Name: AI meets finance: from strategy to impact
Uploaded: 2025-05-15T17:16:08.708Z
Duration: 31 min 22 s
Description: AI meets finance: from strategy to impact

Transcript for "AI meets finance: from strategy to impact": Hello, everyone. Thank you so much for joining us today. Whether you're coming from to us from the financial services industry or you're an AI enthusiast with a curious mind, welcome. We're glad you're here. We We know your time is valuable, so our goal over these next thirty minutes is to make this not just interesting, but actually useful. You won't get a lot of buzzwords from us today. We're gonna keep it grounded and actionable, kind of like a rack system. Okay? That's borderline buzzword, but, we'll call that one a free. Here's what we're gonna be diving into today. First, a quick overview of the AI landscape in financial services. What's happening? What's the hype? What's already driving real value? Then we're gonna get to some common roadblocks, security, compliance, and, you know, the classic how do we even get started feeling that that many of us have had. And finally, we're gonna leave you with some practical advice. What you can do today, this quarter, this year to make AI work for your team. The bottom line is leveraging AI is no longer optional. It's table stakes, especially in the financial industry. So where expectations are high and competition is even higher. So let's kick something kick things off with some introductions. My name is Nic Morales. I'm head of customer experience and partnerships at Cohere. My role is to ensure that our customers are having an exceptional experience with our AI models and solutions, Whether that's from that first moment they contact and form a relationship with here through their onboarding and adoption journey. I work closely with our customers and partners to drive that mutual growth and outcomes. And with me today are two of our two of my colleagues. First up is is Robin. Robin? Thanks, Nic. Hey, everyone. My name is Robin Gainer. I'm an account director at Cohere, and I've had the pleasure of working with several of our biggest strategic customers across different industries, but certainly with the focus on financial services. So excited to speak with you all today. Fantastic. Thanks, Robin. And now let's hear from Laura, our VP of finance. Laura? Hi, everyone. It's nice to see everyone here today virtually. My name is Laura Moss. I'm the VP of finance at Cohere. I lead all things financial from payroll accounting to, investor relations. So really excited to talk about how AI is impacting folks in my role, people on my team, and just kind of more broadly how we're using, AI to help us with our financial, you know, tasks Alright. Thanks, Robin and Laura. And, so, you know, really appreciate having you both on. You both have a really diverse range of expertise, so I I think this is gonna be a really insightful discussion. So let's dig into the next slide and really start to explore the AI landscape in for in financial services. And we're gonna start by zooming out a bit, looking at how AI adoption is actually playing out in the financial industry services industry. So, Robin, let let's start with you. You've been in the front lines with banks, insurance companies, asset managers, you name it. How have you seen adoption of AI change and evolve over your time at Cohere? Now, what's different now compared to say a year or two ago? Sure. So, I mean, I'll go all the way back. I I joined Cohere in January, 2022 as the first member of the sales team. And during the first ten months of my my time here, I I generally spent at least five minutes in every meeting explaining what a language model is and how they could be useful. Then in November 22, CHaTAPT came out, and that year, things just sort of changed overnight. It came back after the holiday break in '23 and conversations were just totally different. Every senior executive sort of had firsthand experience with what the technology could do and and saw the potential. And then 2023, I think, was really the the year of the POC. And '24 was where we really started to see large enterprises move use cases into production and and scale. And then this January a few months ago at Cohere, we released North, which is our agent foundry. We certainly had a strong hypothesis that this was gonna be what the market really wanted and and needed to drive that next level of adoption and and value creation, but the response has really been overwhelming. It's been an incredible to see and we're certainly working really hard to to keep up with that demand. But I think now more than ever, FIs are moving at speed. You know, there's several economic markers recently showing how strong enterprise adoption and momentum are in financial services. I think 2025 is really the the turning point for many, as, organizations move to sort of move beyond pilots into large scale production usage. And I think, you know, senior executives and other leaders, they're no longer sort of debating if, AI and LMs are should be adopted. There's very broad acceptance. This technology is going to be totally transformative to how FIs do business and create value for their customers. And so there's much more focus on the sort of tactical considerations of how do you do this effectively and and drive business value. But with the mind for the specific requirements of of, the financial service industry, which are, of course, very stringent regulatory scrutiny and a rapidly evolving regulatory landscape. There's a lot of different domains of of risk, including the needs to protect sensitive data, both customer data as well as, firms' own own IP and trade secrets, avoiding intellectual property rights violations, and and, of course, making sure the models are fit for purpose avoiding inaccuracies and hallucinations. So there there's a lot there, but I think maybe before we get too far into it, Nic, what what use cases are you seeing our FI customers most excited about as they move forward with implementation and and production