservices. Startups supply specialized building blocks—search and discovery, personalization, pricing optimization—accessed via APIs. This modularity helps experimentation, but echoes the risks seen in shadow IT reports: when different departments adopt overlapping tools without central coordination, data silos and duplicated functionality arise [3]. From the UX perspective, customers simply experience smoother search, faster checkout, and better delivery estimates, associating the positive change with the retailer’s brand. Yet when something fails—such as misrouted deliveries or inconsistent promotion logic—customers rarely understand that a startup provider sits in the middle of the chain.

Digital Health and Traditional Healthcare Systems

In healthcare, the tension between trust in established institutions and innovation from startups is acute. Hospitals and insurers are integrating telemedicine platforms, remote patient monitoring tools, online booking systems, and AI‑assisted diagnostics from startups directly into their portals and clinical workflows. Patients often interact only with the familiar hospital or insurer brand, unaware that a third‑party startup is powering the experience.

The business model implications are profound. Many health systems are experimenting with outcome‑based contracts, population‑health management, and subscription‑like services for chronic disease management. Startups provide monitoring and analytics capabilities under licensing or SaaS models, enabling providers to move beyond fee‑for‑service. Yet regulatory and reimbursement frameworks can lag behind, and complex data‑sharing arrangements raise privacy and cybersecurity concerns [5]. The startup may technically process highly sensitive medical data, even if the patient never sees its name.

From a technology standpoint, interoperability remains a central challenge. Integrations must bridge modern APIs and mobile apps with entrenched EHR/EMR systems designed decades ago. Approaches like federated learning and domain adaptation have been proposed to enable cross‑company AI collaboration without direct data sharing, mitigating privacy and IP risks [5]. In practice, hybrid models emerge: some data remains on‑premise; models are trained or fine‑tuned locally; and only gradients or anonymized aggregates are exchanged. UX benefits—shorter wait times, remote consultations, better continuity of care—are substantial, but if patients later learn that multiple unseen vendors handled their data, trust may be tested, especially in jurisdictions with strong expectations of confidentiality.

5. Quantitative Snapshot and Drivers

Although robust, global statistics on deliberate shadow integrations are still scarce, the data on shadow AI and shadow IT offer proxies for the scale of hidden technology dependence. The 250% surge in shadow AI usage and nearly 50% of agents using unauthorized AI tools in sensitive sectors such as finance and healthcare [1] suggest a structural demand for faster, more capable systems than official stacks provide. Simultaneously, Steelcase’s experience with unsanctioned cloud services underscores how quickly tool usage can proliferate when governance lags [3].

The driver is not simply “startups are faster”; it is a structural misalignment between the cadence of corporate IT governance and the pace of technological advancement. Regulatory complexity also shapes the landscape. Research shows that detailed mandatory disclosure regimes can disadvantage smaller firms that lack sophisticated information systems, because they must reveal sensitive operational information that larger competitors can exploit [4]. This incentivizes some corporates and startups to keep integrations quiet to avoid additional disclosure triggers, deepening the hidden nature of these arrangements.

The table below summarizes selected quantified drivers of hidden technology usage.

These figures, though not specific to white‑label startup integrations, provide a quantitative backdrop: end users and employees both operate under the assumption that sophisticated, often opaque technologies are in play, and they increasingly expect transparency and governance rather than mere denial.

Comparative Analysis

Business Model Trade‑offs Across Sectors

Across fintech, retail, and digital health, shadow integrations reveal a consistent business model pattern: incumbents externalize specialized innovation while internalizing customer relationships and regulatory accountability. In banking, licensing a startup’s scoring engine transforms a cost center into a variable partnership expense while preserving fee income on loans. In retail, outsourcing last‑mile logistics improves service breadth without building full logistics networks. In health, licensing telemedicine platforms enables new care models without reinventing core clinical systems.

The trade‑off lies in strategic dependency and bargaining power. In sectors with many comparable providers—such as generic recommendation engines—corporates can multi‑source and retain leverage. In narrow, heavily regulated niches—such as medical imaging AI with approved clinical validation—the startup may become hard to replace, yet still remains overshadowed at the brand level. This imbalance can discourage startups from investing in differentiated UX or building a direct brand, reinforcing their commoditization risk even as corporates quietly rely on them.

Technology and Architecture: Sectoral Convergence and Divergence

Technologically, all three sectors converge on hybrid architectures: legacy cores wrapped with API gateways and connected to cloud‑based startup services. However, the degree of standardization and interoperability varies. Fintech has benefited from relatively mature API standards for payments and open banking; retail has embraced composable commerce and headless architectures; healthcare remains fragmented, with heterogeneous EHR formats and stringent data localization rules.

Consequently, integration costs and risks differ. A bank integrating a new payment API can often reuse existing gateways and security modules, while a hospital integrating a new diagnostic AI tool may face bespoke interface work, local regulatory review, and complex data‑handling protocols. Shadow AI’s rapid growth [1] further complicates this, as employees may experiment with tools that are not yet properly integrated, creating disjointed flows. Sector‑specific regulations (HIPAA‑like health privacy, financial compliance regimes) push more rigorous integration testing, secure data handling, and auditable logs—raising the bar for startups but also making approved shadow integrations a defensible moat.

