Molly Irelan: Kim, you talk to a lot of banking leaders who are either evaluating Alkami or are actively navigating their AI journey. When financial institutions ask you about AI, what are they really trying to solve?

Kim Reed: When financial institutions ask about AI, it usually comes down to three core pressures, and they tend to show up all at once.

The first is how to better serve account holders. Teams are looking for ways to make interactions more relevant, immediate, and helpful. That’s where conversations around virtual assistants, personalization, always-on fraud detection, and proactive engagement come in. There’s a clear desire to compete on experience.

The second is capacity. Most financial institutions aren’t sitting on excess resources. They’re trying to do more with the teams they have. So the question becomes: how do we reduce time spent on repetitive or manual work and shift that effort toward higher-value activities? AI starts to unlock that, especially in areas like customer support, back-office processing, and internal workflows.

The third, and often most important, is confidence. Leaders are asking: How do we do this in a way that’s safe? Defensible? Aligned with expectations from regulators, our board, and our own policies? This goes beyond what AI can do and starts to lay the foundation of how it will be used at the organization.

Molly Irelan: Where does the hype around AI fall short compared to what teams actually need?

Kim Reed: A lot of the hype suggests AI is going to replace entire functions or teams. That’s not what most institutions need right now, and it’s not where the real value is. What teams actually need are targeted, well-scoped use cases where AI can enhance how work gets done. A good example is customer service. In a traditional model, if you’re trying to understand why a customer or member had a poor support experience, someone might have to review multiple calls, piece together context, and identify root causes. That can take hours. AI can compress that into minutes by summarizing interactions, identifying patterns, and surfacing insights. However, it’s still the human who interprets that information and takes action. The real opportunity is precision and acceleration, not replacement. AI reduces manual work and enhances teams’ abilities to provide answers faster as well as deliver a human touch for high-impact conversations.

Molly Irelan: What’s a common misconception about AI that slows organizations down?

Kim Reed: One of the biggest misconceptions is that AI requires building an entirely new governance model from the ground up. That mindset can stall progress before it even starts. In reality, most institutions already have the foundation they need, they just need to be extended to include AI:

• Risk management frameworks
  • Expand existing risk identification and assessment processes to account for AI-specific risks like model drift, bias, and explainability
• Information security controls
  • Apply existing data protection, access controls, and privacy standards to how AI tools ingest, process, and store information
• Third-party risk management
  • Add AI-specific questions and considerations for 3rd and 4th party vendors
• Change management procedures
  • Incorporate AI models and tools into existing change control processes, including testing, validation, and ongoing monitoring

Molly Irelan: Where should financial institutions actually begin if they’re just getting started with AI in banking?

Kim Reed: Before tools, pilots, or anything else, leaders need to create alignment. You need a shared understanding across the organization of what AI means in your environment, how it will be evaluated, and what “acceptable use” looks like. From there, I’d focus on three foundational moves:

1. Bring the right stakeholders together early
   You need both control and operational perspectives. Risk, compliance, and legal are thinking about protection. Product, technology (tech), and business teams are thinking about opportunity. You need both in the room to map out a list of considerations and priorities.
2. Anchor to a recognized framework
   In February, the U.S. Department of the Treasury released guidance that gives a common language for risk, controls, and outcomes. They also create a level of defensibility if you’re ever challenged on your approach.
3. Establish baseline guardrails
   Include things like a policy and responsible use guide to define clear expectations around data usage, access, and oversight. This doesn’t need to be overly complex, but it needs to exist and align with existing data classifications and standards. Without it, every decision becomes harder than it should be. With it, you can protect your organization from what’s approved/what’s not, as well as encourage employees to be innovative in their use of AI solutions.

Molly Irelan: What does a strong AI adoption playbook look like in practice?

Kim Reed: A practical playbook has four parts that build on each other.

1. Governance and framework
   This is your foundation. Define how decisions are made, who owns what, and how risks are evaluated.
2. Clear objectives and success metrics
   Every use case should tie to a specific business outcome. Are you reducing fraud losses? Improving response times? Increasing engagement? If you can’t define success, you won’t know if AI is working or be able to measure the return on your investment.
3. Disciplined execution
   This is where many teams struggle. You need trained users who understand both what AI can do and where it can fail. Without that, you introduce inconsistency and risk. Focusing on the pre-determined use cases also allows you to curate important lessons learned and share that knowledge across the organization as AI use is expanded among teams.
4. Continuous monitoring and adaptation
   AI systems change over time. New tech emerges. Data changes. Outputs evolve. You need processes in place to validate results, track performance, and adjust when needed. Like any piece of tech, this isn’t a one-time implementation. This should be an ongoing operational capability that creates efficiency gains for the organization.

