Agentic AI: Reimagining Banking Operations in Asia

Speaker A: In this episode of AI and FinTech Learnings, an AI generated conversation breaks down. AI in Asia reimagining banking operations through Agentic AI by McKinsey Company. Welcome back to the Deep Dive. We have a fascinating report on the desk today. And honestly, it feels a bit like we're looking at a blueprint for a machine that hasn't fully been switched on

Speaker B: yet, but when it does, it changes the whole factory floor.

Speaker A: It really does. We are digging into the, uh, next evolution of AI in banking. And before anyone rolls their eyes and thinks, great, another chat about chatbots. Let me just stop you right there, please.

Speaker B: We are absolutely not talking about chatbots.

Speaker A: Not at all.

Speaker B: No. In fact, if you walk away from this discussion thinking about a better customer service bot, we haven't done our job. We need to, you know, completely reset the baseline here.

Speaker A: Most of us, when we hear AI and banking, we have this picture in our heads. It's that little bubble in the corner of your banking app. You type, what's my balance? And it tells you, or I lost

Speaker B: my card and it gives you a link.

Speaker A: Exactly. Is it useful? Sure. Is it revolutionary? Hardly. It's basically a slightly smarter FAQ page. But this report we have from McKinsey focusing on Asian banking is pointing to

Speaker B: something completely different, something fundamentally different.

Speaker A: We are moving beyond gen AI as just a content generator. We're moving beyond predictive AI as just a, you know, a calculator. We are entering the era of agentic

Speaker B: AI, and that is the defining term for this whole deep dive agentic AI. The distinction is absolutely crucial. It's the difference between a tool you pick up and use and a teammate you hire to do a job.

Speaker A: Systems that can actually do things, not just talk about things or summarize things, but actually execute tasks.

Speaker B: Right.

Speaker A: And we're specifically looking at how Asian banks are using these multi, uh, agent systems to completely transform their operations. And I know transform is a word that gets thrown around a lot in tech.

Speaker B: Oh, it gets overused constantly.

Speaker A: But I think in this case it actually fits.

Speaker B: It really does.

Speaker A: Yeah.

Speaker B: And Asia is the perfect testing ground. The report highlights that banks there are at this, this unique inflection point, they're sort of leapfrogging the legacy systems that a lot of western banks are still wrestling with.

Speaker A: But the core ideas, they're global.

Speaker B: Oh, completely. Uh, the principles apply everywhere. If you're a product manager, a founder, or just anyone building something in fintech, this report is basically a blueprint for the next generation of product architecture. This is what you should be building toward.

Speaker A: And that's really the why for you listening today, we aren't just geeking out on tech trends for fun. If you are building a product right now, you need to understand this shift. It's not about making a faster chatbot. It's about building autonomous workflows.

Speaker B: Totally.

Speaker A: So let's map out where we're going. First, we need to get a little technical and define what on earth agentic AI actually is and how it's different.

Speaker B: Yep, the fundamentals.

Speaker A: Then we're going to open up what the report calls the 10 domain playbook and look at how these agents are changing the customer journey, credit, even financial

Speaker B: crime, the real world applications.

Speaker A: After that, the how to building an agent library and this concept they call zero based design. And finally, we'll translate all of this into a practical to do list for all you fintech innovators out there.

Speaker B: It's a packed agenda, but it has to be. The landscape is shifting so fast.

Speaker A: Okay, so let's get into it. Section one, the shift from tools to teammates. The report draws a really sharp line between predictive AI, single LLM gen AI and agentic AI. Can you help us draw those lines? Because I think a lot of people use these terms interchangeably and it must cause a lot of confusion in product strategy.

Speaker B: It causes massive confusion. Let's look at the evolution. First. You had predictive AI. This has been the backbone of banking analytics for, uh, well, for a long time.

Speaker A: Right.

Speaker B: It's essentially a very, very sophisticated calculator. It's great at looking at historical data, you know, endless rows in a database and saying, this customer has a high propensity to buy a mortgage. Or this loan application has a high default risk.

Speaker A: So it gives you a score, a number.

Speaker B: Right.

Speaker A: It doesn't write the email, move the money. It just says, hey, human, look at this. It's like a specialized alarm system.

Speaker B: It's an alarm system. That's a perfect analogy. Then we moved into the era of the single large language model or gene AI. This is what's exploded over the last couple of years with things like ChatGPT.

