Surveillance

Howard Konopka (00:03):

Hello. Welcome to this Watch It Work video on Enterprise Conduct, Smarsh's regulatory compliance solution for global enterprises. With new communication channels being adopted by your organization on an assumingly monthly basis, like Mobile, Teams, WhatsApp, Zoom, and many more, and with messaging volumes growing exponentially, legacy supervision solutions have reached the limit of their effectiveness. Enterprise Conduct, powered by Smarsh's, cognition AI machine learning engine supports both traditional lexicons and proven machine learning models that radically reduce the volume of false positives entering review and which also identify problematic communications missed by lexicon-only based policies.

(00:53):

Enterprise Conduct is one of the optional solutions of Smarsh's enterprise platform. Our fully-managed public cloud SaaS service that meets the communications capture, information archiving e-discovery and compliance needs of the most demanding and complex organizations.

(01:14):

But it's time now for a demonstration of Enterprise Conduct. After logging in, I'll select Conduct from the platform main menu. To keep this Watch It Work video of a reasonable length, I will focus on two capabilities, review workflow and cognitive scenarios. After selecting Enterprise Conduct, you can see application-specific options in the left panel menu, and their availability is determined by user role-based entitlements. Again, entitlements help to simplify the user experience to make it easier to learn and use an application. Since most users of Enterprise Conduct are reviewers, they won't see many of these options, but will instead access their assigned review queues, which is the default option on this menu.

(02:11):

But before I show off the reviewer workflow, I'll quickly show you a queue definition and how they are incredibly customizable. Queues are the main workflow feature of Enterprise Conduct, and they can be used to segregate review across lines of business with both common and separate supervision requirements. For example, if emails are exchanged between investment and consumer banking personnel, and if you set up separate queues as in this example for those two divisions, those emails will be reviewed and tagged separately.

(02:54):

Temporary queues can also be used to test policy changes on a production system without impacting ongoing review. I'll select one of my favorite demo queues to show you the various attributes, the definition, and in the middle of the screen, you can see how you select the employees or the departments whose messages will be reviewed. You can select participants or groups, which usually map to organizational departments. Very flexible mapping.

(03:26):

The policies used to generate alerts for the selected employees are picked on the right. If you add or modify a policy, you can immediately run it to generate additional alerts. And we, of course, have configurable sampling profiles since most customers supervise a random set of non-flagged communications. Employees can also be put on heightened supervision. And on the left are the workflow settings. Our queues support up to five levels of review. This test queue only has one level. And you can see that reviewers can be assigned with specific entitlements. [inaudible 00:04:07] could be an admin in one queue, but a basic reviewer in another. And lastly, customizable review tags can be created and selected on a per queue basis.

(04:19):

But I'll return back to the demo of the reviewer user experience. Go ahead and find my review queues, and I'll just select the first one. When you're in a queue, you can use search to identify all or some subset of messages that you want to actually start reviewing. There are extensive metadata and full tech search fields in this form with support for operators like wildcard and proximity. And you can see here I could search one or multiple full-text fields like subject body and file attachments.

(05:05):

But because I'm in a queue, there are queue-specific fields that I could also add to my search, a lot of them. Let's select a few like policies. I could say, "Hey, I'm interested in the messages that flagged on one or more specific policies." Or I could search for actions that were taken on messages in this queue by me or other reviewers in the queue. Or I could also search on any review tags previously applied. So again, you can run an ad hoc search. You can also save searches. And see, I've got two saved searches in this test queue. I'm going to go ahead and run one.

(05:47):

Results are displayed in a sortable list on the right, and also after running a search, whether it's ad hoc or a saved search, refined results brings up a faceted filtering tab. Filter counts, they're useful for data insight, but you can also filter on one or more values. For example, if a reviewer wants to review just Teams and Slack messages that flagged on a specific policy, well, you could add that criteria to the search. Or after a more basic search, you could apply a filter to come up with that subset.

(06:22):

I'll click on a row to begin review. And the matching lexicon terms of that policy, SD Smarsh test policy, are highlighted throughout anywhere in the search results, including attachments. And we're working on an enhancement to highlight entire sentences when a message flags on a machine learning model. On the right side of the screen, you can see various workflow options. A reviewer can close a message without a concern with one click, the close icon, or escalate it if they think they found a true violation. The update icon provides an annotation field and any queue-specific tags that were set up.

(07:07):

Lastly, the right most panel displays additional info about the current message, including which employees monitored in this queue were correspondents or participants of this message as well as its audit history. And you can see I can use these chevrons to go all the way back to the search screen if I desire.

(07:29):

Even with an efficient supervision workflow, customers don't want their reviewers buried reviewing thousands of false positives. So how we flag messages for review with our policies and scenarios is another differentiator of Enterprise Conduct. Let's select this policies feature, go add policy, and let's say we give it a name, trade rumors. And there is yet to select a policy category, and you have to set a policy type, whether you want to flag messages review or exclude them.

(08:09):

I think I'm going to also point out a feature unique to the Smarsh solution, echo cancellation. This option prevents about 20% of messages from unnecessarily entering review due to repetitive policy flagging on email replies and in social conversations. 20%, that's a pretty big review and cost time saver. Take, for an example, an account rep who sends an email to a client promising, "Hey, I'll triple your investment." Now, that email would flag in a typical guarantees and assurances policy, but if the client were to reply to that email, "Hey, that sounds great," you don't want that reply to flag in the original email content again. So that's what this option prevents. But again, you can turn it on or off for specific policies.