usage? Yeah. Yeah. Great question. You know, very relevant because, ultimately, that use case is what's gonna drive the the value. And and we're seeing these financial institutions who are really serious about Geni. They've moved beyond experimentation, and they're putting it to work in in the real world. So no one common entry point is internal internal knowledge search. So you think about a wealth management firm with hundreds, thousands of of of pages of of documents and compliance documents, investment policy statements, client notes. Teams are using AI to instantly surface that information in plain language, and across all of these sources. So, you know, you think about all these PDFs that they would have to dig through with a mix of text and images and table charts. All of that can now be quickly with natural language, surfaced, searched, and and and, summarized. In the retail banking, we're seeing automation of of routine customer interactions. So things like mortgage prequalification questions or fraud dispute handling. These typically require multiple human touch points. And with AI, we're seeing, organizations, use AI to help triage and respond faster, but still keeping human in the loop when needed. So, you know, as we move forward, we're seeing this interest just grow into more advanced use cases, market forecasting that adapts in real time, an AI assisted portfolio construction that you know, responds to the client goals and their risk profiles automatically. So it it's not just about the efficiency. We're also seeing an unlock of new capabilities that have just been out of reach before. And, you know, these are examples, you know, rather than that highlight the potential of AI in the industry. But and you mentioned this a little bit earlier. We know how difficult it is to move from testing to a POC or or a pilot and and get that now into a full scale production. So, it's not always a smooth journey. You know, Robin Gainer, again, you've been working with with, some of the largest customers in the world and the industry. What are some of these biggest barriers that you're seeing in this transition? So I think what's interesting is the biggest barriers to adoption are are not modeling or sort of technology problems. They are traditional, technology adoption problems like like governance, like policy, like integration and and change management problems. And so Cohere, we've really focused on private deployments, making this technology available to our enterprise customers in their own environments, whether that's through virtual private cloud, VPC, or on premise. We're really seeing that becoming a preferred choice for for FIs because it just instantly mitigates so many of those, risk domains that I I spoke of earlier. It really, engenders a lot of confidence, among different stakeholders in the technology if they know that sensitive data is never leaving the firm's, own environment. And and, obviously, there's risks, with with some adoption. But there's also a risk in really not adopting an enterprise wide strategy for this technology because there's been, many, stories and and sort of studies now showing, if firms don't provide availability to this kind of tech, employees are just using it in personal accounts and which exposes their employers to data loss risk if they're putting sensitive customer or trade secret information into these these public models. And the other thing with private deployment is in many cases, it can also serve as a a cost saving strategy. Certainly, Cohere, we focus a lot on the efficiency and the compute footprint of our models. So as you scale, usage, that really helps reduce total cost of ownership. But I think what a lot of customers are finding too is that just sort of, you know, hitting an API and, sort of token based or use case use space case consumption is not really driving, true business value. Most end users can't use, APIs. So serving, the models in that way really creates this massive barrier, and in turn a burden on other parts of the organization, whether that's IT or sort of tech and operations to to make this technology available to the different lines of business. I think we're also seeing that there is a real need still for customization and and configuration. Like, large language models do provide great performance out of the box on a broad range of use cases, but FIs typically still do want some final, layer of customization, whether that simply lies in kind of the interface layer or in some cases in in driving the model to, to perform better on their particular sort of subdomain or use cases. And there's a lot of different things that can be done to, help models produce better outputs, whether it's sort of the most basic, type things like prompt tuning or prompt engineering, it's called sometimes, to sending custom instructions to the model, in a more, complex or agentic flow, data annotation, all the way to sort of continuous free training, as we did with Fujitsu to build in the custom Japanese model, for example, to really help them drive differentiation and and alpha there. So there's a lot that can be do done to tune this technology for specific downstream applications, and I think that's where it's really important to have a close partnership with the the technology provider. It's gonna really be able to help to guide you what is the right level to intervene at given some set of objectives or goals. Yeah. Thanks for having a great examples and, you know, really demonstrates, you know, that differentiation that, these organizations can can have and, you know, how they navigate through through some of these challenges that are unique to the the financial services industry. Now, Laura, as as as the VP of finance, you're in a unique position to evaluate the return on investment of new technology. How should finance leaders be thinking about ROI when investing in AI? You know, especially given, you know, what we've learned from from Robin around how it's being, moved into production environments. Yeah. That's a great