Speaker A: The copilot model.

Speaker B: Exactly. Your copilot. You can summarize a long document, it can draft an email, it can answer questions based on a knowledge base. But it's usually one model, one brain doing a generalized task. It's a tool you use. You give it a prompt, it gives you an answer, and then the interaction is over.

Speaker A: It's passive. It waits for instructions.

Speaker B: Completely passive.

Speaker A: Yeah.

Speaker B: It has no agency to go and do something else unless you explicitly tell it to.

Speaker A: Okay, so predictive AI is a calculator. GeneAI is a really smart intern who can write well, but just sits there waiting for the next task. What makes agentic AI different agentic AI,

Speaker B: specifically these multi agent systems is a fleet of specialized digital co workers. That's the key. Instead of one giant brain trying to do everything, you have specialized agents that work together to execute complex multi step workflows.

Speaker A: And they do more than just generate text.

Speaker B: Oh, way more. They use tools. They can access APIs, they update databases, they make decisions within certain guardrails. And this is critical. They have memory and persistence. They can pursue a goal over days, not just for one interaction.

Speaker A: The report calls this the atomic agent concept. I found this part fascinating because it's like you're creating a whole new org chart, but for software. You're deconstructing a job into tiny little pieces.

Speaker B: That is the best way to think about it. Deconstruction. The report actually identifies nine key atomic agents. And for any product managers listening, this is so important. You don't build a monolithic mortgage bot. You build a fleet of these atomic ages that can be combined to do a mortgage process or a loan or a credit card application.

Speaker A: Let's run through a few of them just to make it real. They list things like a follow up agent, a monitor agent, an execute agent. These sound like job titles.

Speaker B: They essentially are. You also have a verify agent. Think of that one as the obsessive details person checking documents for inconsistencies. You have a create agent that drafts reports or letters. A coach agent that can provide feedback or guidance, a coordinate agent, the project manager. Project manager agent. Exactly. And even a call agent that can handle actual voice interactions.

Speaker A: And these agents, they talk to each other? Or is it just a bunch of scripts running in parallel?

Speaker B: They talk to each other, but they're typically managed by what the report calls an orchestrator.

Speaker A: Okay, so the orchestrator is the boss,

Speaker B: the orchestrator is the project manager, the brain of the whole operation. It takes a high level goal from a human, like onboard this new corporate client. Then it breaks that goal down into steps and assigns the tasks to the right agents.

Speaker A: So it would say, okay, create agent, you draft the welcome letter. Verify agent, you go check their tax ID against the government database.

Speaker B: And risk agent, you screen for adverse media in the background.

Speaker A: Exactly. It's delegating. It's a huge mental model shift. It's moving from I am a human using a tool to I am a human supervising a Team. But let me play devil's advocate for a second. If I have to supervise 20 different software agents, isn't that just as much work as doing the job myself?

Speaker B: That is the fear, right? The micromanagement trap. But the report has a really powerful statistic on this. It suggests this shift allows humans to spend 80% of their time on strategic activities, decision making, complex customer interactions, empathy.

Speaker A: Instead of the grunt work.

Speaker B: Right. Instead of the current 40 to 60% they spend on non core execution. The supervision is high level. You're not checking every single keystroke, you're checking the final outputs and handling the exceptions, the weird cases the agents can't solve.

Speaker A: So it's about getting out of the weeds of, you know, copy pasting data from a PDF into a web form.

Speaker B: Precisely. It's the shift from being human in the loop doing the work, to being human on the loop supervising the agents. You become the conductor, not the first violin.

Speaker A: Okay, I like that. Let's see what this looks like in practice. Section two, the Customer Experience Transformation. The report dives right into the front office. And they use the example of opening an SME account. A small or medium enterprise account.

Speaker B: Uh, ah yes. Which if you've ever tried to do for a business, you know, can be an absolute nightmare. It's one of the highest friction points in all of banking. Personal accounts are pretty easy now, but business accounts are still so hard.

Speaker A: Oh, absolutely. The report says the current state is just messy. The front end might look digital, you've got a nice app, a shiny website, but the back end is all manual. They say it takes five to eight days on average. Why is there such a huge disconnect?

Speaker B: That delay is almost always because of freeworks. It's the back and forth. The ping pong match. A customer uploads a document, maybe a proof of address, but it's a little blurry. A human operations person reviews it two days later, flags it and sends an email.

