AI adoption in banking is no longer a challenge. Execution is. Across lending, onboarding, and servicing, the story repeats: A loan is approved, but stuck in exceptions. Onboarding is digital but breaks at KYC. Service is fast but disconnected from the actual process.
Why? Fragmentation. AI, workflows, content and communication still operate in silos—so decisions don't translate into outcomes. Leading banks are solving this with orchestration: a control layer that connects everything in real time.
The result:
• Faster turnaround
• Fewer manual touchpoints
• Full auditability
AI won't be the differentiator. How well it's orchestrated and how confidently it stands up to regulation will be.
Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.
Danielle Fugazy (00:11):
AI and banking has scaled quickly, but not evenly. Tier one banks are running one to 300 plus AI models in production, yet less than 30% of journeys are fully automated end to end and model governance costs have risen two to three times with scale. At the same time, regulators are tightening expectations around explainability, bias and fairness and full audit trails. So while AI is delivering gains in fraud lending and servicing, scaling it safely and consistently is becoming a real challenge. To unpack this, we are joined by Anand Raman, Executive Vice President and Chief Operating Officer, Americans at NewGen for today's leaders episode. Welcome, Anand.
Anand Raman (01:00):
Thank you. Thank you, Danielle, for having me.
Danielle Fugazy (01:01):
Thanks for coming today. So let's just start by level setting for everybody. How do you see this playing out across banks and how are large banks versus midsize banks participating in the AI shift today?
Anand Raman (01:16):
Absolutely. You said, you're right. In terms of everybody is participating, I mean, you can't really go too far out and not see AI. I think one thing we see with large banks is that they've been able to digitize their operations across the board. They've invested billions of dollars. They're now setting up these AI governance frameworks. They're setting up these AI data controls with large tech teams in place. If you look at the mid-size banks, they are investing more on specific use cases. So they're working with specialist vendors on trying to solve for specific problems around fraud or lending or around customer servicing. But I think one thing that we find common across both large banks as well as the mid-market banks is that still most of the decisions that are being taken are manual. So about 40, 50, 60% of decisions require manual intervention. So the large banks have taken more of a platform view of AI.
(02:13):
They're investing heavily on that. Mid-market banks have gone in for more point application specific use cases, but the outcomes they're both getting are largely, they're still depending a lot on manual intervention.
Danielle Fugazy (02:26):
So where do you see the opportunity for the most measurable or tangible results here? Where's the impact today?
Anand Raman (02:34):
Absolutely. See, we see opportunity for banks to get AI improvements from the front office to the middle office, all the way to the back office. Traditionally, in the first phase of adoption of AI, banks have focused on the front office in terms of digital self-service, customer assistance for the relationship managers of the banks. What they have not done yet is more invest on the middle office and the back office. So we find that there is potential to, using AI, banks can get 40, 50%, 60% in some cases, depending on the specific areas that they're in, benefits. So if you take, for example, lending from offering structured recommendations to customers, to reducing the documentation errors, to things like accelerating that credit review process, or even how quickly they can close that loan closing package. We see that there's a huge opportunity for banks to be able to drive a lot of efficiency and also improve their customer servicing.
Danielle Fugazy (03:40):
So it's really across the board here.
Anand Raman (03:42):
It's across the board. So if you take commercial lending, you can take consumer lending, you take small business lending, you take deposit account opening for consumer accounts, business accounts, customer servicing. So it's pretty much every major area of the banks. Because finally you're talking about that front office, middle office, back office, all three areas is huge potential.
Danielle Fugazy (03:59):
So the potential outcomes are strong here. You've said that. What is preventing the banks now from scaling across the enterprise? What's the challenge here?
Anand Raman (04:13):
See, banks have been doing a lot of pilots across different lines of businesses. The challenge that they've been facing is that although they've been getting benefits at individual steps in that journey, a specific agent might be giving a good response summarization or an assistance for fraud or for risk or doing whatever the bank is expecting from that agent for that particular step. But unfortunately, these agents are not journey aware. So they're disconnected across the various steps of that process that the bank is trying to do. If you're trying to issue a loan or they're trying to onboard a customer, there are multiple steps in that journey. No agent is aware about the context of that journey. As a result, what's happening is that whatever benefit or gains that they're getting from the agent at one step, they are giving back because then that input is then given to the users who now have to stitch this information together manually.
(05:06):
So it's a lot of information for users. Now they'll have to assemble this and then push it to the next step in the workflow. So as a result, the net outcome of the journey, they're not seeing a tangible out benefit. So then there is confidence to scale this beyond pilots is a little lower. The other kind of correlated issue, if you will, is that banks being highly regulated, they need to make sure that there is enough audit, the outcomes are consistent, and they should be able to explain to regulators, who took what decision, using what data, who approved it. Now, in the absence of ... If the execution is broken, if it is fragmented, if that journey is fragmented, governance becomes bolted on instead of embedded. What that means is that now it's left to the users to be able to take information from these AI agents and then kind of put that case together.
(06:01):
So effectively it just increases cost. So again, in terms of are they getting the benefit at the journey level? I think there's still some opportunity there.
