Deepak Kinger, Chief Strategy Officer and Co-founder of EvoluteIQ – Interview Series

Deepak Kinger, Chief Strategy Officer and Co-founder of EvoluteIQ, is an experienced technology and business leader with more than two decades of experience spanning enterprise transformation, financial services, automation, and artificial intelligence. At EvoluteIQ, he helps shape the company’s overall strategy while leading key technology alliances, solution partnerships, marketing, and analyst relationships. Before co-founding EvoluteIQ in 2019, Kinger held senior leadership roles at GALE Partners, Virtusa, Sapient, Headstrong, Wipro, and Mphasis, where his responsibilities ranged from managing financial services businesses across Asia-Pacific and advising C-suite executives on digital transformation to building new business lines and leading large-scale technology engagements. His experience spans strategy, business development, client relationship management, P&L management, operations, and emerging technologies including AI, machine learning, robotic process automation, blockchain, and data science.

EvoluteIQ is an enterprise AI and intelligent automation company focused on helping organizations automate and orchestrate complex, end-to-end business processes. Its EIQ Platform brings together agentic AI, generative AI, process and decision automation, intelligent RPA, event processing, data fabric, analytics, enterprise connectors, and multichannel application development within a low-code/no-code environment. The platform is designed to enable enterprises to connect existing systems, build and deploy AI agents and automated workflows, monitor outcomes, and scale automation across areas such as customer experience, finance, operations, sales, and supply chain management. Founded in 2019, EvoluteIQ has expanded its focus toward agentic enterprise automation through technologies including its AI Workbench and Agentic Mesh Architecture, and in 2025 announced a $53 million minority growth investment led by Baird Capital to support global expansion and further investment in its Bengaluru R&D operations.

Before co-founding EvoluteIQ in 2019, you spent two decades working across financial services, enterprise technology, and digital transformation. What recurring problems did you see inside large organizations that ultimately convinced you there was an opportunity to build EvoluteIQ, and what did you want to approach differently?

Before EvoluteIQ, I spent 25 years in consulting, with much of it focused on helping large financial institutions on transformation initiatives. That included early digital banking programs as well as trading, risk management and broader end-to-end transformation. Over time, I saw growing demand for automation, and we were helping enterprises implement platforms from some of the largest vendors.

What ultimately motivated us to start EvoluteIQ was the opportunity to move from implementing somebody else’s technology to building something ourselves. We saw an opportunity for a more integrated approach to automation rather than requiring enterprises to bring together multiple technologies, licenses and integration layers. That became part of the motivation for starting EvoluteIQ in 2019.

In our previous Unite.AI interview with your co-founder Sameet Gupte, he discussed EvoluteIQ’s broader vision of moving enterprises from task-based automation toward AI-native, end-to-end autonomous operations. How does regulatory change management illustrate why that shift from automating individual tasks to orchestrating entire business processes is necessary?

Regulatory change rarely affects one isolated task. In 2020, we built a solution around the LIBOR transition. Banks were dealing with a large-scale change that affected existing contracts and required operations teams, paralegals and business teams to review and remediate them. Many institutions were essentially responding by throwing more people at the problem. The lesson we took from LIBOR was that the specific regulation changes, but the underlying challenge repeats. Regulatory change and compliance are ongoing operating requirements.

With the more recent FinCen proposed changes to Anti-Money Laundering/Countering the Financing of Terrorism (AML/CFT), we are seeing a move away from a highly prescriptive, “check-the-box” compliance model. This actually raises the bar on compliance rather than making it a lighter lift because institutions have to interpret the regulation, articulate their own risks and determine what those risks mean for their operating model. This shift inherently reaches different departments and functions across an organization, not just the compliance and audit teams.

You have argued that banks should stop treating each major regulatory change as a new transformation project and instead build regulatory change as a permanent operating capability. What would that operating model look like in practice, and why have financial institutions historically struggled to build it?