Speaker A: Right. And the customer is busy running their business. They see the email a day later and maybe they reply with the wrong file by mistake.

Speaker B: Exactly. Then the ops person is busy, so they check it three days after that. That's where your eight days go. It's not the work itself, it's the idle time between the steps.

Speaker A: So how does the Agentix solution fix this? It can't just be a faster email system.

Speaker B: No, it removes the ping pong entirely. Imagine that orchestrator we were talking about. The customer starts the application immediately. An autofill agent starts pulling data via APIs or OCR from their business registration documents. You don't have to type your company address. The agent fetches it from the official database.

Speaker A: Okay, less data entry for me. That's already a good start.

Speaker B: But here's the kicker. In real time, while you're still on the screen, a verify agent is checking those documents. If a document is blurry or the name on it doesn't match the tax ID you entered, the agent catches it instantly.

Speaker A: So no two day wait, no wait at all.

Speaker B: It pops up a message right then and there. Hey, this proof of address looks a little blurry. Can you please re upload a clearer copy? The feedback is immediate.

Speaker A: So the verify agent is doing that tedious grunt work instantly.

Speaker B: Exactly. And at the same time, a risk agent is running adverse media screenings and checking sanction lists in the background. And if the customer gets distracted and drops off the application, which happens all the time. All the time. A communication agent chases them. But not with a generic come back and finish email. It sends a very specific message. Hey, we just need that one tax document to finish processing your application. Here's a secure link to upload it.

Speaker A: The results they quote in the report are just staggering. They talk about reducing that account opening time from five to eight days down to 24 to 48 hours.

Speaker B: It is a game changer.

Speaker A: And REWORKS reduced by 60 to 70%. That's not just an efficiency gain. That's a fundamentally different product experience.

Speaker B: It creates a wow moment for the customer. And that's just onboarding. The report also highlights sales and distribution. Talking about the super empowered relationship manager.

Speaker A: I loved this concept of the morning huddle planner. Right now, a relationship manager, an RM comes in, opens their email and just reacts to whatever fires are burning.

Speaker B: It's totally defensive, totally reactive versus proactive. In this agentic model, an agent analyzes the RM's entire portfolio overnight. It looks at transaction patterns, it scans news alerts about their clients. It checks for upcoming loan maturities.

Speaker A: And it builds a to do list for them.

Speaker B: It builds a prioritized to do list and says here are the five people you need to call today. And here's exactly why. Mr. Smith just had a large deposit come in. You should suggest this investment product. Mrs. Jones's business loan is maturing in two months. You should start a conversation about refinancing. Provides the what and the why.

Speaker A: But it goes even further than just telling you who to call. They talk about post meeting automation. This sounded like pure magic to me.

Speaker B: This is the killer feature for anyone who's ever worked in sales an agent can listen in on the client meeting.

Speaker A: With permission, of course.

Speaker B: With permission, yes. It summarizes the key points, it updates the CRM automatically. So no more. I'll log my calls at the end of the day.

Speaker A: And then forgetting the bane of every sales manager's existence.

Speaker B: Absolutely. And then it draft the follow up email with the correct product offer attached ready for the RM to review and send.

Speaker A: They're talking about salesforce effectiveness jumping by 50 to 100%. Yeah, it's not by making them better salespeople, really. It's just by removing all the administrative burden of being a salesperson.

Speaker B: It lets the salesperson actually sell and build relationships rather than be a data entry clerk. It returns the human to the human part of the job.

Speaker A: Okay, let's move into the deeper plumbing of the bank. Section three, risk and money. This is the middle and back office. The report hones in on lending and credit operations. Now, the credit memo is historically this piece of holy grail banking paperwork. And it's also the biggest bottleneck.

Speaker B: It is the absolute bottleneck. The current state is that a human analyst has to manually collate all this data. Bank statements, tax returns, credit reports from different bureaus.

Speaker A: It's a treasure hunt.

Speaker B: It's a total treasure hunt. Then they have to spread the financials, which means manually keying numbers from a PDF into an Excel model, which is hugely prone to error. Then they check it all against policy. And then they have to write this long memo to pass around for approval. It can take three to five days just to get the paper ready for a decision.

Speaker A: And the future state they describe is this AI led dossier. How is that different from just, you know, a digitized by vjooqic word document?

