Achieving Artificial Intelligence Return on Investment (ROI) through Value Co-Creation

Generative Artificial Intelligence (GenAI) has moved rapidly from technological curiosity to strategic priority in banking. Since 2023, institutions have launched numerous pilots, from internal chat assistants to automated document analysis. While these projects show impressive capabilities, many struggle to turn experimentation into measurable economic value. This gap reflects not technological immaturity, but misalignment with regulatory, organisational, and contextual realities. As hype fades, pressure grows to justify investments with tangible returns. This paper argues that 2026 represents a critical inflection point: only institutions demonstrating measurable ROI from GenAI will succeed in embedding it into core financial processes, while others will remain confined to isolated proofs of concept. Success depends on contextualised applications, robust controls, and value co-creation with specialised software providers.

1. Looking Back | Lessons from the Hype

1.1. From Exploration to Economic Accountability      

The rapid adoption of GenAI in banking closely followed the classic technology hype cycle, characterised by inflated expectations and extensive experimentation [cf. Gartner 09.2025]. Early use cases focused primarily on horizontal productivity tools, such as chat interfaces, code generation, or text summarisation. While these applications delivered incremental efficiency gains, they rarely addressed core financial decision-making processes. At the same time, regulatory scrutiny and internal governance requirements limited the scalability of many pilots [cf. ECB 2024]. As a result, banks now face a growing recognition that GenAI initiatives must demonstrate clear economic value rather than technological novelty.

1.2. Why Context Determines Success or Failure

A key lesson from early GenAI deployments we have seen is the central importance of context. As shown in Figure 01, general-purpose language models excel at generating plausible outputs, but plausibility alone is insufficient in regulated financial environments.

Risk management, finance, and treasury functions rely on precise semantics, well-defined assumptions, and traceable data lineage. Without embedding these elements, GenAI systems risk producing outputs that are difficult to validate and explain, increasing operational and model risk [cf. BIS 2025]. Research consistently shows that domain-specific AI systems outperform generic models in regulated industries, particularly where decision accuracy and explainability are critical [cf. Liu et al. 2024].

2. Looking Ahead | A Value Co-Creation Framework

To move from experimentation to sustainable value creation, this paper proposes a framework in which ROI from Generative AI emerges at the intersection of Context, Control, and Collaboration.

As illustrated in Figure 02, an overview of the proposed C⁶R framework, ROI from Generative AI does not result from individual technological capabilities, but emerges at the intersection of Context, Control, and Collaboration, with credibility, capability, and continuity arising as secondary effects.

---

2.1. Context: Embedding GenAI in Financial Reality

Context represents the degree to which GenAI applications are grounded in the financial, regulatory, and methodological realities of banking, and – more specifically – of the bank using the system. This includes domain-specific data structures, risk semantics, explicit modelling assumptions, and institutional knowledge accumulated over time. In regulated environments, outputs that are merely plausible are insufficient; they must be interpretable within established financial logic. When GenAI systems are contextualised in this way, by reusing available risk data for example, they support expert decision-making rather than replace it, reducing validation effort and increasing adoption [cf. Liu et al. 2024; cf. McKinsey 2023].

2.2. Control: Governance as an Enabler of Value

Control refers to the mechanisms that ensure GenAI systems operate within clearly defined boundaries. These mechanisms include human-in-the-loop decision structures, explainability, auditability, and clear accountability for outputs. In banking, such controls are often perceived as constraints on innovation. However, experience from early GenAI deployments shows that the absence of adequate control leads to rework, delayed approvals, and heightened operational risk, ultimately eroding economic value [cf. ECB 2024]. Robust control therefore functions as an enabler of ROI rather than a barrier.

2.3. Collaboration: Value Co-Creation Instead of Procurement

Collaboration captures the shift from traditional software procurement toward value co-creation between banks and specialised software providers. GenAI solutions evolve through iterative use and refinement, requiring close interaction between domain experts, users, and technology providers. Banks contribute business ownership, regulatory expertise, and accountability for decisions, while software providers contribute domain data models, embedded financial logic, and AI engineering capabilities, as shown in Figure 03. This collaborative approach focuses GenAI investments on concrete high-impact use cases and shortens time-to-value [cf. Gartner 03.2025, HBR 2023].

2.4. Emergent Properties: Credibility, Capability, and Continuity

Within this framework, credibility, capability, and continuity are not treated as independent design objectives but as emergent properties. Credibility arises where contextualised GenAI operates under robust control, producing outputs that are trusted by experts and defensible in audits. Capability emerges at the intersection of control and collaboration, where governance structures and iterative development enable effective human – AI interaction. Continuity results from the combination of context and collaboration, embedding GenAI into core processes and ensuring that value creation persists beyond isolated pilots.

Conclusion

Generative AI is entering a decisive phase in the banking industry. As the hype cycle matures, the focus is shifting from experimentation towards sustainable value creation. In 2026, success will no longer be defined by the number of pilots launched, but by the ability to demonstrate measurable ROI and embed GenAI into core financial and risk processes. Institutions that fail to make this transition risk remaining trapped in isolated use cases with limited strategic impact.

The proposed C⁶R Framework highlights that economic success depends not on model sophistication alone, but on contextualisation, governance, and value co-creation between banks and specialised software providers. Where these dimensions converge, credibility, capability, and continuity arise as natural outcomes, enabling GenAI to scale beyond experimentation and become part of the operational infrastructure.

With the rise of requirements-driven development, building GenAI features is closer than ever to the fingertips of banking subject-matter experts, enabling solutions that align with real operation and risk management needs. In this environment, established partners and longstanding vendors play a critical role by providing scalable platforms, embedded domain knowledge, and delivery capabilities. Together with banks, they form the foundation for a co-creation model that turns GenAI from promise into performance.