Molly Irelan: How should bank and credit union teams prioritize their first AI use cases?

Kim Reed: You want to balance business impact with governability. Start by asking:

• Does this meaningfully improve performance (ex: revenue, cost, risk, or experience)?
• Can we clearly explain how it works and defend it if needed?

The best starting point is where the impact is tangible, data is understood, and controls are manageable. That’s why many financial institutions begin with internal use cases — things like document review or internal communications — where the risk is lower and outcomes are easier to measure. Where teams sometimes get into trouble is jumping in too quickly without having the structure to support them.

Molly Irelan: What differentiates a strong AI use case from a risky one?

Kim Reed: Strong use cases are well-defined and contained; having clear inputs and outputs, known users and workflows, controlled data environments, built-in human oversight, and measurable outcomes over time. Risky use cases tend to lack those characteristics. A good example of a risky starting point would be embedding AI directly into decisioning processes, like credit decisions, without understanding how those outputs are generated or how to explain or defend them. If you can’t explain it, you shouldn’t operationalize it.

Molly Irelan: How should leaders think about AI governance without slowing innovation?

Kim Reed: The goal of AI governance is to make progress repeatable and defensible. The key is structure without rigidity. A few ways to do that:

• Use risk-based tiering
  Not every use case needs the same level of scrutiny. A low-risk productivity tool shouldn’t go through the same process as a high-risk decisioning system.
• Standardize intake and review processes
  Create templates, questionnaires, and clear workflows so teams aren’t starting from scratch every time.
• Define escalation paths
  Teams should know what they can approve independently and what needs broader review. Similar to the risk-based tiering approach, teams should understand what level of scrutiny a new AI tool should fall into by its characteristics, such as the dataset it has access to, whether it’s an open or close-looped system, etc.

This creates consistency and reduces the burden on governance teams over time.

Molly Irelan: What are the biggest compliance concerns right now when it comes to AI?

Kim Reed: There are three that come up consistently.

• Data privacy and security
  Financial institutions want to know exactly where their data is going and how it’s being used. Is it staying within controlled environments? Is it being used to train external models?
• Documentation and explainability
  If a regulator asks how a decision was made, can you clearly show the process—from design to deployment to monitoring?
• Third-party and fourth-party risk
  It’s not only about your vendors but also their dependencies on third parties. Are you aware of how AI is being used across that ecosystem? Are you familiar with how your contracts were established? Did you contract through your vendor or directly with their partners? Understanding these aspects is paramount in ensuring tools adding to your tech stack adheres to the risk framework you’ve established.

Molly Irelan: How can financial institutions move forward with confidence while regulators are still catching up?

Kim Reed: Regulators may still be evolving their guidance, but they’ve been clear on what they expect – accountability, traceability, and control. You can build confidence by anchoring to established frameworks and showing how your program aligns with recognized standards. Use real examples that demonstrate how controls actually work in practice. From what I’ve seen, regulators aren’t expecting institutions to avoid AI. They’re expecting banking leaders to adopt it responsibly.

Molly Irelan: What does responsible AI look like in day-to-day operations?

Kim Reed: Aside from policies, responsible AI shows up in how work actually gets done. That includes clear ownership for each use case, defined guardrails around data access and usage, ongoing monitoring for accuracy, drift, and anomalies, feedback loops so users can flag issues, and integration with existing incident and change management processes. AI should be embedded into operations, not treated as a separate initiative.

Molly Irelan: What’s the real cost of waiting on piloting AI?

Kim Reed: There are operational and risk implications. Operationally, you fall behind peers who are already improving efficiency and the digital banking experience. From a risk standpoint, you create an opportunity for shadow AI — teams using tools without oversight because there’s no approved path. Even if you decide not to adopt AI, you still need policies and controls in place. Because AI usage can — and will — happen regardless. Setting these guardrails protects both institutional reputation and employees in their day-to-day work.

Molly Irelan: What would you say to teams that feel behind in their AI journey?

Kim Reed: Regulatory guidance is still evolving, and maturity varies widely across the industry. The focus should be on building a strong foundation centered on governance, data protection, and accountability. Then proving value through a few well-executed use cases. When you do that, you’re not choosing between innovation and control, you’re reinforcing both.

AI doesn’t create advantages on its own. Instead, it succeeds or stalls based on how well it’s operationalized. Leading financial institutions are building a repeatable way to evaluate, deploy, and manage AI — grounded in governance, aligned to real use cases, and supported by clear accountability.

However, there’s one piece that underpins all of it: data. Without a strong data foundation that is unified and accessible, AI becomes inconsistent, difficult to scale, and harder to trust. With it, AI becomes something entirely different. A new mechanism in your organization’s growth engine; turning everyday interactions into insight, and insight into action.