The New Rules of Enterprise Technology Buying

To explore how enterprise buying decisions are evolving in the AI era, Opus Technologies Chief Revenue Officer Prakash Devulapalli sat down with industry analyst Ronak Doshi, Partner at Everest Group.

The discussion examines how AI is reshaping buying cycles, commercial models, software engineering in the banking, payments and fintech ecosystem, and what technology leaders should prioritize as they make investment decisions.

Prakash: AI has fundamentally changed the conversation around technology investments in banking, payments and the fintech ecosystem. How are enterprise buying priorities evolving in 2026?

Ronak: Enterprise buying priorities in 2026 are moving beyond generic AI experimentation toward a much tighter focus on business value, resilience, and scalable operating change. In banking, payments, and fintech, AI is increasingly being tied to modernization of core platforms, customer experience, fraud prevention, compliance, and enterprise productivity.  In payments specifically, I am seeing strong intensity around real-time scam prevention and merchant risk operations. The broader shift is from “AI as a pilot” to “AI as a lever for measurable operating and commercial outcomes.”

Buying priorities are evolving around re-evaluating build versus buy, embedded versus native AI stack development, and a holistic AI control tower for costs, outcomes, and governance.

Prakash: Buying cycles appear to be getting shorter in some areas and longer in others. What patterns are you seeing across enterprise technology decisions in the banking and payments ecosystem?

Ronak: As priorities shift toward measurable outcomes, buying cycles are becoming more distinct. On one side, banks and payments firms are moving faster on bounded, workflow-level AI decisions where the use case is clear, the data is available, and the governance model is manageable. In some of my recent conversations with banking and financial services (BFS) buyers, I have seen governed copilots in service and fraud moving into production in as little as six weeks when the business case is visible and execution risk is contained.

On the other side, buying cycles are lengthening for decisions tied to enterprise architecture, regulated decisioning, large-scale data modernization, and end-to-end operating-model redesign. Buyers are spending more time pressure-testing data foundations, model governance, auditability, integration complexity, and total cost impact before making bigger commitments. At the same time, expectations for faster ROI realization are compressing traditional deal durations and creating demand for more micro-transformations. Where buyers can quantify outcomes quickly, decisions are moving faster. Where value depends on broader transformation, they are requiring more evidence before committing at scale.

Adding to that, BFS buying is becoming multi-threaded, with several stakeholders weighing in, each with a more deliberate AI outcome orientation. Business and P&L owners increasingly sponsor AI, Chief Data & Analytics Officers (CDAOs) drive data investments, and risk leaders own regulatory priorities, while CIOs orchestrate the technology foundation. Buyers expect their suppliers to tailor their solution value proposition to each stakeholder group, and shape opportunities with deeper thought leadership.

Prakash: Traditional RFPs focused on scope, effort, and cost. In an AI-first world, what are enterprise buyers in the banking and payments world evaluating differently?

Ronak: Enterprise buyers are evaluating production readiness in their unique organization context much more rigorously than they did in traditional RFPs. Scope, effort, and cost still matter, but they are no longer sufficient. In an AI-first environment, buyers want proof that a provider can operationalize AI in their unique regulated enterprise setting, connect it to their real workflows, and sustain it under governance, audit, and risk requirements. In our recent research, we found that a little over half of enterprises are looking to integrate supplier’s AI-native solution in their environment. Buyers are evaluating suppliers on their ability to deliver pragmatic AI solutions that work in the real-world complexity of their organizations and provide a high degree of transparency on cost, outcomes, and quality controls.

They are also placing more weight on explainability, compliance readiness, and the provider’s ability to deliver an audit-ready operating model. In some of my recent conversations, I have explicitly seen “trust-first, responsible AI” as a lead theme, with emphasis on compliant, explainable, and audit-ready AI for regulated BFS environments.

Prakash: AI is changing not only the technology but also the commercial model for banking and payment services companies. How do you see deal structures evolving over the next few years?

Ronak: As buyers focus more on measurable outcomes, AI is shifting BFS deal structures from paying for effort to paying for value delivered. Traditional FTE- and transaction-based models will increasingly evolve toward hybrid constructs combining smaller human teams with AI agents and provider’s IP. I am seeing buyers prioritize value realization and governance over pricing flexibility, and expect greater adoption of outcome-based pricing, gain-share, AI/platform consumption fees, and committed productivity models tied to metrics such as cost-to-serve, release velocity, straight-through processing, and fraud reduction.

