KYC für alle: Sicherer Self-Service zu Kundendaten

Good morning. Welcome, everyone. Good morning, good afternoon, good evening, depending on where you're joining us from.

Welcome to the webinar. We're going to give it a few minutes for some additional folks to join. We see some joining.

So I have 8:00 AM in my time zone, 11 Eastern, wherever you're joining us from. So we're at the top of the hour. We're going to give it a few more minutes before we actually kick things off.

So we appreciate your patience. We just want to make sure we give everyone a chance to dial in.

So give us about two more minutes, and we will get things going. Hello again. Welcome.

I see a few more folks joining the webinar. Thank you so much. We're going to give it about one more minute, maybe two more minutes.

I see a couple of folks joining in. We want to give everyone a chance to get in. Don't worry, we have plenty of time to get through all the content.

But welcome. We'll get things going here in about another minute or so. Thank you, everyone.

Again, good morning, good afternoon, good evening, depending on where you're joining us from. Thank you to those that are joining the session. We're going to give it about 30 more seconds.

I see a couple of folks joining, and then we will kick things off. Don't worry, we have plenty of time to get through the content, and I think we'll actually get you all out of here early. So again, we appreciate you joining, and we will get things going here very quickly.

All right, team. Welcome. We're ready to go.

I see we've got our last person or people joining. Good morning, everyone. Welcome to the webinar.

Excited to have you all here from-- Or good evening, wherever you're joining us from. I want to cover a couple of things before we hand it over to our presenter and the rest of the panel. So today's agenda, I'm going to cover a couple of housekeeping items, just so everyone's familiar with the webinar questions, how to answer questions and things like that.

I'll cover that in a minute. I'll then introduce our team. You can see some of them on the chat or in the window now.

We'll cover those introductions so you understand who's here. And then we'll get to the heart of what we're here to cover, which is our Actian AI Analyst. Give a quick little overview, and then we'll get into the demo.

We'll wrap the call with some call to action. We've got a little homework. Kidding, we don't have homework, but want to make sure that you're able to get all the information that you need on our Actian AI Analyst, and then we'll go through and we'll answer any questions that you all have during the course of this webinar.

So let's jump into the housekeeping items. So for those of you new to Actian webinars, we're going to keep you all muted, unfortunately. So the audio's muted, don't worry if the audio isn't working.

This is a Zoom session. You can ask questions throughout the course of this. We'll take questions.

I've got a panel standing by to answer your Q&A. We do have a Q&A window for those new to Zoom. You can get to the Q&A window through the menu at the bottom of the screen.

You should see three dots and a More words. You can click on that and open up the Q&A window. You can type your questions live throughout the course of this session, and we will be here to answer.

We are recording the session, and that recording will be sent out afterwards. We're going to run maybe 45 minutes, probably a little bit shorter than that, so there'll be plenty of time to get through questions, cover the content, and give you back a little bit of time in your day. And again, all the resources that we cover, the recording, any additional materials, we will make those available to the participants, the attendees after the session.

And then one last quick thing. If anybody has any technical challenges or need to get in touch with me to help you through those challenges, please either use the chat capabilities at the bottom. You'll see the chat.

If you cannot, there's my email address in order to reach out to me, and I can handle or address any of your questions. All right. Let's meet our team.

So today's speaker, I'm excited to introduce Alex Seidler. He will be presenting, and he's going to be supported throughout the course of this webinar by two amazing sales engineers. We have Phil Wall and David Neideffer manning our Q&A chat window.

So again, any... questions that you have or anything that comes up, please type your question into the Q&A box and we will be sure to address those questions. And so without further ado, I'll kick things over to Alex.

Alex, take it away. Thank you very much, John, for the introduction. And good morning, everyone, or evening, time zone.

And thank you very much for joining today. So over the last few years, we have seen an explosion of interest around generative AI and AI assistance in enterprise and daily life. Today, almost every platform claims you can simply chat with your data, but in highly regulated industry like banking, the real challenge is not generating answers.