The new operating model would understand that change is continuous; it’s no longer a process that needs to be rolled out quarterly or when the auditors come knocking. Building a solution for LIBOR showed us that while one regulatory change may eventually be implemented, the need to adapt does not go away. Understanding the change, translating it into operational requirements and updating processes will be required for the next regulatory change.

The new operating model must also not be siloed within compliance. Under a more risk-based model, the compliance function may still ultimately own adherence, but operations, legal and, in some cases, counterparty risk management all become more important. One of the mistakes institutions can make is underestimating how broadly regulatory change affects the organization.

Historically, that has been difficult because enterprises have built up fragmented technology estates and siloed operating models. When different business units rely on separate RPA, BPM, reporting and other platforms, responding to a new requirement becomes an integration project in itself.

When a new regulation is introduced, a bank may need to determine which customers, products, contracts, systems, and processes are affected. Where can AI materially improve this impact analysis, and which parts of the process still require deterministic rules or human judgment?

AI can significantly reduce the amount of manual work required to understand and implement regulatory changes. More specifically, AI can help analyze and categorize large volumes of information and help zero in on where human focus should be concentrated. For example, EvoluteIQ’s regulatory response capabilities can categorize and prioritize contracts by client and asset type, connect that information with enterprise systems and external data sources, and then orchestrate the downstream work across legal, risk and operations. That gives institutions a way to move from identifying what may be affected to actually managing the resulting changes across the process.

Human control and oversight are still essential. There are parts of a regulated process where an institution needs clearly defined business rules, conditions, exceptions, and escalation paths. The model we believe in combines the adaptability of AI with deterministic automation and human judgment where appropriate. AI can help interpret and surface what may require action; deterministic rules can govern how defined decisions are executed; and exceptions or decisions requiring judgment can be escalated to the appropriate person.

Regulatory implementation often cuts across legal, compliance, risk, operations, data, and technology teams. How can AI and process orchestration help these groups work from a common interpretation of a regulation rather than creating separate workflows and potentially conflicting business rules?

One of the biggest implications of a more risk-based compliance model is that responsibility spreads further across the organization. Compliance may still be where accountability ultimately sits, but operations, legal and counterparty risk teams all have an important role. That means institutions have to avoid treating implementation as something the compliance function can execute in isolation.

This is why an integrated platform is key. In a large organization, different systems of record and business rules have to be integrated in this new model, but that also means taking into account security complexities. A unified approach can reduce the risk that different parts of the organization are implementing the same change through completely separate technology stacks.

Generative AI can interpret complex regulatory documents, but banks cannot afford hallucinations or inconsistent interpretations when compliance decisions are involved. What safeguards should be built around AI systems used for regulatory change, and where should human experts remain in the decision loop?

First, do not deploy AI agents in isolation, as they may create agent sprawl without proper governance.

Second, what an organization can automate is not just constrained by its technical capabilities, but also by regulation, privacy and what data can be legally shared.

The core design principle that needs to be followed is that AI should accelerate the work of interpretation without owning the decision. Accountability for compliance stays with named people, as regulators expect. Safeguards fall into two groups: controls on what the AI produces, and deliberate placement of human decision points. Every output AI produces needs to be grounded in an authoritative source with input and output guardrails set in place within the AI system.

Lastly, for every output, an audit trail of the source clauses, model and prompt versions, confidence, reviewer, decision and rationale should be maintained.

Auditability is especially important in financial services. If AI helps interpret a regulation, generate business rules, identify affected customers, and initiate workflows, how can institutions preserve a clear evidence trail showing regulators exactly how and why each decision was made?

Traceability has to go beyond documenting the final decision in highly regulated industries. Instead, it’s much more valuable to see a reconstructed path that led to that final decision by showing the information that was used, which rules were applied, the decisions that were made, changes made throughout the process, and where humans intervened and what they approved.

As multiple agents, business rules and workflows interact with each other, traceability becomes even more important. Auditability should be built into the orchestration layer. Decision trails can track decisions, rule changes and their rationale, while end-to-end agent traceability provides visibility across a chain of autonomous actions.