Buyers will increasingly expect AI-led productivity gains to be built into the commercial baseline, pushing providers beyond labor arbitrage toward technology- and IP-led value creation. Leading deals will likely combine a baseline managed-service fee, AI/platform consumption, and outcome-linked upside; aligning both parties around continuous productivity and reinvention. As AI takes over more structured work, buyers will have a higher degree of transparency in the commercial models and productivity commitments and are discarding black-box commercial models.

Prakash: AI agents are rapidly becoming part of the enterprise stack. Where do you see agents creating the greatest value, and where is the hype still ahead of reality?

Ronak: AI agents will create the greatest value where workflows are high volume, rules-informed, data rich, and operationally repetitive, but still benefit from judgment support and orchestration. That is why the strongest early momentum is in onboarding, know your customer (KYC) and perpetual KYC (pKYC) processing, underwriting, financial document analysis, servicing operations, and risk workflows. In our latest agentic AI research, I am seeing enterprises already moving several of these use cases from proof of concept into production, particularly across onboarding, compliance, and risk functions.

The hype is still ahead of reality in fully autonomous, high-stakes decision environments that require complex exception handling, legal defensibility, and strong human accountability. Areas such as autonomous credit decisions, broad multi-line orchestrations, and enterprise-wide agent swarms remain promising, but they demand much stronger orchestration, cross-functional ownership, and maturity control. So, the opportunity is real, but the value today is highest in governed, domain-specific agents with human in/on the loop, not in broad autonomous enterprise decisioning.

Prakash: AI is finding its way into almost every enterprise conversation. But when should banking, payments and fintech organizations deliberately say no to AI?

Ronak: Banking, payments, and fintech organizations should deliberately say no to AI when a workflow does not actually require non-deterministic technology. The starting point should be to break the workflow into meaningful components and determine the right technology for each one. Organizations can address many rules-based and predictable activities more effectively through workflow redesign, robotic process automation (RPA), analytics, traditional machine learning, or better systems integration. Only a minority of activities typically require the judgment, interpretation, or adaptability that generative or agentic AI provides.

Even within that smaller set, organizations should avoid defaulting to the most advanced models. Smaller, lower-cost, or open-weight models can address many AI-enabled tasks, while organizations should reserve frontier models for the narrow set of use cases where their additional capability materially improves the outcome. This creates a more disciplined architecture: use deterministic technology wherever possible, AI where necessary, and the most expensive AI only where it earns its place.

Prakash: AI is reshaping every phase of the Software Development Lifecycle. Which parts of the SDLC are seeing the greatest transformation, and where is human expertise still indispensable?

Ronak: The biggest SDLC transformation is happening in the parts of software engineering that are structured, repetitive, and documentation heavy. Code generation, test creation, test-cycle reduction, migration support, documentation, and engineering productivity are seeing the fastest impact. In some of my recent work, I’ve seen proof points such as 30-40% faster delivery through AI-assisted blueprinting and a 60% reduction in test cycles with Gen AI in core modernization contexts.

That said, human expertise remains indispensable in architecture, enterprise integration, regulated requirements, control design, cyber risk decisions, and any software change that could create customer harm or compliance exposure. In financial services, software engineering is not only about speed. It is about trust, resilience, and evidence. Human judgment is still essential to decide what should be automated, how much autonomy is acceptable, and where oversight must remain.

Prakash: Beyond technical capability, what qualities will differentiate technology partners in an AI-driven market?

Ronak: Technical capability is now table stakes. The partners that will truly differentiate in an AI-driven BFS market are the ones that combine domain knowledge, workflow understanding, enterprise integration capability, and governance discipline. Buyers increasingly want partners that can connect AI to operating-model change, data activation, decisioning intelligence, and process orchestration, because value comes from how the model reshapes execution in a controlled environment.

All my engagements with enterprises point to driving predictability of cost, outcomes, and controls that deliver trust in the AI solutions.. That means the strongest partners will be able to bring audit-ready compliance tooling, partner ecosystems, domain-specific accelerators, and a credible path from pilot to scaled adoption.