question. So, I mean, there's there's two ways to think about it, and I'll tell you a little bit about how I think about using AI just on our team. But I think more generally as, a finance leader at a, you know, any company really or especially in the financial services space, I think the leaders, you know, are thinking about AI as, you know, what can it enable and unlock for our our company. Obviously, there's, the aspect of AI that allows us to do our jobs better. You can do more. You can think of more interesting answers to problems. It frees you up to have, you know, an assistant, a knowledge assistant as as Robin was talking a little bit about. Obviously, there's also a ton of time savings that can happen, especially when you're searching through large, reams of data, SEC filings, contracts, invoices. So, you know, really some cost savings advantage there, and maybe even offloading repetitive workflows, so your teams can spend more time on work that's adding value to your company. I think another thing that's really important to to have nowadays is a budget for AI. So as you kind of think of the strategy of your organization, making sure you're setting aside, some money to be doing this investment. As Robin talked about, sometimes there is a lag between trialing, figuring out the workflows that you're going to be embedding AI into and then seeing the returns come, later from an efficiency and revenue standpoint. And I think also being willing to experiment. You know, a lot of folks are excited to use this technology and it really is super helpful. And so, kinda giving the the teams the the freedom to do that, to figure out where it works best for their day job. Because I think, ultimately, the way we're all using AI in our lives now is very organic, and it it's fitting in. Right? And it's fitting in in a way that, search has fit in in the past. You know, new technology like phones have kind of just become part of us. And so I think this will really be the same, when you think about it from a work interaction, how you interact with your work and the applications and the data sources that you need to do your work from a financial point of view. Pivoting a little bit to talk about how how I've used it and how my team has used it, I think last year, we started with very, you know, simple use cases, like using it to write a job requisition, using it to review, expense reports that come in, figuring out if it aligns with the policy or doesn't align with the policy. Now we're using it, in in many more ways including, like, data translation into various formats. That's, like, a great use of it. It's really easy. You can automate it, simplifies your workflow, like, okay. You get information in this format. You need to translate it into this other format to then upload it into your ERP of choice. Like, we use NetSuite, so, like, gotta get it in the right formats for the for our financial reporting. On the, you know, other kind of ways that we use it is we have a lot of PDF, you know, documents that are related to contracts or customer agreements or, you know, even vendors. So we use, our our internal solutions to kind of go through all of that information on a monthly basis and synthesize it into a management report that will kind of highlight, okay. These are new contracts, for example. These contracts have, unusual clauses in these ways, and so it really, really helps kind of our month end close process. I would say we're also using it in our procurement workflows, with a similar, you know, kind of actual use case in mind where you're going through the agreements and pulling things out, that you want management to look at. And then, of course, being in finance, we do do a lot of ad hoc reports and analysis and write ups. And so leveraging, AI to both find that information and start populating it into formats that we know are are helpful to the executive team is kind of where we are right now. And that's really that's really gonna, you know, change, I think, how my team works, and I've found that it it really frees up time for our team. You know, one task that could take one person, you know, a week can take now an hour. So, really, it's, I think very, very helpful, and I'm really excited to see what it what it does in all of your industries and all of your companies too. So I think it's gonna be a great journey. You know, Nic, employees can be resistant to new tools. Like, I had to kinda coach my team to use it and say, hey. You know, we're not gonna do this other thing. We have to use this. How how does Cohere support our customers with enablement and getting value out of this technology? Yeah. Yeah. No. That's a great question and and great example set of use cases. You know, and you're right. Because even the most powerful AI solution, it can fall flat if employees aren't set up to on how to use it effectively. And, you know, just as you demonstrated all the different ways that you and your team are using it, it it does start with with use cases. And and we do work closely with our customers to identify those areas where, you know, AI can make a real impact. So you gave us a few examples. You know, others were working with customers where, you know, we'll say, hey. Let's how do we help you reduce the time the analysts spend reviewing some of these forms? The 10 k's, the 10 q's for risk assessments. How do we help you streamline the commercial loan underwriting process or help automate some of the prep prep work that wealth managers or wealth advisors have to do before meeting with clients. They have to summarize customer portfolios. They have to do market insights reports. All of these now that improvements are they move to from, you know, nice to have that they used to, have to you know, these are high frequency, high effort processes that, we're able to work with our clients to automate. So when we're able to first start with a use case and then tie that use case to a success metric and measure ROI, you know, that's the first piece of of adoption and enablement. But again then again again, even