Danielle Fugazy (06:09):
Yeah, you're losing your gains there a little bit. So there's a lot more questions I want to get through. I want to get to the risk piece of it that you just talked about. But before we get to that, can you talk a little bit about how banks are solving for this problem?
Anand Raman (06:25):
See, the leading banks, in our view, what we are seeing is that they understand that AI agents will not solve a fragmented process. The agents are only going to be as effective or as context aware as the underlying journey infrastructure that is provided to them. And agents need that to be able to execute that journey coherently. So most of the leading banks are investing, are building out these, or providing these orchestration layers to these agents. So what that means is that think of it like AI as a brain and think of the orchestration layer as the nervous system, which means the brain is responsible for thinking, gathering the facts, and then the nervous system is really what is executing it. So how do you get work done? And so that's where what happens is that this orchestration layer will coordinate the capabilities of the AI agents across things like extraction, summarization, prediction, all the different agents that they have.
(07:22):
It will also connect them across the workflow steps. It'll provide that context of that journey to these agents. And it'll then also intervene in terms of when do we get in humans involved at which step? So I think that's what some of the leading banks are doing.
Danielle Fugazy (07:38):
I like this analogy. It is where AI is headed, right? Yeah. The brain and the nervous system. So getting back to something that you talked about, which I think is important as we move closer and solve these problems, as you introduce more autonomy, how do you do that without increasing risk?
Anand Raman (07:59):
I think most leading banks are trying to build autonomy as part of the journey rather than as a standalone decision. Basically, there are three things there. The first is in terms of what can the agent do? What will trigger an approval or exception which requires human intervention? And then three is what must always be done by a human. I think those are kind of the three big blocks. And most banks, what they've done is they've taken a three-tier approach to autonomy. Level one is what we would call as an assist, where the agent would be more of a digital assistant. It would help with the heavy lifting in terms of read a 50 page tax return, for example, and get the relevant information from that, or a huge bank statement. It would do summarization. So that's kind of like a level one. It would be supporting with information and data.
(08:49):
Level two is where an agent becomes context aware, journey aware. So it starts making recommendations, which means that it can now provide that, hey, this borrower is trying to apply for a loan, but look, the income is not matching, the cash flows are not matching. So hey, you can offer an alternate. So when we are talking about structuring different terms, so it starts recommending. Again, the decision is always going to be taken by the human, but now it becomes, rather than assistant, it becomes a partner. So the AI becomes like a digital partner, if you will. And so that's like a level two. Then a level three is where AI agents can start taking action of act. So basically, can it trigger a task? Can it identify that, hey, this requires additional human approval, or this requires flagging to a different department. So it is now with guardrails in place, the AI agents are able to now take action.
(09:43):
So I think those are the three levels.
Danielle Fugazy (09:45):
How far are we from level three with banks?
Anand Raman (09:49):
So most banks we see will start with level one, then go to level two. So I think most banks are at level one today and many are the leading banks, as you asked, are at level two. I think level three is, of course, the next step in that journey. I think the important thing is that instead of treating it as a switch, most of the leading banks at least are trying to build it as part of the workflow itself. So it becomes a configurational, like a workflow gate, if you will, which means that as your confidence increases on the agents and your governance and you're clearing those audits, you can start tightening or loosening those gates based on how your businesses and your regulator ... I mean, the risk teams are reviewing it.
Danielle Fugazy (10:30):
That makes sense. So can you tell our audience a little bit about Newgen and where it fits in to this dynamic?
Anand Raman (10:38):
Absolutely. So Newgen helps banks fix their fragmented processes. We provide an orchestration layer which helps connect that front office, middle office, and back office so that the journey execution is done in a faster, more efficient manner, and all this is done in a manner which is built with strong governance, trust, and controls. So what that means is that banks can use the orchestration layer to automate major areas of their business such as commercial loan origination, small business loan origination, consumer loan origination, account opening, customer servicing, and be confident that the orchestration layer, the NewGen one platform that we provide will take care of things like auditability, predictable outcomes, consistent outcomes. And how we do that is that we have this unified platform which brings together content, process, and communication into one single platform, helping customers to be able to achieve their, automate their journeys in a much more efficient manner.
Danielle Fugazy (11:36):
So as a final takeaway, what will separate the leaders going forward?
Anand Raman (11:40):
I think the leaders will start treating AI as an operational capability and not just a technology capability. So what that means is AI in itself is not going to be the differentiator. Everybody's going to have AI. The big issue is how consistently will it execute? How reliable is it going to be? How confidently can it pass those audits? So that's what we see as the differentiator. So the best banks are not necessarily the ones which are going to have the most AI models or be the earliest, fastest adopters of AI, but the ones who can operationalize AI in their production environment at scale reliably. I think those will be that will be the winner.
Danielle Fugazy (12:20):
That's a great call out and a great way to end this. Thank you so much for joining us today and for sharing your insights with our audience. I'm sure they're going to find them valuable.
Anand Raman (12:27):
Thank you, Danielle. Thank you for having me.
Danielle Fugazy (12:29):
Thank you for joining us.