Speaker B: It's a radical change in the workflow. Agents extract the financial data directly from the documents. Other agents spread the financials. They map the data from the PDF directly into the bank's credit model. No human typing involved. Then another agent checks for policy compliance. Does this loan to value ratio fit our rules? Is the debt service coverage ratio acceptable? And then this is the coolest part. The agents draft the memo. They write the narrative summary of the deal.

Speaker A: So the human underwriter isn't writing anymore.

Speaker B: The human becomes a reviewer or an editor, not the writer. They step in for exceptions. If an agent flags something unusual. Hey, revenue dropped 50% in Q3. This is an anomaly. The human digs into that specific issue. But all the baseline work is just done. And the impact the report suggests decision cycles drop to less than a day.

Speaker A: That's a massive competitive advantage. If I can approve a business loan in four hours and my competitor takes four days, I win that business every single time. It's that simple.

Speaker B: Speed is the new currency in lending. But it's not just about speed. It's also about better risk management. The report also touches on financial crime, sore fraud and anti money laundering.

Speaker A: And there was a scary stat in there. Digital fraud has increased fivefold in some regions like India over the last five years.

Speaker B: The fraudsters are using technology, so the banks have to as well. It's a constant arms race. The big problem with current fraud detection systems is the number of false positives. A simple rule says block any transaction over $10,000 and suddenly you're blocking legitimate business payments.

Speaker A: And then a human has to investigate each one, which takes hours. And in the meantime, the customer is angry because their card didn't work at the store.

Speaker B: Right. The agency fix is to have agents conduct that initial investigation automatically. An agent can check the transaction history, compare it against the customer's typical spending profile, which look at the merchant's reputation, check IP addresses.

Speaker A: All in milliseconds.

Speaker B: Exactly. And then it writes a risk summary for the human analyst. The report cites a three to four times improvement in actual fraud detection rates. And this is huge. A 40 to 50% reduction in those false positives.

Speaker A: It's just more thorough. It's not a binary yes, no rule. It's a full investigation happening at machine speed.

Speaker B: And it scales. Yeah, you can't just hire enough human investigators to cover a five fold increase in fraud attempts. You need the agents.

Speaker A: There's one more domain in this section, Collections, which is usually the most unpleasant part of banking. It's just a blunt pay us now.

Speaker B: It uh, is usually very blunt and impersonal, but Agendic AI allows for what they call hyper personalization. Instead of a generic script, agents analyze the customer's behavior and history. They can determine the best tone to use the best channel. Maybe this person responds better to Ah, a WhatsApp message than a phone call. And even the best offer to structure a repayment plan.

Speaker A: That almost sounds more empathetic.

Speaker B: Weirdly enough, it can be. It's about maximizing recovery while trying to maintain the relationship. If you treat a customer with respect when they are in financial trouble, they're much more likely to stay with you when they recover.

Speaker A: Okay, let's move on to section four, the internal transformation. This is about corporate functions. HR Finance. These are the departments often seen as cost centers. The places where innovation goes to die.

Speaker B: The back Office at the back office. Right. But the opportunity here is massive. It's about moving from reporting the past to predicting the future.

Speaker A: The budgeting example in the report really brought this to life for me. Usually the budgeting process is one poor analyst chasing 30 different department heads for their spreadsheets for three months.

Speaker B: Have you sent me your Q4 projections yet?

Speaker A: It's a coordination nightmare. Just a mess of emails and version control issues. In the agentic model, you have a data collection agent that just pulls the actual numbers automatically from the company's financial systems. Then a budget generator agent creates the baseline forecast.

Speaker B: But the real power comes from the next step.

Speaker A: The scenario ager.

Speaker B: The what if machine.

Speaker A: Exactly. The human CFO can now ask questions. Instead of building spreadsheets. They can say, run me a scenario. What happens to our P and L if interest rates go up by 1%? Or what's the impact if we decide to expand into Thailand next quarter? The scenario agent runs the models instantly.

Speaker B: So the even CFO is spending their time on strategy, reviewing the scenarios, not managing the spreadsheet strategy over spreadsheet management. That's the goal. And the report mentions HR too. Agents handling scheduling, onboarding, even acting as personalized skills coaches for. For employees.

Speaker A: But there's a big structural idea here that connects all these different domains. Which is the concept of shared services or a center of excellence. A, uh, coe?