The real challenge is whether those answers can actually be trusted. That's especially true for financial anti-money laundering and know your customer investigation, where business definition matter, governance matter, lineage matters, and also context matters. So today, we want to show you how a compliance team member could accelerate their investigation with the Action AI analyst, which is enriched with trusted data, business understanding, metric definition, governance details, and also the context of our business.

So imagine I'm the head of financial crime compliance working for a retail bank. Tomorrow morning, we have a compliance committee review ahead of an incoming regulatory audit. One of our analysts identified something potentially concerning.

Several customer classified as low risk and are generating transaction activity close to or above suspicious activity thresholds. Now, in theory, this should not happen. If low-risk customers suddenly behave like high-risk customers, there are only a few possibilities.

The classification is outdated, the onboarding data is incomplete, the enrichment process failed, or suspicious behavior is emerging earlier than expected. The problem is that investigating this usually takes hours, sometimes days. So the teams need to navigate multiple dashboard, query multiple system, reconcile conflicting definition, and manually piece together the investigation.

So let's see how our AI analyst agent can help on this one. So let's go through the demo. So what we can do right now is going on our demo.

So pretty simple here. Starting our investigation, we need to better understand what customer are classified as low risk, but are exceeding the suspicious activity reporting, SAR, for example, threshold. So let's ask to our agents here, for example, do we have, here we go, a new task.

For example, here we go. Bank analyst, and I'm checking my agent right now because we can have multiple analysts, and we'll get back later on that on how we can produce that. But just starting investigation, we need to better understand what customer are classified as low risk.

So turning to the Action AI analyst, we can ask this question: Do we have customer classified as low risk who exceed the SAR threshold in the last 30 days? So what I will do, I will just enter this prompt, and as you can see, my AI analyst is working for me. So it's creating a plan for its analysis, and we'll step through it to address my question.

This for covering our data, aggregating it, gathering customer level data, create charts, and build out a compliance report. So now this is what we need AI to do, not just search raw tables guessing at what we mean. We need AI that can understand our business context.

What a low-risk customer means, risk factor inferior 12. How our SAR exposure is calculated, what threshold definition apply, and which trusted data set should be used. Now, this analysis might take a minute or two, so we have pre-generated here.

So let's see the results. And as you can see, we have multiple prompt to get the funnel. But just to show you, if you check the executive summary, with Action AI analyst, we are not stuck finding data, validating the right data across rule and regulation.

We are not joining tables and writing SQL. Our valuable time is applied to reviewing result and next step. And that's pretty interesting because as you can see, you have your executive summary, all the key metrics that answering my previous question, all the risk factor distribution, notable analytic observation, and recommended compliance action to help you to get to the next step.

And you can have dedicated information like dedicated report, et cetera. And everything was created in two minutes because if you go back here, if I go back here, for example, to get back to the result on the first low-risk expectation, as you can see, it's starting to create my report, and I'm going to see the same resultThan before, because everything is about context, et cetera. And as you can see, it's writing and it's doing the job by itself.

So this is where governed data in context becomes critical. Context is the understanding of the business, our definition, our KPIs. Without governance and AI, our LLM could easily use inconsistent definition, query the wrong data set, produce conflicting interpretation, or straight up hallucinate a whole analysis based on something it's read off Reddit.

And as you can see, it's exactly the same result than before that I pre-generated, and everything is pretty well. I'm not cheating. It's demo live.

So let's go back on this one. Or let's keep in this way, because I'm going to keep prompting. So this is why organizations are leveraging Action AI Analyst.

We gain our user trust because our answer analysis and reasoning, it's all based on your business trusted definition, its governance data, its KPI metric definition, and the most important, its context. So now that we are starting to see some concern in our customer base, let's drill down a bit further from detecting issues to prioritizing our investigation. So turning to the Action AI Analyst, let's ask him this kind of question.

Show me the top five customer by ID with the highest mismatch between KYC classification and transaction behavioral. And so he has memory, so we will keep the work to get the final root cause analysis, et cetera, to get the right decision for you. So as you can see, the AI analysis agent get right to work analyzing our data.

