Back

Financial Services Improve Operations

Achieving operational efficiency through human-led agentic delivery

11/18/2026

by Susanne Turnbo and Mikaela Fidler

Banks are under constant pressure to do more with less. Improve operational efficiency. Reduce costs. Accelerate delivery. Find new ways to innovate.

And what’s being treated as the golden ticket to achieving all of this? Agentic delivery. And for some organizations, it might be. However, before we start asking what AI agents can take off our teams’ plates, shouldn’t we first understand why that work is on their plates in the first place?

Some of the biggest problems slowing banks down have less to do with how quickly work gets done and more to do with how the business operates. Agentic delivery has the potential to optimize how banks operate and how day-to-day work gets done. But it can’t fix a business process that was broken to begin with. It still requires people to determine what should change, what can be automated, and where human judgment and oversight are needed.

Understand your business before leveraging agentic delivery

When a commercial banking client came to us and said, “We want to do agentic delivery,” we asked the obvious next question: “Why?” They couldn’t answer it. After multiple discovery sessions, we finally got to what appeared to be the goal. They wanted to increase operational efficiency. Problem solved. But not quite.

“Increase operational efficiency” is a direction but not a specific problem to be solved. What does increase operational efficiency actually mean? Faster delivery? Lower operating costs? Fewer manual activities? More importantly, what’s preventing the business from achieving those things today? Is it the work itself, the processes surrounding it, or the decisions and priorities that are holding everything up?

So we asked why again.

Although tedious, and sometimes rather annoying, asking “why” until everyone has run out of answers is one of the most important habits when implementing any type of change in your organization.

What we learned was that the client wasn’t simply looking to move faster. They were tired of the inconsistent red tape slowing them down and the effort spent navigating overly complex processes, which stole time from the innovative work their people were hired to do in the first place.

As a result, we shifted the question to “How?” How was work actually moving between teams? How were decisions constantly stalling? And how consistently were teams communicating and collaborating?

Once we mapped their current processes, we started to see where the problems stemmed from. Teams needed clearer ownership and accountability for decisions, more consistent approval processes, and more streamlined standard operating procedures. These were foundational business issues that needed to be addressed before we could introduce agents and expect them to deliver the value the client was looking for.

We learned if you can’t clearly explain the business problem you’re trying to solve, you’re not ready to choose technology that could solve it.

Turn problems into operational efficiency goals you can measure

Once the business processes are clearly mapped, governance is established, and pain points are identified, we can get specific about what success should look like and how to measure it. Organizations often start with whatever metric is easiest to capture, like how many tasks an agent completed or how much time an automation saved, and work backward to justify the investment.

However, if the problem is delayed approvals, those numbers don’t tell us whether anything has actually improved. We should be measuring how long decisions sit waiting, how often work gets sent back, how many tasks require someone to intervene, and how long resolution takes.

These aren’t questions organizations should only be asking when agents are first implemented. If your organization isn’t able to see or speak to those improvements, you should be questioning whether the business processes were ready for agents and whether the right governance was in place, not just the agents.

Change, rather than eliminate, human governance

Once we’ve identified what needs to improve and how we’ll measure it, the next question is what work should actually be handed over to agents. Some routine activities can be automated from start to finish, while others may benefit from agents gathering information, analyzing issues, or recommending next steps. But there’s a critical difference between helping make a decision and having the authority to make it independently.

Before handing work over to agents, organizations need to be clear about what they’re actually trusting them to do, especially in banking, where decisions can impact customers, regulatory compliance, and the business itself.

Just because agents are taking work off our teams’ plates doesn’t mean we’re taking away their responsibility for it. As agents take on more work and are given more authority to make decisions independently, the way we govern that work needs to change with them.

Human governance doesn’t mean having someone approve every action an agent takes. That defeats the purpose of automation. It means understanding the business processes we’re introducing agents into, establishing clear rules around what they can and can’t do, and making sure the right people remain accountable for the work. Without strong governance in place, we’re giving agents more responsibility than the business may be able to manage.

There are four areas we’d recommend focusing on:

  • Clear ownership and accountability. Every agent-assisted process needs a person responsible for its performance and outcomes, including the decisions the agent is authorized to make. Just because an agent completes the work doesn’t mean the business owner is no longer responsible for it.
  • Validation against business requirements. Completing a task doesn’t necessarily mean completing it correctly. Organizations need clear requirements, testing, and validation procedures to ensure agents are producing the intended results.
  • Escalation and exception management. Agents will encounter situations they weren’t designed or authorized to handle. When that happens, someone needs to know there’s an issue, who’s responsible for it, and how it gets resolved. Otherwise, we’re creating another bottleneck rather than removing one.
  • Documentation and auditability. Organizations need to know what agents did, what decisions were made, and where human intervention occurred. Without those records, how can we identify recurring issues, evaluate performance, or demonstrate accountability?

And these responsibilities don’t end once agents are live.

As agents take on more work, organizations need to ensure their governance practices are keeping up. When the same exceptions continue to appear or teams are repeatedly correcting an agent’s work, someone needs to determine what’s causing the problem and whether the agent should continue operating as it is. That may mean changing the process, adjusting the controls, or limiting what an agent is allowed to do until the issue is resolved. 

Human governance should give organizations the ability to expand what agents can do without losing accountability for how the work gets done.


We can’t overlook the responsibility banks have to understand how their businesses operate and govern the technology they’re introducing. Giving agents more authority doesn’t change that responsibility. If anything, it makes understanding our processes and establishing the right governance even more important.

Banks need to build strong business foundations, establish the governance that keeps people accountable, and then use technology to scale, innovate, and grow. Connect with one of our consultants today to get greater value out of your technology investments.