Sponsored by: ServiceNow | The Patchwork Enterprise is Holding Your AI Agents Back

Summary

ServiceNow presents its unified data fabric, which connects to over 450 enterprise systems without copying data, as the foundation for closing four enterprise AI roadblocks: fragmented data silos, governance gaps, missing context for AI reasoning, and the insight-to-action gap.
This video introduces ServiceNow's autonomous governance capability, which keeps data trusted and discoverable for AI agents operating at machine speed, and shows a zero-copy connection to ERP systems and Databricks for advanced analytics workloads.
ServiceNow positions itself as an AI control tower for business reinvention, combining Now Assist, Moveworks, and industry-specific workflows to move organizations from AI-assisted toward fully autonomous workforce operations.

Chapters

FAQs

What four AI roadblocks does ServiceNow identify for enterprises?

This video identifies four roadblocks preventing enterprises from realizing AI potential: data silos that fragment information across hundreds of applications, governance that cannot keep pace with agents acting at machine speed, missing context that prevents AI from reasoning accurately across systems, and the insight-to-action gap where analytics never reach the point of workflow execution.

How does ServiceNow's data fabric connect to Databricks without copying data?

ServiceNow uses a zero-copy connection approach to integrate with Databricks and ERP systems, allowing data to be enriched, harmonized, and made available for AI reasoning without replicating it into another store. This video shows how Databricks can serve as the analytics and AI layer while ServiceNow provides the workflow execution and governance fabric on top.

What is autonomous governance and why does it matter for AI agents?

Autonomous governance in ServiceNow's framework refers to governance controls that operate at the speed of machine execution rather than requiring human review for every action. As AI agents act across enterprise systems, autonomous governance keeps data trusted and discoverable in real time, preventing agents from operating on stale, ungoverned, or improperly accessed data.

How does ServiceNow help enterprises move from AI-assisted to fully autonomous workflows?

ServiceNow describes a spectrum from AI-assisted workflows—where humans use AI suggestions to make decisions—to fully autonomous workflows where agents act without per-step human approval. This video positions ServiceNow's platform, including Now Assist and Moveworks, as the AI control tower that manages this transition by providing the data fabric, context engine, and workflow execution layer that autonomous agents require.