It run two queries simultaneously using SemQL, an abstraction layer over SQL, to address the question. First, as you can see here, it pulled the full customer level details with a computed mismatch score, KYC TL label, and how far above the 10,000 SAR threshold each customer sits. Second, it prepare a ranked data set specifically shaped for the chart visualization.

It then present the result at both as chart and details table. Let me show you the following because I pre-generated here, and basically, you're going to see all those two charts. Now that it's turned, we can see applied a clear methodology to rank the customer.

All right, here we go. If I go in the mismatch score methodology, what the data tell us. So the mismatch score equal SAR amount multiply 12 under risk factor.

This compare each customer KYC classification against their 30 days SAR exposure. All five share risk factor one, the lowest clearance level standard due diligence only. No enhanced due diligence in place, and every one of them has still breached the 10,000 SAR threshold, with score ranging from 121,797 to 121,997.

Now, this is what we need AI to do, understand our business context, not just at raw table. Low risk customer risk factor below 12, as you can see here. No EDD required.

SAR exposure, three-day rolling total pulled from our pre-aggregated roll-up model. Flagging threshold 10,000 applied across all the risk tier. And also trusted data set, the gold layer transaction roll-up, not raw transaction record.

That's amazing because instead of reviewing hundreds of potential alerts manually, I can immediately focus on the five customer who's the strongest depreciation between their official KYC classification and their observed transaction behavior. Now, we can see this customer warrant enhanced due diligence. If you keep on going like that.

That's pretty interesting because of what appeared to be structure pattern at the customer level, which should be logged with our financial intelligence unit. These are actionable insights that I have been able to gather with Action AI Analyst in minutes. Again, this is not AI gathering data.

This is AI interpreting business meaning, determining analysis methodology, and revealing discrepancy based on our business rule and processes. So let's keep on going. I wonder is there a specific channel that customer are using based on this activity?

So let's ask him a question to keep on going on the information. So as you can see, right now, it's still working and presenting the result, and it will present me the result. So waiting that, I can ask him what transaction channel are driving the exposure.

So as you can see, sometimes it could take some time because more you will have data, more you will need to get some results. So we'll get back later on how we can improve this kind of process and how we created this kind of result. But let's do it.

Let's keep it finishing just for the demo. Let's do it the first. What transaction are driving this exposure?

And so the Action AI Analyst automatically correlates this customer and this transaction based on the different channel we have in banking, likeATM cash deposit and withdrawal, wire foreign out and in, Zelle debit card, et cetera. And as you can see here, I have all those information. It's all correct information about their standard banking activity, no retail purchases, no bill payment, no standard debit card transaction.

So standard transitional dashboard are great at displaying key metrics, KPI, and insights, but they only addresses the question they were built to answer. When business user or analyst have deeper questions or want to see different perspective on the data, they are forced to request another report to build, download data, and DIY the analysis. Analysis is rarely linear.

This is what makes conversational AI so powerful because it enable the user to ask the question that have not been answered yet. And so the last thing that I wanted to show to get the root cause analysis, because okay, we make some search, we have some information quickly than expected because what I'm showing you today, it could take weeks to get those kind of information trusted. Now, in less than 10, 15 minutes, I did everything.

Now let's go deeper on the search and said, okay, but why? Why it happened? So we can ask why were this customer initially classified as low risk?

This is where most AI system typically fail. This question blends to predictive nature of AI with deterministic reasoning. Action AI analyst is able to address my question because it has to govern context of my business, like our onboarding policies, our KYC, AML definition, our enrichment process, our enterprise data environment, our metrics and KPI definition, our business glossary, and our business knowledge and context.

So that's why here you can get this kind of result directly. So now reviewing the result, I see some of the factor that the customer are all US-based, as you can see. They are using retail account, and this customer were onboard on 2021, 025.

What this mean for the investigation? The KYC model might have been exploited. 2022 and 2024 onboarding clustering around 23,000 dollars on that.

So that's pretty interesting here. $20 per year, suggests a sustained multi-year enrollment operation rather than one time event. So sorry, it's not dollar, but people.