If a regulator asks why a particular customer, contract or transaction was handled a certain way, the financial institution should be able to trace that outcome back through the data, rules, AI-driven decisions, approvals and actions that produced it. That is what turns explainability from a model-level issue into an operational capability.

Large banks operate across decades of legacy infrastructure, with critical customer and transaction data spread across many systems. How can they build a reusable AI-driven compliance layer without attempting a costly replacement of the underlying technology estate? 

Any new enterprise capability has to operate within the existing infrastructure. It’s unreasonable to expect otherwise. Consolidation is also important because we increasingly hear from customers that they have multiple RPA platforms or a legacy BPM system with multiple, separate reporting technologies across the enterprise. Our customers are being more vocable about focusing on the tools and systems that generate ROI rather than finding a replacement. A reusable compliance layer must therefore work across the existing landscape rather than assume everything underneath it will be replaced.

You have worked with financial institutions through multiple generations of technology, from traditional core systems to robotic process automation, machine learning, and now agentic AI. What is fundamentally different about the current AI transition, and where do you think financial institutions are at risk of repeating mistakes from previous automation cycles?

The earlier generations of automation were for work that was already well defined. Core banking systems codified records and transactions. RPA scripted keystrokes across screens. Machine learning made narrow predictions, such as fraud scores and credit risk, on structured data with measurable accuracy. In every case a human or a fixed rule decided what happened next, and the technology carried out a known path.

Agentic AI changes that in a few ways:

• It handles unstructured, judgment-adjacent work more effectively. Banks are no longer just automating tasks. They are delegating outcomes, and the system chooses its own route.
• It fails quietly. When a bank changed a screen layout, an RPA bot broke loudly and someone fixed it. A language model that is wrong produces a fluent, plausible and confident answer. That is a problem.
• Its behavior is probabilistic and hence grounding it in enterprise context to avoid hallucinations is extremely critical.
• Finally, building using AI has become cheap and business users can now create working agents in an afternoon. Automation used to be bottlenecked by IT capacity; now the bottleneck is governance capacity.

Financial institutions should also be watchful that lessons learnt from the traditional automation era are not forgotten. Some examples of these include:

• Automating broken processes. RPA’s biggest legacy was thousands of bots layered over processes nobody had redesigned. That created brittle technical debt, which banks are still paying to maintain. Putting agents on top of the same fragmented workflows repeats the error at a higher level of sophistication. Banks should redesign the work first, then decide what to delegate.
• Chasing vanity metrics: Proofs of concept stall without a funded path to production. Counting agents deployed repeats RPA’s habit of counting bots instead of measuring real business outcomes.
• Underinvesting in data foundations. Machine learning taught that model quality is capped by data quality. Agents that retrieve from outdated policies, duplicated documents or conflicting sources will reason confidently from bad premises.

Looking ahead, do you see regulatory compliance evolving toward a form of continuous or even partially autonomous compliance, where AI systems detect regulatory changes, assess their impact, update workflows, and escalate only the exceptions that require human judgment? What would need to happen before banks and regulators are comfortable with that level of autonomy?

Yes, but the autonomy will arrive in stages. Detection, triage, impact mapping and workflow updates for well-defined changes will become largely continuous and automated. Judgment on ambiguous rules, risk acceptance and regulatory positions will stay with humans. A realistic end state is “supervised autonomy,” where machines run the cycle by default and humans govern by exception.

For this to happen, regulators also need to publish rules in structured, versioned formats, building on initiatives like the FCA’s and the Bank of England’s digital regulatory reporting work.

There will also be the need to have reversibility and kill switches built into the execution systems so that autonomous workflow changes must be easy to roll back when anomalies, error rates or model changes are detected. Comfort will grow faster when supervisors use similar tools themselves. That creates a shared language for testing, and possibly direct machine-to-machine supervision over time.

Thank you for the great interview, readers who wish to learn more should visit EvoluteIQ.