Full transcript

[00:08] Thank you all for coming in. Appreciate it's late and we have a cozy crowd here. My name is Juan Cicada. I am the principal data strategist in the new data analytics business unit that Service Now has. Uh raise your hand if you're a Service Now customer. All
[00:21] right, raise your hand if you know that Service Now is in the data analytics space. Yeah, probably don't know and that's why we're here. So the main takeaway I want you to all have is that service now is also in the in the data analytics space. I'm going to introduce
[00:33] to you what we're doing here. Um just quick legal note have to put this in there. Uh I'll I'll be talking about what we have but also a little bit about roadmap. So just a heads up on that. Um so one to to kick off here is that look
[00:47] every or has been investing in AI and everybody's expecting to provide the ROI. Now, there's kind of these two sides kind of the way we see this is that uh you have your AI assisted, which is kind where a lot of people are
[00:59] starting from, but we really want to be able to get to that fully autonomous workforce. And that's where we're that's what we're doing with with Service Now. Uh Service Now, we are the AI control tower for business reinvention. And what
[01:12] I want to introduce to you to a what service now is traditionally people think about service now as just an IT but I want to share with you high level kind of our our market on what service
[01:24] now is a service now AI platform. So if we start at the top we have service now auto which is a a combination of our now assist and our recently acquisition of move works. It's your front door. You're open to all your employees and users.
[01:37] For example, data bricks, they actually have they actually use service now auto move works. They call it internally. I think it's called R2-DB. So service now. So data bricks is actually a customer of
[01:49] service now using this. We have workflows for all industries and we have workflows focusing on IT on CRM security and risk employee experiences and entire app development framework. So when you think that service now is it we do way
[02:03] more than it and we have these main pillars sense because we want to be able to bring in any data decide so you can do reasoning on top of any model and act so you have on any type of workflow we have a security for any type of system
[02:17] identity uh we made acquisitions recently around uh Armis and Vza for security we connect to run on any cloud on bring in data from any system of record um so but what happens is when
[02:29] Start thinking about this autonomous workforce. We talk about autonomous IT, HR, CRM, risk and security. These are the areas that we focus on service now. So again, we're not just it. We have all
[02:41] these different areas and data and analytics is part of every single element, every single part of these workforces that we're doing. So we built data analytics not just as a standalone data product, not as an add-on, but as
[02:54] really the foundations to make all these autonomous workflows possible. So what I want to introduce to you is what we're doing here at at the data analytics side. Um but we know that when it comes to autonomous it really can't work.
[03:07] We're seeing these kind of high level four high level situations is of course data is trapped in so many silos. I think we've say this over and over again for decades and decades. Number one with so much data we able to we have lacks of
[03:19] controls and and governance. Of course now context is the key word we hear all the time. AI all needs to have this enterprisewide context. And the fourth and a really important one that I want
[03:31] the data analytics community here to really understand is that insights need to reach where the work is happening. And what happens a lot inside of the data analytics world is that you think we think that the goal is to bring data
[03:43] together and generate insights, generate your your your dashboards and generate your machine learning models to find your churn risk or whatever. But so what at the end you need to drive action with those insights and
[03:57] we see four capabilities that are required to be able to address these particular issues. So we need to be able to connect to any data source uh which is which is in reach. Second we want to
[04:10] be able to control that data so we can provide trusted governing governed access to data so we can make that data AI ready. But third, your da your data needs to have all that context that
[04:22] semantics so AI can be able to understand that. And finally, we need to converge so all those decisions happen where the work is actually happening. It doesn't just end inside of where your analytics is. And that is a service now
[04:36] data analytics stack what we're going to be presenting here. So this autonomous enterprise needs all of these four things working together. And we have two product lines for service now data analytics. Our workflow data fabric is
[04:48] how you're able to connect to any data through zero copy. We'll be able to talk about this stuff. Control the data catalog and governance contextualize to provide more semantics and with our database Raptord Pro Raptor DB Pro be
[05:01] able to converge all this particular data. So let's go through each one of these. So connect for decades we know that every system right every organization has all these silos data is at probably your Salesforce and sales
[05:14] data. You have workday for HR, you probably have SAP for finance. So you want to be able to connect to all these particular data sources that you have. So our connect layer part of workflow data fabric, we have uh integration hubs
[05:28] and spokes where you can actually now start connecting to any particular system through APIs and not just read their data but actually have actions to be able to go right back to any of those systems. External content connectors, so you can be able to connect into any
[05:41] textual type of data to be able to kind of bring that in. Third, stream connects. So, anything that needs real time uh uh events, we have over a hundred different connectors for zero copy connectors. So, you can connect to
[05:54] data that's in data bricks and keep that data where it is in the zero copy fashion. We have an entire we've introduced MCP registry. So, we can be able to register third party MCP servers. And we also have MCP clients, different ways to be able to connect to