And there might be missing KYC classification rational field in our onboarding model. So with AI action analyst or Action AI analyst, our AI analyst, it's not telling us this customer looks suspicious. It's actually explaining why they are within our business environment, our business context, and based on our business rules.

It's this enriched context through governed data and knowledge that set the Action AI analyst apart from a generative AI system or a homegrown solution. This is why enterprise are looking to Action AI analyst to be trusted enterprise-wide data analyst system for conversational analytic. So if you want to see, I'm still working on-- it's still working, so you can see that I have exactly the same result for this first question, and I can ask him the third question.

We'll not have the time to go through the last question, I'm very sorry. But anyway, you have seen the result, and just to show you that it's exactly the same kind of result. Because it's not hallucinating on what I've explained it to you just before.

As you can see here, wire transaction channel are driving the exposure, and here it's exactly... Let me show you. Okay.

Show me the top five customer. It was this one. Sorry.

Here we go. So we have the same information. There you go.

So it's not inventing any kind of information. So now, so how does Action AI analyst become a trusted and informed SME on our business? How does it understand our business, our regulation, our rules, our data?

That's where we have a dedicated interface called the studio, and the studio come into focus, and we set the context for our system through our semantic model. So let's go in studio. So Robi Agent.

And as you can see, I'm on the administration page. The administration page, this is where I can create multiple agent, because right now it's on the KYC, AML, but I can create multiple agent for any kind of business or anything else that I need. That's pretty interesting.

So let's focus on the bank analysis. And here, the Action AI Analyst Studio set the foundational context for our conversational analytics system through a few different perspective likeThe instruction. Here we go.

So you can give him multiple instruction, and those instruction give your analyst context and guidance for better analysis. It will help you to get the right results. More you give you instruction, more the result will be better.

We have also the models. This is where we are connecting to your data, its dimension, and measure. So you can select all the data where you are connected, and we have dedicated connector for that.

So you can see that I'm connected to my Snowflake right now on those tables. It's pretty interesting. We have, sorry, the metrics.

Those metrics is here to understand your business, your calculation rule, your business rule, your definition about, to understand the metric, and this analyst can use all those information. The last thing, pretty interesting, is the glossary, because the glossary is here to enrich its context with your glossary term that the analyst should reference. So it's speaking the same language than you, and you can configure anything that you want.

And the last thing is you can manage your access because as you can see, the AML/KYC, sometime people need certification to get access to those kind of data, so you can create agent that have access to specific people. All those combined enables our Action AI Analyst to understand our business and data semantics model. It's this information that enable our conventional AI analyst system to understand our business, concept, and relationship, or more simply, our context.

This is what allow our AI to reason with business meaning rather than simply generating SQL queries. Now, what we have shown so far is the result of a well-governed semantic and analytical foundation. But at this point, most organization usually ask the same question.

How difficult is all of this to implement? Because traditionally, building semantic model, glossary, relationship, governance mapping, analytic context, and AI-ready business layer requires significant time, expertise, and manual configuration. And honestly, that complexity has historically been one of the biggest barriers to enterprise AI adoption.

This is exactly why we introduce our Steward agent, and as you can see here, it's another agent. It's the agent in the agent. Rather than manually configuring every layer of scratch, organization can now describe the business problem they want to solve, and the Steward agent helps accelerate the setup of the semantic governance an analytic foundation requires behind the scenes.

So instead of showing you another static configuration screen, I'd like to share a short video, one minute and 30 seconds, to show you the power of this Steward agent, and how you can solve the problem, and how we can build this faster and with efficiency. So let's do it. In this example, we start for a non-empty environment.

The user simply describe the analytical use case they want to enable, and the Steward agent assists in connecting relevant data source, identify business concept, proposing semantic relationship, generating governance mapping, and preparing the AI-ready analytical context required for investigation like the one we demonstrated today. The goal is not to replace governance team. The goal is to dramatically simplify and accelerate implementation.