(08:56):

Administrators also need, when modifying or creating a policy, to specify the query criteria to generate alerts. And that's stored in an object called a scenario, and I can go ahead and select one from the dropdown list. But let me show you the UI where scenarios are created or modified. I'll go ahead and select it from the application menu. And you can see a little bit of a comprehensive user interface here. Oh, by the way, when I'm using the term scenarios, there are two types. Standard scenarios use a combination of lexicons and metadata query criteria, but cognitive scenarios primarily use machine learning models to generate alerts. And available models include guarantees and assurance, rumored secrecy, customer complaints, and many more. Customers subscribe to the models they require, which are constantly refined and updated by Smarsh.

(10:01):

So let me go through some of the basics of this screen, including, well, how do you create one of these cognitive scenarios? How do you potentially augment it with metadata and lexicons? But first, let me mention that here I have a number of test lexicons. And lexicons, of course, are a standard vendor feature with support for terms, names, email addresses and more. But I like to highlight our support for regular expressions, which if you're unfamiliar with them, they're a way to find complex patterns in a message like account numbers or PII like social security numbers and even emoticons. If I go down here, how about this one here? This says slightly smiling face. This is how you would find smiley face icons in a message, pretty common in social media.

(10:51):

Now, again, standard and cognizant scenarios, they are created pretty easily with a drag-and-drop user interface. Let me go ahead and, literally in about a minute, show you how you'd create one. I'm going to go ahead and give it a name and say, Watch It Work, trade rumors, create scenario.

(11:14):

And so let's think of some criteria. Let's say I'm only worried about messages with less than seven recipients. Well, I could find here in the metadata section the metadata field number of recipients. That can even be very specific, external, internal. I'll just stick with number of overall recipients and just drag and drop over here. And let's do less than seven. Save. So I've got my first criteria.

(11:44):

Now, let's say that I want to leverage Smarsh's machine learning rumor model, but I also want to use my own lexicon that I've highly refined, maybe picking up some messages that I want to make sure will enter review. So I'm going to pick the bullion or operator one up and drag and drop that. And then I'll go ahead and pick the machine learning model and find it on a dropdown list, rumor, and say, "Hey, if there's a match to the model, I want it to be true." If I want to flag it for a review, I can even set confidence thresholds and hit save. And I want to use my own special lexicon that I've highly refined to find rumors. And I'll just pick one off of a list, scroll down, hit save, and boom, I'm basically done.

(12:42):

The change is automatically saved and published in production versus other vendor solutions that sometimes require support tickets to make policy-related changes. By the way, Smarsh offers extensive training and advisory services to help customers learn all about policies and scenarios and the other great features of Enterprise Conduct. And it also bears repeating that this solution works consistently on the communications captured from 100-plus different networks and channels we support, including audio. One solution to help manage all your communications risk.

(13:23):

Well, the benefits of machine learning-based surveillance are somewhat under the hood and difficult to demo. So in just a minute, I want to quickly give you my personal top three reasons why I think our cognitive scenarios and machine learning models are really terrific.

(13:40):

Number one, our cognition AI solution is from Digital Reasoning, the compliance surveillance leader Smarsh acquired in 2021. We acquired both terrific natural language processing and machine learning technology, as well as highly effective surveillance models developed with and validated over many years by dozens of customers using their high-volume, real-world customer communications.

(14:08):

Number two, great models generate outstanding alerts. The volume of false positive hits is reduced up to 90-plus percent, and up to 300% more risky communications are identified. And this is true across all channels, email, social, audio and business collaboration, even with the content and style differences of those channels.

(14:32):

And my third personal reason, which I showed you in the demo, is that customers can augment the use of models in a scenario with additional metadata fields and lexicons to either A, exclude unwanted model results and/or B, include results not currently returned by the models against their data. This flexibility, the ability to augment, helps customers achieve desired compliance outcomes ahead of our continuous model updates. Smarsh's technology, experience and model maturity gives us a clear lead and advantage over other vendors. And again, that's been recognized by industry analysts.

(15:17):

I'll close this Watch It Work video with a nod to reporting. Enterprise Archive and Enterprise Conduct applications provide, again, various dashboards and reports to help customers manage their archive, review teams, policies and workflow. Defensible alerting and our governance reports help customers meet their internal standards and even prep for regulatory audits. There is also an outstanding feature to schedule reports I'll just highlight, schedule reports, conduct intel. I'll hit add scheduled reports, and I could select one or more compliance reports and select recipients and have them delivered by email on a regular cadence. There's no more excuse not to review reports ahead of a status meeting.

(16:11):

In this Watch It Work video, I focused on key capabilities of interest like search scenarios and workflow. Enterprise Conduct will lower your operational costs, reduce, again, false positives, wasting review time, deliver quality alerts to better identify and manage risk, and provide integrated surveillance across all communication channels. If I piqued your interest to learn more, the Smarsh sales team stands ready to meet with you at your convenience to discuss and demo Enterprise Conduct and the other solutions in our product portfolio. Thanks for watching.