UX and Trust: Different Starting Points, Similar Tensions

User expectations around trust and transparency differ by sector, yet the underlying tension is similar. In finance and health, trust is central and heavily regulated; in retail, convenience and price often dominate. Still, the statistic that 84% of consumers assume brands use their data for AI [2] cuts across domains, signaling rising awareness of invisible technologies.

Banks and hospitals typically enjoy strong baseline trust but can suffer greater backlash if hidden integrations lead to perceived violations of privacy or fairness. A mispriced loan or a controversial diagnostic decision traced to an unseen algorithm can trigger regulatory scrutiny and reputational damage. Retailers may face less regulatory heat but more rapid customer churn if personalization feels creepy or if logistics failures—often rooted in integration glitches—erode reliability. In all cases, the invisibility of startup partners complicates attribution: users blame the visible brand, even when failures stem from third‑party systems.

The table below contrasts how shadow integrations shape UX and trust in the three sectors.

Case Studies

Case 1: A Regional Bank’s Invisible Credit Engine

A mid‑size regional bank sought to launch a new consumer lending product targeting thin‑file borrowers. Building an internal machine‑learning credit model would have demanded multi‑year investment, specialized talent, and extended regulatory validation. Instead, the bank integrated a startup’s credit scoring API. The partnership was structured as a white‑label, with no external communication about the startup.

Within six months, the bank launched the product and reported significantly faster approval times compared with its legacy scoring approach. Default rates remained within acceptable bounds, aided by continuous model updates provided by the startup. From a business model perspective, the bank converted a fixed R&D expense into a variable fee structure; the startup gained recurring revenue and access to a large data stream to refine its models. Architecturally, the integration added a microservice to the bank’s API layer, with data flows tightly governed to satisfy regulatory auditors.

UX outcomes were positive: customers perceived the bank as more modern and responsive, attributing the improved experience to the institution rather than to any partner. However, the bank’s risk committee later recognized a new concentration risk: if the startup changed pricing or strategy, or if regulatory scrutiny demanded greater model explainability, the bank would face high switching costs. This realization triggered a broader review of the bank’s shadow integrations and led to a more explicit vendor diversification strategy.

Case 2: An Omnichannel Retailer’s Hidden AI Stack

A large omnichannel retailer faced stagnating online conversion despite heavy investment in marketing. Rather than replacing its entire e‑commerce platform, the retailer layered in two startup solutions: a recommendation engine and a dynamic pricing service, both embedded via APIs and fully white‑labeled. No public announcement was made; the retailer presented the improvements as part of its ongoing digital transformation.

Over 12 months, the retailer observed a modest but meaningful uplift in average order value and conversion rates. The startups were compensated through a mix of subscription fees and upside‑linked payments tied to performance metrics. Internally, however, the retailer’s IT department struggled with a growing patchwork of external services, some adopted through formal channels, others emerging from past shadow IT experiments. The Steelcase‑style response—auditing services, scoring them by risk, and consolidating vendors—became necessary to manage complexity [3].

From the customer’s perspective, the experience simply felt “more like a modern e‑commerce site”: better product suggestions, more relevant discounts, and smoother checkout. Yet when an integration glitch temporarily misapplied promotions during a peak season, customers expressed frustration on social media, blaming the retailer. Investigating the issue required coordination across multiple partners. This incident led the retailer to formalize incident response protocols with its startup suppliers and to consider minimal co‑branding in future to manage expectations.

Case 3: A Health Insurer’s Telemedicine Pivot

A national health insurer wanted to reduce emergency room overuse and improve access to primary care. Building a proprietary telemedicine platform was deemed too slow and risky given regulatory constraints. Instead, the insurer integrated a startup’s telehealth and triage solution into its member portal under a complete white‑label arrangement. Patients booked “virtual visits with network doctors” without realizing that a third‑party startup platform powered scheduling, video, and initial symptom triage.

The integration enabled rapid rollout: within a year, telemedicine visits accounted for a meaningful share of primary care encounters, and ER utilization for non‑urgent conditions declined. The insurer paid the startup through a per‑consultation license and platform fee, while the startup benefited from steady volume and valuable real‑world usage data—handled under stringent privacy controls. Technically, the platform interfaced with multiple EHR systems across providers, highlighting the value of the startup’s interoperability capabilities.

UX improved markedly: shorter wait times, better continuity of care for chronic patients, and higher satisfaction scores among digitally savvy members. Yet questions emerged around data governance and informed consent. As consumer awareness of AI and data sharing grew—echoing the broader finding that most consumers assume their data is used in AI [2]—patient advocates questioned whether members should be informed about the involvement of third‑party platforms. In response, the insurer updated its privacy notices and began piloting a minimal co‑branding approach, listing core technology partners in accessible language without overwhelming users.