So I'm letting you see the result of how we can build in five minutes, 10 minutes, maybe, a whole semantic model and starting to generate value in your context and in your company. And now you can connect to the model. It will create the semantic.

We already created the semantic model, and as you can see, what are my sales growth? I query the things, and I get all the information, and I just publish for my end user. Really easy.

So what we showcased today, let me represent that. Here we go. So what we showcased today was not just conventional AI analytics.

It was trusted AI analysis and reasoning powered by governed enterprise context. Within hours, Action AI Analyst can be the foundation to your AI strategy for enterprise data and analytic. A trusted AI, this is not simply generative, but deterministic, which combine semantic understanding, governance, lineage, trusted data, and AI-assisted stewardship.

This is the Action AI Analyst, the future of enterprise AI for data and analytics.Thank you very much, everyone. I'm letting the lead to John, just to presenting you the rest, and we will go move to the Q&A just after, if you want. Great job, Alex.

Pretty cool stuff. I know there's a couple questions coming in. We got the team standing by to answer those.

Again, if you have any additional questions, feel free to type them in the Q&A window, and we will actually ask Alex live some of these questions that are coming in. So next steps or calls to action, right? I joked around earlier about some homework.

But really, if you need more or additional information, the entire team is here to help. So I wanted to let you know myself, John Lysenslee, my email address. If you've got any questions, want any additional information, please feel free to reach out to myself, Alex, our presenter, his email address.

We also have Phil. Again, thank you, Phil for, I think I saw you answer one of the questions that came in. So we've got Phil on tap.

And then we've also got David Neideffer here to address your questions. So as anything comes up, any additional information you need, please feel free to reach out to us. Next slide.

So what else can you do? So this is a three-part series. Each part is meant to stand on its own.

Today was the first session focusing in on our Actian AI Analyst. We have a couple of other sessions coming up. Maybe you've registered for them, maybe not, but later this month, in June, we're going to focus in on understanding your customer.

But how do you build context around that? We're going to have Betty take us through our Actium Data Intelligence platform. And then later in July, Scarlet is going to focus in on data observability.

So if you haven't attended, or if you haven't registered, please feel free to register for those events. And then my last call to action before we get to some additional questions, next slide, Alex, is as you noticed, our Actian AI Analyst is extremely easy to use. There's tons of information on our website, and we are also offering you a free 14-day trial.

So it's very easy to go to the website, sign up for the trial. You get 14 days to leverage the technology. And I think you'll quickly see how it's very easy to start working with your semantic model, and start using your analyst to ask real-world questions and get back real-world information.

So that trial is available to everyone. So that's our call to action. Finally, next slide is going to be Q&A.

So Alex, I think we do have a couple of questions, if you don't mind answering those. And again, for those participants that are on the webinar, if you have any questions, please feel free to ask those. Alex, here's one that came in.

Looks like it's around databases and data warehouses. It says, "Which databases and which data warehouses does Actian AI Analyst support?" We have a pretty broad coverage, like Snowflake, BigQuery, Databricks, Redshift, Azure, Synapse, Microsoft Fabric, so all the modern data stack, and relational database like PostgreSQL, MySQL, MariaDB, SQL Server, ClickHouse, MotherDuck, and even CSV files.

And also our own Actian platform. So yeah, pretty broad coverage. Awesome.

And we are keeping developing new connectors. That's great. Thank you.

Again, to the attendees, any other questions, feel free to type those. Looks like security is always top of mind, Alex, I think. This one is around is our data safe?

Does the data, or does it leave basically our four walls, right? Does it leave our system? Yeah.

So, first question, is our data safe? Yes, of course. And does it leave our system?

No, of course. So your data never gets copied to Actian system. The AI provider only ever sees the specific rows returned by a query, not your whole database, and nothing is used to train AI models.

So connections are read-only and TLS encrypted, and the platform is ISO 27001:2022 certified. Great. I think I touched on this very high level when I was talking about the 14-day free trial, but a question around value, right?

How quick do I get to value? So the question is, how long does it take to set this up, and when do I start getting value? Very good question.