the best use cases won't drive value if if folks aren't feeling confident and supported. And that's where the enablement strategy comes in. So, you know, we recommend that we work with our customers on a comprehensive, you know, digital and an on demand learning experience. It's designed for different roles. So you may have, you know, compliance analysts and financial advisors or a developer in your IT department that is supporting, your your line of business. And each of these personas then may be at a different stage in their own AI journey. You know, we may have someone who's brand new to AI or someone who's been experimenting with advanced prompts. So they act so for us, it's important to get that right content to the right level through a digital experience. Second, we do run a train the trainer program, for your internal champions. So these are those trusted team members who understand both the technology and the business. They help us and they help the organization demystify the the tools that, that are now available through AI, and then they're able to support the adoption from within the organization. And third, we offer white glove enablement experiences. So think about, you know, live expert led sessions, personalized coaching, hands on support that where we adapt to the different learning styles of of of our of our customers. So some folks are gonna want a technical deep dive, and API access. Others are gonna want, you know, single real world example, and it's important that we meet the learners, and users where they are. So that's really where the unlock happens. Yeah. Well well scoped, high value use case that's paired with a thoughtful strategic enablement. And that's what we have started to see, you know, AI become a a daily tool for people to, that they actually trust, and there are use cases that, gives them immediate value. And so, Laura, you you gave us great examples of use cases how your how your team is using Cohere to save time internally today. I don't know if there there's more you wanna add there or maybe talk a little bit about, some of the emerging capabilities that you're most excited about as as we move ahead. Yeah. I think I mean, I can build a little bit on the use case, which will just then touch touch into the emerging capabilities. So I talked before about how we were using our internal AI solutions to review expense submissions. They come over the form of email. We have to compare them to our policy, say yes or no based on, you know, the amounts and the other parameters that are in there, and then provide a response back to the customer. So, you know, maybe even four months ago, the response back to the customer was, handled handled by one of our accountants. And today, we are actually now using our our North solution, which is an agentic AI platform, to to draft that email. And then we still have our accountant review it first. Like, so we're still in the human in the loop phase, for us internally. Maybe soon we'll be out of it because it's it's doing really well and saving her a ton of time. But so then it drafts up the email, right, and gets it all ready for her to send, populates, you know, the subject line, all of that. And so it works, by searching both, like, our internal Notion pages using our Gmail and, you know, looking at what has been spent before. And so it really has been saving a ton of time and actually, like, moving us into not just, finding information and extracting information or even generating a response, but it's the workflow that is getting tackled that is what is most helpful. Because it's not just what you do within one tool. We found that that is helpful. It's really the ability of AI to support us as we move across different tools in our workflows on a a monthly basis, on a daily basis, kind of whatever it is. So, you know, I love the contract example because that's what I spend a ton of time doing. And so I just interact with North and say, like, hey. What's the clause in this contract that talks about x y z, and it can find it and pull it out for me versus me being like, is this the one that was executed? Like, let me check all the folders. So I feel like that's been been super helpful, and I really, you know, I just enjoy that I'm not spending time personally doing things that are or with that our AI can do now. Right? Like searching and reading. I feel like that is a large part of what AI is, solving for all of us, you know, human thinkers. So that that's been really great. And then, you know, when we think about, our the emerging capabilities that I was talking about, So earlier this year, Cohere launched an early access program with North, which you see kind of a snapshot here on the screen with RBC. We announced it publicly in January. It was incredibly exciting. I think both, ourselves and RBC are really, like, excited about what we're building together, to to streamline, some of their internal tasks and and accelerate how they're using AI. So that that's been really exciting. It's a all in one Secure AI workspace that empowers employees to, you know, do all of the work that I've just been talking about faster, more easily, more accurately. So for me, when I think about the finance use cases, it's about finding the information, the right information, you know, because I'm sure you all know. Right? It's like, okay. I need that particular piece, that's relevant to the inquiry that I have in mind, and then pulling it out, being able to analyze it, and put it in a a way that is is very easy to understand and act on, because I think that's kind of the point of of finance in general is getting the right information, understanding what it's telling you, and being able to act on it. So I think of, Agentic AI here as just sort of really taking what we're doing to the next level and making it easier and more impactful. And, you know, if you're in investments, you know, finding that edge over your competition and finding the trends that other other people aren't seeing yet. So, yeah, I think it it's pretty I think it's