Prakash: What’s the biggest misconception enterprise leaders in banking and payments industry have about AI investments today?

Ronak: The biggest misconception is that access to strong models is the hard part. The real work sits in data, process, talent, and change management, not just technology. The harder part is turning AI capability into enterprise value through data readiness, workflow redesign, governance, talent, and adoption. Many leaders still assume that once technology is selected, value will follow quickly. In practice, most of the work begins after the model decision, especially in regulated sectors such as banking and payments.

I am seeing buyers increasingly fund AI in combination with secure platforms, data modernization, and operating model change rather than as standalone deployments. I am also seeing firms invest in scaling AI, not just experimenting, which reflects the need to build supporting enablers such as governance, talent, and infrastructure.

A related misconception is that AI success is mostly a technology decision. It is an operating-model decision and the leaders who understand that early are the ones most likely to move beyond pilots and capture durable returns.

Prakash: How are financial services clients thinking about compliance / auditability of AI. How do they handle a regulatory exam?

Ronak: In the US market, financial services firms are preparing for AI scrutiny through existing regulatory and supervisory exams rather than a single standalone AI exam. The key implication is practical: if an AI system influences a regulated decision or control point, regulators will examine it through the same lens they apply to the underlying business process, with added focus on explainability, traceability, governance, and human accountability.

Firms are responding in three ways. First, they are building AI asset inventories and risk classification frameworks to identify which use cases require heightened scrutiny, especially in credit, fraud, anti-money laundering (AML), KYC, and customer servicing. Second, they are investing in governance tooling that supports model lineage, audit logs, fairness testing, drift monitoring, and automated compliance reporting. Third, they are tightening controls around gen AI and agentic AI through prompt filtering, output controls, secure API use, and review mechanisms.

In the US, this plays out across existing exam structures. The SEC Division of Examinations and Financial Industry Regulatory Authority (FINRA) can review AI where it affects recommendations, communications, surveillance, AML, trade controls, and books and records. Bank regulators, including the Office of the Comptroller of the Currency (OCC), Federal Reserve, and Federal Deposit Insurance Corporation (FDIC), can examine AI through model risk, fair lending, consumer compliance, third-party risk, and operational resilience reviews. In insurance, state market conduct exams can scrutinize AI in underwriting, claims, and customer treatment.

A regulatory exam is becoming the real design test. Examiners are unlikely to accept broad statements about responsible AI. They will ask firms to reconstruct a specific decision: what data was used, which model or agent version was active, what controls were applied, who reviewed the output, and how the final action was approved. Firms that build this auditability by design will scale AI faster and with less regulatory friction.

Prakash: Looking ahead, what advice would you give technology leaders, buyers, and business leaders making technology investment decisions over the next 12–24 months?

Ronak: Over the next 12–24 months, technology and business leaders should be much more deliberate about where new technology is additive versus where it can replace or consolidate existing parts of the technology stack. The risk is that AI and other emerging technologies simply become another layer of spend on top of legacy platforms, point solutions, and fragmented tooling. Every major investment should therefore come with a clear view of what it can simplify, retire, or consolidate over time, alongside the value it creates. That will be critical to managing total technology costs and avoiding further sprawl and complexity.

At the same time, organizations need to address data, process, and talent debt in parallel with technology adoption. New technology cannot deliver its full potential if it is operating on fragmented data, layered onto inefficient processes, or deployed into an organization that has not redesigned roles, skills, incentives, and ways of working. New AI investment cycle is a good opportunity to simplify processes, strengthen data foundations, and build the capabilities required for employees to adopt new tools effectively.

This also means being more selective about where to invest. Rather than spreading capital across a large number of pilots, leaders should focus on a smaller set of high-value domains where there is a clear path from technology investment to business outcomes, measurable productivity or growth impact, and ultimately simplification of the broader operating and technology environment.

Ronak Doshi is a Partner at Everest Group, covering technology and the tech services industry, with a proven record of identifying emerging trends shaping the modernization and process-led transformation of enterprises early on. Ronak also leads the global banking, financial services, and insurance (BFSI) industry research.