[06:06] this data. We also have an expanding network of what we call the workflow data network where we have all of the outofthe-box integration hub spokes to all these different systems already
[06:18] external content connectors to bring in data to documents and stuff from shareepoint from Google drive. We have zero copy connectors to so many of these different sources of data obviously including data bricks partnering a lot
[06:31] when it comes to enrichment harmonization integration also for data quality and we have what now our passport program which is uh because we're in a consumption mode you can any of the tokens that you have uh inside of
[06:43] service now you can use that to consume also any of our partners. So this is the connect phase of it. Now control well you connecting the connecting to the data is not the same as making that data AI ready right we
[06:56] need to be able to provide all of that catalog and that and that and the the governance around that one of the things I'm super excited is that we announced uh a month ago our service now data catalog and personally I came so service
[07:11] now acquired uh my our company data.world world last year which was a data catalog and governance platform uh that I joined. So I I was a principal scientist over there where I've been I've been working on semantics and
[07:23] ontologies and knowledge graphs around 20 years now and everybody's been talking about this. I'm really excited and so this is now in GA inside of service now. So within the data catalog you can start defining your uh we have
[07:35] connectors to be able to bring metadata from all these different systems. You can start building your data products connect your data through that connection to zero copy. You define all your metadata, define your data contracts in the data products and then
[07:47] we actually make use of all this metadata, right? Activate that metadata for data discovery for search and you create your data marketplaces, create your lineage of all your data, all the standard features that you want inside of a data catalog is part of service now
[08:00] and be able to go use your agents, your workflows and do analytics over all this metadata. Now, one of the things that's happening that AI is making it so much more complicated, too, is that there's a lot more governance that needs to be
[08:13] happen. And making our data AI ready is not just a one-time cleanup. It's a continuous approach. And that's why we're introducing autonomous data governance where we'll be able to have autonom automated uh profiling when it
[08:26] comes to data quality uh real-time pipeline monitoring for observability being able to do a lot of integration and in combination with different data uh uh harmonization and also be able to define your policies and be able to go
[08:39] execute your policies all within Service Now. Again, when it comes to policies and workflows, you want to be able to leverage the power of Service Now from the workflows. Now next is the contextualized layer. So
[08:51] even the clean governed data is not enough for for AI and that's why we need to be able to provide all of your understanding your semantics of your business. Be able to bring in your processes, your policies, your relationships, your metrics, your
[09:04] dependencies, your decision histories around this. And that's why uh in May we introduced the context engine. The context engine for uh the context engine is this living graph or think about it
[09:18] almost as the graph of the different graphs of everything that we have understand about your business. So we already have been building out a knowledge graph. Uh so inside of service now if you have your CMDB right you're
[09:31] already bringing in all of your particular uh registry your items your assets your business uh uh your how everything is connected through the data catalog you can define all your semantics define your ontologies and bring all that semantic layers inside of
[09:44] knowledge graph. Second we have all your users we understand your employees and we can put this all together. Third is with security and our investments with VZA and and Armis. We understand now all
[09:56] the security your identity management and finally the decision graph. This is where we have your decision traces. You are running your workflows inside of service now. We understand what has been done and we can we do process mining and
[10:09] be able to define what these decisions. So when you start combining all of these things together we have a very powerful the powerful context about your business. So the typical example I bring up is you say when somebody ask a
[10:21] question about a customer you can say well what do you mean by customer we have all these definitions about customer which comes from the knowledge of the semantics but we know that this particular user is probably from the marketing department therefore we know
[10:33] they probably care about the definitions of customer from the marketing and we know from the security if they actually have access to the data or not around that and we know what decisions that what what what other people have been doing with that data make suggestions of the types of decisions that could be
[10:45] happening. So this is that that context engine that we should be that we're able to go do very powerful to provide all that context for AI. Now all of this which is built inside the context engine is an recent
[10:58] acquisition that we also did with pyramid analytics. Uh pyramid analytics is a business intelligence tool that we acquired a couple months ago and that is now part of our autonomous data
[11:10] analytics which one of the leading semantic layers implementations of this. What's really interesting is that if you look at the Gartner Magic Quadrant for BI, Pyramid Analytics is the only BI vendor who consecutively has gone up to
[11:24] the right for for for two to three years. So, we've actually been able to bring in their marketleading semantics and BI inside of Service Now. So, we really are providing to you all a very powerful data analytics stack.
[11:39] And finally is converge. Your data is connected and it's governed. Your agents have all this context that semantic meaning. But the moment an agent needs to act, that data still needs to be accurate, needs to be timely, needs to be nearby. And our Raptor DB Pro is one
[11:53] engine for any type of data. So we've actually built out an HTAP database. So Raptor DB Pro is a is an extension from a Postgress database where you have column stores where you can do HTAP. So