pretty exciting. So Yeah. I I'm equally excited about North and, you know, agents really do feel like, you know, this next frontier, moving beyond some of these use cases, where we're answering questions and, actually now taking it to the next level by making decisions and working across these systems. You know, Robin, you know, how are you thinking about agents at Cohere and the use cases for them, with our our our FI, customers and in industry broader? Yeah. So agents are super buzzy right now. I can help demystify them a bit, I think. Agents are really the next evolution of of RAG or retrieval augmented generation, where what you're what you're doing is grounding the output of a model on specific data or source material. So agents combine that with reasoning and with access to tools. In terms of tools, you can think of things like, like a web search tool or a Python interpreter if you want the model to, write some code or perform some calculation, produce a chart. And another type of tool is connectors to company wide systems. So think of the the applications that we all use every day, things like SharePoint, OneDrive, Google Drive, Slack, email, and so on, all with role based access control so that, employees are only seeing the the data and the information that they're, supposed to based on their permissions. And so agents just make it make it, you know, really simple to build out more and more complex workflows. So starting from, like, a very basic one, just imagine, like, a single document, question answering use case. Let's say I have to ask questions about a weekly economic report and and put feed that into some other downstream, thing. And and now imagine you have to do that as a regular part of the job of your job on, on, like, a weekly basis as these new economic reports are released. Instead of doing that manually over and over, even if you're doing it, with LLMs, with that sort of, question answering ability, you can actually set up an agent just to run that repeatedly either on demand, or, on some schedule you define, like, when that new economic report is released each week. And so you can see how also this framework sort of can expand as the workflow becomes more complex. Like, think about, an agent to do a more complex task, something like like plan my week. Like, there the agent might need to connect to my calendar, my Slack, my email, and do some reasoning among those different sources to sort of make a plan. So it's less about sort of a very discreet simple task with, clearly defined steps, and you're passing over a little bit more of the the reasoning or problem solving ability to to the agent. And so agents are always gonna follow that sort of, react style framework where they're doing, some reasoning based on the prompt or instructions they're provided. They're performing some action, and then they'll do that in a loop until they produce the the outcome, that the user is looking for. And so, as you scale these agents out from, more simple and discrete tasks to more and more open ended and complex ones, they really redefine how businesses, can work. And we think that, of course, this is really gonna produce the next step change in efficiency gain gains. Moving away from sort of simple, question answering or just, like, knowledge retrieval to actual and and, like, augmentation to real automation and freeing up, much more time for folks to focus on the things that are gonna drive more value for for their employer. And so employers can really use agents to anything from, like, conducting research and analysis on public materials across the web or, in global organizations, searching across documents in different departments in over a hundred languages, to drafting and writing reports and emails, data transformation, and some of the use cases Laura was describing, really performing increasingly complex tasks across different domains and doing so in a way that is, private and secure. So, maybe just grounding this a little bit, in in use cases, and I'll I'll go back to the example of RBC that that Laura touched on. So, RBC was already very sophisticated, when it came to technology and AI before they engaged Cohere. There's an organization called, the Evident AI Index that ranks financial institutions on their their readiness and sophistication when it comes to AI. RBC is number three on that list globally. But what they really found was they were looking to turn over more of the power to the subject matter experts in the different lines of business to be able to create, share and monitor these agents. And so they have used cases across every line of business from capital markets, risk management, personal commercial banking, technology and operations, asset management, And really what what North and, agents are gonna allow, deployed securely in their own environment is to allow them to, put more control in the hands of the subject matter experts in these different lines of business and in an, secure, safe environment to be able to no code, sort of create, share, and use these agents across the different lines of business. And so, we're really excited about that. Awesome. Thanks so much, Robin. Agents are definitely opening exciting new doors for the industry and the space. And, you know, that's a great place to wrap up. So to quickly recap, you know, financial sector's on the edge of an AI driven transformation, and those that are gonna see the most value are the ones who implement strategically. And, you know, that means balancing innovation with compliance, proving value through internal efficiencies, and, you know, planning for scale right from the start. So, you know, as we close, I wanna thank Robin and Laura for sharing their insights and expertise. All of you in the audience, we really appreciate you joining us. If you wanna learn more, I recommend downloading the financial services one pager. It's in the docs section of this, web chat and, you know, connecting with our team. So, again, really appreciate you joining us. Thanks again, and we'll see you next time.