In the “capability to profitability scale” spectrum, are banks using a sharper economic model for AI adoption?

Nilesh Sapar, VP of BFSI Delivery

The Ambitious Road of ‘ROAI’

ROAI or Return on AI investments is becoming one of the most important questions in banking (starkly the same for enterprises). 

In early 2025, the Reserve Bank of India (RBI), surveyed 612 regulated banks and NBFCs. Only 20.8% were actively deploying AI, while 67% were exploring or seeking to move toward AI use cases. That’s barely 1 in every 5 institution that had moved AI into active deployment while nearly 2 in 3 still exploring. Globally, IDC had forecast that spending on AI-centric systems would surpass $300 billion in 2026, while Deloitte’s State of AI in the Enterprise 2026 report shows the gap between ambition and value clearly: 74% of organizations hope to grow revenue through AI in the future, but only 20% say they are already doing so.

The gap between accelerating AI investment and slower value realization is widening every quarter. And that gap is forcing a harder question for banks whether they are using the right economic model to understand how AI value is created in the first place.

The ROI Trap in Banking AI

For years, tech investments in banking followed a familiar economic logic. Define the business case, approve the budget, implement the system, measure the return. The destination was visible before the journey began and that logic worked for many traditional technology programs because the path was largely known.

Today, AI changes that equation. AI value is rarely fully visible at the point of investment. It is discovered through use, strengthened through adoption, improved through data, and multiplied through scale. The first investment often creates capability before it creates profit. That makes traditional ROI models useful, but incomplete.

The danger is that banks may reject serious AI opportunities because they are being judged by financial models built for predictable technology projects. This is where the ROI conversation becomes too small. The issue is that banks may be measuring too early, too narrowly, and too close to the surface of work.

Why AI Value Appears Late

Most AI conversations still begin with technology and progress around models, copilots, automation, agents, platforms, productivity tools. But banking leaders know the harder truth that a model can work, and the business can remain unchanged.

Several factors come into play - legacy processes that resist change, users who do not yet trust the output, unclear decision rights, governance slowing adoption, data that is not ready, and compliance expectations that demand explainability the solution may not have been designed to provide. This is where AI ROI begins to break since the operating model was not ready to absorb it. 

The Six Connected Lens of AI ROI

AI economics should therefore be understood through six connected lenses: 

1. Realizing AI ROI progressively
2. Recognizing the penalty of inaction, 
3. Investing in AI capability, 
4. Measuring workforce productivity differently, 
5. Building AI-centric financial models, 
6. Achieving profitability with scale. 

Together, these lenses represent the economic progression through which AI capability is transformed into measurable business value.

The Progressive AI ROI 

The first lens is progressive ROI. AI adoption is not a single investment decision where complete value is visible upfront. Value is realized as organizations learn, innovate, implement, improve, and scale. The AI journey begins with the first step.

The Penalty of Inaction

Most organizations focus on the cost of AI adoption. The greater economic risk may be the cost of delayed learning. Every quarter of hesitation postpones more than deployment. It delays data maturity, workforce fluency, governance experience, business confidence, and the discovery of which use cases matter. In AI, learning has economic value.

The banks that move early do not only deploy earlier. They understand earlier. They learn where AI breaks, where users resist, where controls are needed, where data is weak, and where the business case is real. This is where the learning compounds.

This is the second lens: the AI inaction penalty. The late mover does not just start later, they start with less institutional memory. This does not mean banks should move recklessly. In financial services, caution has value. Trust, explainability, resilience, auditability, and accountability matter. But waiting for perfect readiness is built through disciplined work.

Capability - The New Economic Asset

The third lens is capability investment. Technology alone does not create AI value. Sustainable adoption requires investments in architecture, engineering, AI leadership, data foundations, model operations, governance guardrails, workforce fluency, partner ecosystems, and a culture of continuous learning. 

These investments may not always generate immediate returns, but they create the foundation upon which future AI value is built. Traditional ROI models often overcount the visible cost of capability and undercount the compounding value of readiness. 

This is where banks need to think more deeply about abstraction. The real question is not only, “Can AI automate this task?” It is, “What should this workflow become now that intelligence can be embedded into it?”

A loan application, compliance exception, customer escalation, risk review, or service request may still move through the same screens, approvals, queues, handoffs, and manual checks. AI may be inserted into one step and expected to produce transformation. But that is automation trapped inside old design.

The larger opportunity is to strip work down to its underlying structure: receive, verify, decide, approve, record, monitor, respond. Once the business flow is visible, leaders can decide where AI should assist, where humans must intervene, where accountability must remain explicit, and where the process itself can be redesigned. That is when AI starts to change the economics of work.

Workforce Productivity Economics

The fourth lens is workforce productivity economics. The easiest AI benefit to measure is time saved. That is why productivity dominates many AI business cases. But in banking, the stronger return may come from better judgment at scale.

A credit team that spots risk earlier creates value. A compliance team that reviews exceptions faster creates value. A service team that resolves issues before escalation creates value. A relationship manager who understands customer signals more clearly creates value. A delivery organization that plans faster when conditions change creates value. These are efficiency gains with improvements in decision quality. The strongest AI programs will remove the most friction from expert work.

The smartest AI investment is often talent made more capable by technology.

AI-Centric Financial Modeling

The fifth lens is AI-centric financial modeling. Most organizations still evaluate AI through traditional project-level ROI. That approach often underestimates value because AI benefits are distributed across operations, customer experience, compliance, risk management, productivity, innovation velocity, and organizational maturity.

Banks therefore need to manage AI less like a collection of disconnected business cases and more like a value portfolio. Some use cases should deliver near-term productivity, some should reduce operational risk, some should improve customer experience while some should strengthen compliance and auditability.

A serious AI portfolio needs discipline. Every initiative should have a business metric, a baseline, a workflow connection, an adoption owner, and a clear view of what will be measured after launch. Without that discipline, AI remains activity. With it, AI becomes an economic system.

Profitability with Scale

The sixth lens is profitability with scale. Individual AI use cases can generate measurable benefits. But the greatest economic value emerges when AI scales across functions, processes, controls, and customer journeys.

From ROI to ROAI

As adoption expands, productivity gains multiply, operational costs decline, decision quality improves, governance becomes repeatable, platforms become reusable, and business value compounds. That is when AI stops looking like a technology spend.

The winners in banking will be the institutions that continuously learn, govern, improve, and scale faster than their competitors. True winners in banks and institutions will be those who are building the organizational ability to realize AI value where cost appears first, capability develops next, value follows progressively, profitability emerges with scale. That is the new economics of AI adoption.