[12:06] so transaction analytical story. With uh Raptor DB Pro, you can now do live connect which you can actually write write SQL queries directly to service now data to our database. You have also have live archiving. So when actually
[12:18] you have a lot of data that you actually just want to archive for because it's all old data you but you still need to keep it. You have all the archiving and it's multimodal. We're doing operational analytical having a graph also time series around this. So what is really
[12:32] important here is that when we're putting this all together, we know that a lot of customers take data outside of service now uh to be able to do more analytics outside of it. But what we're telling you is that a lot of those
[12:45] operations that you're trying to go do is analytics. You don't have to take that data outside of service now. You can actually then connect it to your data in data bricks be able to go provide all that semantics inside of service now. You also have all the
[12:57] analytics that we're providing or you can also bring in your thirdparty BI tools to do this. And what we're also doing is that we're packaging all this up with workflow specifics data apps. And what we're
[13:10] doing is really providing these applications out of the box starting with things like uh industrial analytics, customer service intelligence, advanced ITSM analytics and sec ops intelligence. So this is out
[13:23] of the box well definfined domain models that are being mapped to all the data you have inside of service now and any external data you're bringing in workflows that are doing analytics and workflows that apply to those analytics
[13:35] that are actually driving action back into service now. So these are the features and capabilities that we're offering. But one of the things I want to kind of I can go through so many different examples. But one of my one of the great examples that I have one this
[13:48] is a customer example. It's a large logistics company but this is actually FedEx. I mean they were at our at our knowledge conference last last month. FedEx was there at stage with our CEO Bill McDermott. H we're talking about this particular use case. So um they
[14:01] came to us because they had a revenue leakage problem. So they understand that they have a bunch of unbuilt invoices. Now this data to be able to go do this is part in some different Oracle systems. They have analytics that's
[14:13] happening in data bricks. So what we're what we're able to go do is to have first that agentic engagement, right? You be able to can go talk to your agents to be able to say look uh I want to identify all the unbuild invoices in
[14:26] North America for the last 30 days and let's sort this by dollar amount. to be able to go ask those questions and then the analytics that is powered here by the context can understand which data is needed is there an existing data product
[14:39] for there do we need to create a data product for all this stuff and that's why we're leveraging the entire context engine and your data catalog all together then with workflow data fabric and raptor DB what we can be able to do is that we already have data that exists
[14:52] inside of service now for example you have all of your customer cases and so forth we have your billing information in Oracle you have your analytics that you're doing things inside of data bricks and we can actually through zero
[15:04] copy connect to to data bricks through a zero copy to ERP connect to things like an Oracle and be able to identify what are the what are the particular missing invoices and some of them may be as simple as oh there was an issue here
[15:18] that I can just automatically go send update something a PO inside the billing system go do this maybe there's something more complicated that we need to go create a particular open up an incident inside of service now and assign This is actually doing the analytics and
[15:32] doing the work and taking the data, your analytics, your governance and your context to the work where the work needs to happen and that is the power of service now where through workflows and through our agents is where you get work
[15:44] done data and analytics governance it's all for the service of this. So with that I also want to share you a demo of where we're heading right now. Uh uh
[15:57] this this this this is a particular demo we have. Um we're partnering right now with McKenzie. So McKenzie uses a big is a big customer of data.orld from the catalog and also from from service now.
[16:10] And this is a this is an example where we're heading towards and I'm going to talk about three different personas. This first persona is a procurement officer. And the procurement officer is in teams where they start their day. And you can see here that James, they get a
[16:23] message saying, "Look, we have a particular vendor that we work with, a data vendor called Meridian, that there's some part there's issues happening with this particular vendor and the data that they're delivering to us." So, we can go in and check what's happening here. And the first thing that
[16:36] we're doing is that we're analyzing the the history about this particular vendor, all the vendors that we have, and we see that we've done some analysis, and we can see that Meridian Systems, their delivery has been declining for 90 days. And we can see I
[16:49] have these open escalations. This is some analytics. This is something that you're that you' be able to get out from some from your BI tools. What we can see here is the sources how this was done. We can see that we're act to be able to
[17:01] go create this analytics. You had to combine data from different sources. Snowflake data bricks and so forth. You can then see also the semantics that are being used to kind of understand all this particular data, the measures, the dimensions and so forth to be able to go
[17:13] bring this all together. But what I really want to go do is drive action with this. So I should be able to say is like what is going to be the client impact of having this data not being delivered on time. So McKenzie as
[17:26] a consultant they want to know they have different vendors uh uh and they have different part practices and they said look we actually have three different practices that can be affected by this. Note what's happening. We are connecting
[17:38] your technical your metadata about your data systems but we're connecting it also to your c to your users to your customers that you already have. This is the entire knowledge that we're bringing together. And we can say look there's