So the only setup is to connect to your data source, then build the semantic layer. And as you can see, we did it in five minutes, without the acceleration of the video, through the Steward AI agent, because it will help to automate everything. So for a focused use case with a clean data source, teams have gotten their first AI analyst live in a day or two, maximum.

And more complex environment with many data source take longer, but the Steward agent significantly accelerate the semantic layer build. So pretty easy to generate value in less than one or two days. Great.

Thank you. Thank you. Again, any other questions?

We've got a few more minutes. There's some nice ones. Give me a second to read through this.

Actually, I'm going to read this one first, and then we'll get to that. Question around who's this really geared towards, Alex? So who's this for?

Is this for data teams? Is it for business users, or is it really for both? Who would benefit from leveraging something like Actian AI Analyst?

I can say clearly it's for both, and they have different interfaces. Data team use the studio to build and maintain the semantic layer and configure AI analyst, multiple AI analyst. And the business user use the explorer to ask question in plain language and get charts, report, and general insight.

No SQL needed. You can also embed the analyst in your Slack or Teams. So if you don't want to use a new application, we have a dedicated integration with Slack and Teams.

Just ask him, and ping your AI, and it will answer it. That's great. That is great.

I think that I see Phil's addressing the last, or one of the last questions in the Q&A. Again, if there's any other questions, we're here for maybe the next couple of minutes. Otherwise, like I promised at the beginning of the session, we can give you back a lot of time.

Let's give it another minute or so. And again, for those that might have asked your question and I might have overlooked it, I've been trying to man the window. I know I've got Phil and David manning them.

But if we did happen to miss your question, we will be sure to follow up and address those questions in the materials that we provide. Like I said, we'll provide the attendees with a number of resources, including the link to the recording, and we will address any other additional questions that we may have missed. I see Phil's in the process of still answering Yeswant's question, so hang in there, Yeswant.

The answer is coming. It's a good one. I got to type it, though.

Well, you can address it live if you'd like as well. So- Yeah ... we've got a little time.

Go for it, yeah. All right. So this is a multi-part question that came in from our attendees.

It was a nice session. One of the questions was: How does Actian AI Analyst automatically discover and classify sensitive data across structured and unstructured sources? So this is a multi-part, and this is only part one of a few parts to this question.

So to start off with, Actian AI Analyst is really looking at your structured data, right? So we're not necessarily hunting through your systems for sensitive PII, GDPR-like information. When Alex was showing the build-out, our glossary and our metrics definitions, those can help to inform the agent about those policies, definitions, and calculations if you end up using them.

But it is against more traditional structured data in that respect, not unstructured. We're not reading PDFs and trying to figure out if it is or it isn't classified data. So that's part one.

Additionally, does the Actian AI Analyst platform provide autonomous or AI-powered data quality agents that can detect, assess, monitor, and remediate data issues? Actually, John had talked about some of our other upcoming webinars that are coming up. We do have an Actian data observability webcast coming up that will talk about how we address data issues.

Actian AI Analyst works in concert with other platform products that Actian offers. We can also be standalone as a point solution if you need it. But we are consuming that business context that helps us provide the trust by using governed data, and we have, in our product suite, tools that help to establish your data governance environment, the framework, and to document the additional context.

Those things get reused throughout our platforms. So- Right ... it was a longer one to type.

Sorry, I can't type that fast. That's good.

Listen, hey, that was a good one to address live, Phil. As you said, that was a pretty detailed one. Okay.

Anything else? Any other questions from the attendees? Alex, great job on the presentation.

Phil, David, appreciate you manning the Q&A window. For the attendees, hopefully you found this worthwhile, are excited about the Actian AI Analyst. I encourage you all to go out, leverage that 14-day free trial.

Register for our upcoming webinars. Reach out to us with any sort of questions. We'll give it one more minute to see if anything else comes in.

Otherwise, again, we appreciate the participation, and we can give you back some time in your day. Thank you very much, guys. Great.

Thanks, everyone. We appreciate it. Thank you all.

Have a great rest of your day. Bye, all.