Limitations

The analysis in this paper is constrained by the limited availability of direct, quantitative data on deliberate shadow integrations between corporates and startups. Most public metrics relate to shadow IT and shadow AI more broadly, such as the 250% increase in unauthorized AI usage and the proportion of agents using unapproved tools [1]. While these figures illuminate the underlying appetite for hidden tools and the governance gap, they do not map one‑to‑one onto formally contracted but invisible partnerships.

Furthermore, many corporate–startup integration agreements include confidentiality clauses that restrict public disclosure of technical and commercial details. As a result, the case studies presented here are either composites or anonymized extrapolations grounded in known patterns rather than named contracts. This limits the ability to generalize precise financial outcomes, adoption timelines, or failure rates.

Regulatory environments are also evolving rapidly. The implications of frameworks such as the Corporate Transparency Act, new AI regulations, and sector‑specific privacy laws are still playing out in practice [4][5]. Their ultimate impact on the viability and design of shadow integrations—especially regarding disclosure obligations and data sharing—is uncertain. Finally, user sentiment data, such as the 84% of consumers who believe brands use their data for AI [2], is often collected in marketing or CX contexts and may not capture nuances across all regions or demographics.

Implications

Despite these limitations, the emerging picture has clear implications for strategy, governance, and innovation management. First, leaders should recognize that their competitive position increasingly depends on how effectively they orchestrate invisible capabilities rather than only on visible offerings. Mapping existing shadow integrations—both sanctioned and unsanctioned—is a prerequisite to understanding real dependency structures, innovation levers, and risk concentrations.

Second, governance frameworks must evolve from a binary stance on shadow IT (“block or ignore”) to a spectrum approach: discover, assess, integrate, or retire. Steelcase’s experience illustrates that systematic audits, risk scoring, and selective integration can harness innovation while mitigating exposure [3]. For startup partnerships, similar rigor is needed: clear SLAs, security requirements, data‑processing agreements, and exit strategies should accompany even white‑label arrangements. Technical measures such as federated learning can enable cross‑company AI collaborations without raw data exchange, aligning innovation goals with privacy and IP protection [5].

Third, user trust emerges as a strategic asset. With 84% of consumers already assuming AI‑driven data use [2], the question is less whether to disclose and more how to do so in a way that is intelligible and proportionate. Transparent but concise explanations of how data is processed and by whom, along with clear redress mechanisms for errors, can differentiate brands. Corporates that proactively manage the narrative around invisible technologies may enjoy a reputational advantage over those that only respond after incidents.

Conclusion

Shadow integrations—where corporates embed startup capabilities behind their own brands—are becoming a defining feature of modern innovation ecosystems. They blur the traditional boundary between “legacy incumbents” and “disruptive startups,” revealing instead a layered reality in which value creation, technology, and UX are co‑produced by networks of visible and invisible actors. In parallel, the documented rise of shadow AI and shadow IT underscores that hidden technology usage is not an exception but a structural response to gaps between formal systems and operational needs [1][3].

Across fintech, retail, and digital health, the three‑layer framework developed here—business model, technology/architecture, UX—highlights common patterns and sector‑specific variations. Business models shift toward capability markets and variable partnerships; architectures evolve into hybrid stacks of legacy cores and startup microservices; UX improves in speed and personalization but carries new trust and transparency challenges. Power imbalances and regulatory pressures complicate these dynamics, particularly for smaller startups constrained by disclosure and compliance burdens [4].

For leaders of corporates and startups alike, the actionable insight is clear: the most consequential frontier is no longer “corporate vs. startup,” but “visible vs. invisible” innovation. Strategic advantage will accrue to organizations that can (1) map their hidden integration landscape, (2) align it with explicit business and risk objectives, and (3) communicate its implications to regulators and users with honesty and clarity. In doing so, they can turn the hidden layer of innovation from a source of unmanaged risk into a deliberate, defensible competitive asset.

References

[1] Zendesk / Forbes – The Rise Of Shadow AI In CX: 4 Benefits To An Integrated Strategy – https://www.forbes.com/sites/zendesk/2024/11/04/the-rise-of-shadow-ai-in-cx-4-benefits-to-an-integrated-strategy/

[2] Modern Marketing Today – Shadow AI Is Real: Are Your Customers In The Dark? – https://modernmarketingtoday.com/shadow-ai-is-real-are-your-customers-in-the-dark/

[3] OneLogin – Getting A Handle On Shadow IT With Skyhigh And OneLogin (Steelcase Case Study) – https://www.onelogin.com/documents/getting-a-handle-on-shadow-it-with-skyhigh-and-onelogin-case-study-155581.pdf

[4] Columbia Business School / Forbes – Innovation At Every Size: Why Small Firms Struggle To Innovate In Today’s Regulatory Environments – https://www.forbes.com/sites/columbiabusinessschool/2024/07/22/innovation-at-every-size-why-small-firms-struggle-to-innovate-in-todays-regulatory-environments/

[5] Arxiv – Federated Learning with Domain Adaptation for Cross-Company AI Collaboration – https://arxiv.org/abs/2107.03912