[17:50] actually a client deliverable that we know that something is due in 18 days and something's happening with this right now. And I can say well show me more of the details in here. What's happening? I can say look we've had a missed data feed syncs. There's some
[18:03] quality issues around this stuff right now. And what's happening is that with this vendor, this vendor we're in a renewal process. So what the procurement person can say like recommend some actions that I can take right now for
[18:15] this renewal process. So this is where we can now leverage the entire decision traces, the workflows that we know they've done in the past and we can make suggestions of what you could go do, what actions you can go take and you can say look you can either renegotiate with
[18:28] this vendor or maybe we can bring in a different vendor but you know what our recommendation is to renegotiate right now and here are the people that you should be talking to who are related to this. Again, you can see we're connecting your decision traces with a
[18:41] particular user graph within your within your organization and say, "Okay, let's guide me what this can happen here." And what the system, what auto is going to be doing is packaging this all together and saying, you know what, I'm going to create a partnership brief. This is
[18:55] takes all those analytics and we put it together. So, you can send it to David who is a partnership person you need to talk to. We can then see what is the particular uh meeting, set this up, create the the meeting proposal, uh
[19:07] schedule it and we're ready. This is now how we're combining the business problem of the procurement, bringing in your data, how it's affecting your business and actually getting that work done. But let's see kind of what happens underneath the hood a little bit around
[19:20] this. So this is now Sophia who is a data governance professional and what she is seeing is that there is a particular uh agent here that client benchmarking agent has three data assets that are performing below McKenzie's
[19:33] governance baseline and what when we see the details we see all these gaps that are happening with all these particular data sources in there the current escalations around that but we're actually going to be able to offer some suggestions and this is autonomous data
[19:46] governance right now in action is that we can make suggestions of how to go fix these governance gaps. Right now, what auto is doing is that we're building a plan for you. We're saying look, you should actually be connecting to some of
[19:58] this external data using zero copy. Uh you should be we could automatically classify some of the sensitive data around this. There's some generic descriptions. We could generate some data quality rules. You can edit the
[20:10] particular plan or you can go follow it here. If we go in and we're going to go build it, we're now going to follow those steps that we have presented in there. And what you can see is that it's actually done. It's creat the data in created data products. It's creating the
[20:23] sensitive classification. We need to have a data steward, data ownership. Again, we know the people and we know how they've been related. We can make suggestions that can say, you know what, Priya is probably the the data owner who should be here uh because she's actually
[20:35] worked with this vendor before. So, we can assign it. So, now we're assigning Priya to this uh to this new data product. But hey, we actually have an we can identify the errors because we have the data quality rules and we can say
[20:48] this there's some records that are missing. We can actually see what's happening here because we can then pull in all the lineage and we realize that there's some particular data that is missing but we can actually we know other data that you have and we can say
[21:01] for example there's S&P global data that you can use to fill in the blanks over there. So we can go say yep let's go ahead and let's go add uh let's use the S&P data and now we can we can now turn all this data ready and it can be
[21:13] published for this. Now let me go to a third persona and this here is Jack who is your AI steward. Jack is in the AI control tower where you're keeping track of your AI
[21:25] agents which can be AI agents inside of service now but can also be any agents outside including anything you're building building for example here in data bricks and what you can see is that there is now an update to that client benchmark agent that you have so can go
[21:39] in and can go uh re can get the the analysis about what has happened here and he says look there has been a change to this agent and the confidence score has increased uh everything looks really
[21:51] green from what is being told from the AI control tower. He says, "This is good to go. I'm going to go ahead and I'm going to approve this new agent around us." So now it's approving the agent. It's notifying the uh the data steward
[22:03] who built it. Actually suggesting then, you know what, we should probably set a confidence score. So if there's an issue that it goes down, we can go trigger if there's any new issues. So have that threshold and be able to support that. Confirm the threshold. And then now
[22:16] we're being able to send this back to Sophia and we're ready to go. three different personas, one platform where we're actually getting your work done. That's the power of service. Now again,
[22:29] combination of data links and we're doing much more than just it here. So to really to wrap up, look, many of you are customers like Service Now is trusted by so many different companies. I think
[22:41] it's like 90% of the Fortune 500 are Service Now customers. Uh, and what I really want you to leave here today is knowing that with Service Now is we're in the data analytics space and think about data as you need to connect your
[22:54] data, control your data, contextualize that, and converge to drive action. If you want to learn more about what we're doing at Service Now, please, if you're Service Now customers, a lot with our new AI native pricing, you probably
[23:06] already have access to all of these uh, offerings right now. And to learn more, our booth is right next to where the puppies are. and uh please come join us and I'm here for any questions you have. Thank you very much.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.