Speaker B: Yes. And this is crucial for making it all efficient.

Speaker A: Explain that. Why do we need a coe? It sounds like more bureaucracy.

Speaker B: It's actually the opposite. Think about that Verify agent that's really good at checking IDs. If you build it for the credit card department, you shouldn't just keep it there.

Speaker A: The mortgage department needs it too.

Speaker B: The mortgage department needs it. The corporate account team needs it. HR needs it for checking new employees credentials. The idea is you centralize these atomic agents into a common library so the whole bank can use the same set of proven, reliable digital teammates.

Speaker A: That makes perfect sense. Don't reinvent the wheel in every single department.

Speaker B: Exactly. It's modular efficiency at an organizational scale.

Speaker A: So we've covered the what, These atomic agents. We've covered the where, front, middle and back office. Now let's get to the how. Section 5, implementation and strategy. The report introduces this philosophy called Zero Based design, or zbd. Uh, now this sounds a little like consulting jargon. What does it actually mean?

Speaker B: It is critical jargon though. CBD means you don't just automate the existing messy process. That's what we call paving the cow path.

Speaker A: Paving the cow Path. I like that image.

Speaker B: Yeah. Think about a path made by cows walking across a field. It winds around rocks, it takes the long way around a hill.

Speaker A: Yeah.

Speaker B: If you just come along and pave over that exact path, you get a smooth road, but it's still a winding and efficient route.

Speaker A: Right. If a process takes 20 steps because of old paper shuffling rules and you just make a robot do those same

Speaker B: 20 steps, you haven't really solved the problem. You've just made a bad process faster. You crystallize the inefficiency.

Speaker A: Got it?

Speaker B: Zero based design says wipe the slate clean. Forget the old path. Assume you have this fleet of intelligent agents. How should this process work? What steps can you remove entirely?

Speaker A: So instead of asking how do we make this 20 failed form easier for the user to type in, you ask, why is the reform at all? Can't the agent just pull the data from the source?

Speaker B: That is the exact mindset shift. The goal is a lean scalable organization. The report even suggests a Future ratio where one human might supervise 20 to 30 agents.

Speaker A: 20 to 30 agents per human. That is a massive force multiplier. But getting there isn't just about buying some software. The report is very clear on this. It outlines five essential capabilities you have to build.

Speaker B: It does. And if you skip any of these, the whole thing will probably fail.

Speaker A: Okay, let's run through them. Number one, a business driven roadmap.

Speaker B: Don't start with the technology. Don't have a meeting and say we need an agencic AI strategy. Start with the business value. Say we need. We need to reduce SME onboarding time by 50%. If you start with a clear business problem, you'll find the right solution.

Speaker A: Number two, talent.

Speaker B: You need new roles. You need people who can think like agent architects. And you need to structure them in what the report calls garage teams. Small, fast, cross functional squads.

Speaker A: Which leads directly to number three, the operating model.

Speaker B: Right, the garage model.

Speaker A: Yeah.

Speaker B: McKinsey really emphasizes this. You don't do a massive two year waterfall project where you write requirements for six months and and then build for a year. You put a business person, an ops expert and a tech person in a room, a virtual or real garage, and you say, build the agent that fixes this specific problem. You have three months.

Speaker A: So it's about rapid iteration. Build, test, fail, fix, deploy.

Speaker B: Exactly.

Speaker A: Number four, technology.

Speaker B: This is about building that central agent library. We talked about creating those reusable components so you're not constantly rebuilding the verify agent 10 different times in 10 different ways.

Speaker A: And finally, number five, which might be the hardest adoption.

Speaker B: It's almost always the hardest one. Change management. If your employees don't trust the agents, they won't use them. Or even worse, they'll double check every single thing the agent does, which completely defeats the purpose of the efficiency game.

Speaker A: That trust factor is huge. If an agent makes one mistake and the human employee gets blamed for it, that human will never rely on it again.

Speaker B: Precisely. That's why that human in the loop model is so vital, especially at the start. The agent acts as the maker, it drafts the work. The human is the checker, the final approval. Over time, as trust builds, the human checks less and less. But you never fully remove the human from the ultimate accountability loop.

Speaker A: Okay, we've arrived at our final section, section six. This is where we translate this whole McKinsey report into practical actions for our listeners, the fintech product managers and founders. We've synthesized five key takeaways. These are your action items, what you should be thinking about on Monday morning.

