Saris

Transcription:


Danial Jameel (00:10):

Hi everyone. My name is Danial; I'm from Saris AI. I'm going to make this to the point so you can go have your drinks. I know I'm the last one.
At Saris AI, we basically build AI workflow agents for banks and credit unions. Our primary focus is back office operations, which are the tedious, monotonous, complex workflows. As I think one of the judges put it, this is the unsexy stuff, but it is the backbone of the industry, and it needs to get done. That's where we help automate and increase the productivity of team members there.

So, two things to note before we jump into the demo. Number one, Saris AI is not a specific platform or a UI where users log into. We actually sit on top of your existing systems, so it's minimal change management, and you'll see what that looks like in a second. Second, we use a combination of different AI technologies. It's not just LLMs; we're using computer vision, machine learning prediction models, and more traditional classifiers as well. Our team has over a decade of experience. We publish papers, we have patents, so we kind of know what we're doing and we're having fun with it. With that, I'll jump into the first example. Here we go.

(01:27):

This is a workflow on a proof of insurance. This bank is using secured messages through Outlook. Again, we sit on top of your existing systems. Normally, when a proof of insurance or even a loan application comes in, the next step for your team member is they click on this email, they download the application, they'll do a quick check and review. If everything is good, they'll upload it to a specific system. If there are issues with it, they may need to do the verification. They will do a write-up, correct the application, send it back to the customer, and it kind of goes back and forth. But with Saris AI, all you have to do is click the reply button, and the magic happens. Now, all you do is click the reply button, and Saris AI kicks in. Now it's doing all of that heavy lifting.

(02:16):

It's using computer vision to download the document. It's doing the verification. It's updating part of the core or LOS where it needs to. If there are issues with it, it's going to flag those. It's going to do the write-up, and here you go. It's even going to attach the corrected document. In this case, this took about 23 seconds. "Dear Brianna, thank you for providing proof of insurance. Upon further review, the document does not mention collision coverage, and the expiration of the policy is not specified." A task that took, we actually timed this, about 15 to 18 minutes, and loan applications can take more than two hours, you can do this in a matter of minutes. What I want you to think about is what we call the future of work, which is humans and AI working together, where AI is doing the heavy lifting, and your team members are focused on the high-impact stuff. You want your team members to focus on the final review.

(03:09):

They can update this rather than going back and forth between different systems and documents. That's what's really changing with the use of agent systems. I'll move on to another example. This is an unauthorized transaction. Again, we get these all the time. "Here's an $80 charge from Walmart that I don't recognize." Your team member has to verify the transaction from the core, then create a dispute claims form, pull in the information, and then send it out to the customer for signature. Again, this takes about 15 to 20 minutes. Now, with Saris AI, all you do is click the reply button, and the magic happens. Saris AI is doing the review with the core, checking the transaction. It's even creating that dispute claims form. The idea here is that all your team member needs to do, again, is the high-impact work. Review this form, make sure everything is correct, send it out. Don't waste 15 to 20 minutes, or sometimes an hour. You can do this within minutes.

(04:13):

Moving on to a more complex example. This is an example from one of the top 10 US banks. It's part of the commercial funding team. They have to go through about a hundred or so data points to pass and fail specific documents. On average, this takes about two to three hours. We do the same job in about five minutes. A way to think about Saris AI, if you can see this little icon right here: it sits on your desktop. The analogy I like to use is back in the day, if you ever used Microsoft Word, it had a paperclip called Clippy. That was just there to help you understand and do everything in Word. It's the same way. Based on your role, if you're a lending officer or in the deposit team, you can click this. It'll show you the workflows it can automate for you.

(05:03):

You can assign a task and go about your day. So, basically, Saris AI is giving each of your team members an associate to whom they can give a task. It will perform those tasks and come back to you. It's really cool. All you do is, as I said, click. You can assign the specific workflow, and it even gives you a live audit trail. As it's running, you can actually track what the AI is doing. In this case, there's a whole range of different corporate resolutions. It can scan images, passports; there's a whole range of things it can do. You have this live workflow map that shows you exactly what it's doing each step of the way. Once it's completed, it will actually show you the pass/fail in this case. Another unique thing about Saris AI is that it gives you an audit trail for everything it does.

(05:51):

Now, you can just click on a highlight, and it tells you why it passed something. You have an explanation at the top, and it even highlights the exact data points that it used. So, your team members don't have to go through different pages, exactly the system you're using, exactly the process, but it's tracking everything. From our experience, most credit unions and banks don't have 100% coverage over their mortgages and their loans, and that always annoys auditors. But with this, every single variable can be tracked in real time. Your team members don't have to go through pages. It just highlights, it gives you the explanation, and so on. Again, this goes back to our thesis: this is the future of work in this industry. Finally, I'll show you another complex one. This is around UCC filing.

(06:41):

Just as a context, this organization had a backlog of about 700 and growing. We powered through that in less than two weeks. All you have to do is put in the documentation. For UCC filing, you have to go through hundreds of different bank statements. Each state has its own compliance. You have to extract certain information. You have to redact certain personal information. Then you have to issue the demand letter. Saris AI understands the nuances of specific state compliance. It will go through each and every specific bank statement. You can extract certain information that you want to update onto your core or specific system, and it even does things like redaction. In this case, I believe they wanted to redact the account number, and you'll see it right here, right there. So, it even does all of these tasks.

(07:32):

Again, the idea here is that as a human agent, you shouldn't waste time going through statements and so on. You should be focused on the final demand letter, make sure everything is correct, send it out. This is what I mean by the future of work, which is humans and AI working together, where AI is doing the heavy lifting. Thank you again. If you're interested, we have a little booth in the corner. Happy to welcome you and show you what else we can do. There are about 50 different workflows that we automate in the back office. Thank you.