Speaker B: Let's do it.

Speaker A: Okay, key takeaway number one. Audit your roadmap for atomic agents.

Speaker B: Exactly. Look at your current product roadmap. Look at all the features you're planning to build that require human input or manual work. Stop and think. Can this be deconstructed? Look for those reusable tasks, verification, summarization, chasing people for information. Don't build a monolithic feature, build an atomic agent for that task and start creating your own internal agent library.

Speaker A: So don't just build a loan application helper. Instead build a document verification agent that can then be used for loans, credit cards and your KYC process.

Speaker B: Exactly. Modularize your AI. Think in terms of Lego blocks, not giant custom built statues.

Speaker A: I like that. Okay, key takeaway number two. Use zero based design for all new features.

Speaker B: When you're designing a new user flow, your first question should not be how do we make this UI better? Your first question should be, how can our agents do this for the user? Can the user just upload a single PDF and the agents fill out the entire application? Can the agent negotiate the interest rate within a pre approved range? Shift the burden of work from the user to your agents.

Speaker A: Shift the burden. I love that framing. Make the software do the work. Okay, key takeaway number three. Start designing the supervisor ui. This seems like a really big UX challenge.

Speaker B: It's a massive design challenge and it's a new one for most teams. It's. If humans are going to manage 20 agents, the UI can't look like a simple data entry screen. It needs to look like a Dashboard a control center. Users need to see what the agent did, why it did it, and have a simple way to verify and approve it. We need to start designing UIs for supervision, not just for execution.

Speaker A: It's like designing a cockpit for a pilot rather than just a steering wheel for a driver. You need visibility on all the system statuses, not just the road right in front of you.

Speaker B: That is a great analogy. The UI needs to surface things like confidence scores. The agent is 98% sure this ID is valid, but only 60% sure about this address. That context is everything for the human supervisor.

Speaker A: Okay, key takeaway number four. Start with human in the loop.

Speaker B: Don't try to achieve 100% automation on day one. It's too risky and frankly, it scares people. Build the agent to do the first draft of the work, but let the human be the editor, the checker. It dramatically lowers the barrier to entry. It builds trust, and it also gives you invaluable training data to make the agent smarter over time.

Speaker A: Makes sense. And finally, key takeaway number five, focus on reusability.

Speaker B: Don't let your teams build a support bot and a sales bot. Incomplete silos. They're solving similar problems. Instead, build a context retrieval agent. They can figure out who the customer is and what their history is. Build a tone adjustment agent that can make communication more formal or more friendly.

Speaker A: And both the support and sales teams can use those same agents.

Speaker B: Exactly. If you build them as atomic reusable agents, you build them once and you can use them everywhere. It's just smart, efficient product architecture. It's how you avoid technical debt down the line.

Speaker A: So we've covered a lot of ground today, from atomic agents to zero based design. What's the one sentence summary here?

Speaker B: I think it's that we are at a genuine inflection point. The winners in fintech over the next few years won't just be the ones who are using AI tools, they will be the ones who learn how to orchestrate AI agents. The 10 domains we discussed, from 24 hour count opening to AI led credit dossier, that's the roadmap. That's where all the value is hiding in plain sight.

Speaker A: It really feels like the industry is finally waking up to the fact that AI isn't just a shiny toy to put on your website. It's the new engine of the entire operation.

Speaker B: It's a new workforce.

Speaker A: I want to leave our listeners with a final provocative thought to chew on. We mentioned that statistic from the report. One human supervising 20 to 30 agents. Chris, if that ratio is even close to being true. What does the org chart of a unicorn fintech look like in, say, 2028?

Speaker B: That's a profound question. Today a unicorn might have 500 or 1,000 employees to get that valuation in 2028. M maybe it has 50 human employees, but 5,000 agents doing the work. The organizational structure might look more like a server farm architecture than a traditional corporate hierarchy.

Speaker A: Wow.

Speaker B: You wouldn't have layers of middle management, you'd have layers of agent orchestration and supervision.

Speaker A: A, uh, 50 person company doing the work of a 5,000 person bank. That is both incredibly exciting and frankly, a little terrifying to think about.

Speaker B: It is the future of operations that

Speaker A: brings us to the end of this episode of AI and FinTech Learnings. An AI generated learning experience. Thanks for listening.