Data Management & AI Readiness in Intelligence with Lori Wade

[00:00:00] Lori Wade: Quality data is one of the number one factors for the success of AI and a lot of the technologies that are coming online, coming toward us right in the here, now, and in the future are pretty data, reliant, right? And so how do we in the intelligence community take the steps that we need to as a data organization to have data interoperability, have data that's consumable by humans and machines.

How do we work with the private sector? Our defense partners, our Five Eyes partners and others. To make sure that the data that we have discoverable, accessible and interoperable. And then how do we ready the workforce? 

[00:01:03] Ryan Connell: This is Ryan Connell from the Chief Digital Artificial Intelligence Office, joined here today with Lori Wade. Lori, how you doing?

[00:01:08] Lori Wade: I'm doing great. Thanks for having me. I'm excited about our discussion today.

[00:01:12] Ryan Connell: yeah. Excited as well. Appreciate you being here. You want to just give a quick overview of your role and what you're doing?

[00:01:19] Lori Wade: Absolutely. So I'm the chief data officer for the intelligence community. And what that means is basically all 18 intelligence community elements. Have chief data officers, I chair the council, for the CDOs, I bring them together. We're looking at the IC data strategy that was released last year.

We're currently in the second year of the implementation planning for that. We're starting to work on the next iteration of the strategy. And we also, in June, the intelligence community directive 504 was signed out. And so I've turned a focus into what are the intelligence planning guidance that has to come out of that.

directive on how we're going to manage our data across the 18 elements.

[00:02:04] Ryan Connell: Awesome. So you mentioned the IC data strategy set was released last year. 

[00:02:08] Lori Wade: That's correct. 

[00:02:09] Ryan Connell: yeah. So what's like the overarching summary there in terms of what you all are doing to manage?

[00:02:14] Lori Wade: I think it's an important point about the strategy and a lot of people, found it very useful. Uh, we intentionally going into it, we made it unclassified so that we could talk about it openly like we're doing here today so that we could emphasize the relationship between the public and private sector.

That's actually one of the four goals and focus areas of the strategy. And the idea is, we wanted to take a focus on the fact that the intelligence community is a data organization. So how do we have an intelligence community that is data driven, how do we unlock mission value and insight, use our data?

As we go forward and how do we have our data become AI ready, right? Quality data is one of the number one factors for the success of AI and a lot of the technologies that are coming online, coming toward us right in the here, now, and in the future are pretty data, reliant, right? And so how do we in the intelligence community take the steps that we need to as a data organization to have data interoperability, have data that's consumable by humans and machines.

How do we work with the private sector? Our defense partners, our Five Eyes partners and others. To make sure that the data that we have discoverable, accessible and interoperable. And then how do we ready the workforce? And so that's really the focus of the strategy in a nutshell. And the other factor here is that, that I made the strategy go to 2025.

And a lot of people thought that was a typo when we were putting it out. Cause you know, a lot of it follows a budget cycles and everything else. And I didn't because I wanted, uh, To signal that we have an urgency behind this, right? If we need to get this work, started, uh, we need to be in a good position as we go into 26 and out.

And so I had to signal the pacing challenge, not only with. Where we're looking at where we are in the world, we talk about near peer competitors, but they're actually a strategic competition now, right? And then also the pacing challenge of the technology itself. So that was a couple of unique factors about the strategy when we put it out.

[00:04:22] Ryan Connell: That makes a lot of sense. And, I can't even imagine. I mean, there must be, so much complexity around managing, uh, you talked about managing 18 different elements, within the IC community and the types of data that they, I assume gather is all different. So I can only imagine struggles in terms of interoperability and making different pieces of data connect.

[00:04:43] Lori Wade: Yes, it's, I don't even know if across the intelligence community, when we take a step back and we look at the complexities. That's why a lot of people will look at the data, the IC data strategy. It's only eight pages if you take the front and back off, right? And that's by design too. We don't need 40 pages to say what we need to do.

And I wanted to quickly get. To a place where we were implementing and we had the CDOs come together and we looked at what needs to be implemented against the four, focus areas and break that down into what can be done each year, do a new one, and then sign up right all the way up to the top of, Each element had to sign off on the prioritized items, and that came at the deputies level.

And then EXCOM level, which are the heads of the agencies. Everyone talks about the fact that it's very practical approach that we're following. And that is really just getting down to the core basics of data management and end to end data management. And so if we, if I have a chart that I've, laid out that shows basically from the point of collection and all collection, right.

And they, acquisition of data, whether it be open source or commercially available all the way through transport all the way through, uh, exploitation all the way through disseminating and for us, we have to care about, the disposition of the data. So if we lay that across the intelligence cycle and then we follow, because we are a data organization again, and we follow the data life cycle, what are we doing, uh, across the intelligence community to make sure that we're collecting for purpose, and that we're actually getting the data into a place where we can use it?

And then how are we getting it to a place where we can scale the analytics? And that's through the use and the availability of. AI and ML come, you know, that's come to us, right? So how are we doing that? But in a much more deliberate way. An intentional way than we have in the past. We've enjoyed a long history of collecting data and talking about collection and analysis and throwing our hands up, right?

Well, because of the volumes of data that we're dealing with, because of the availability of the technologies that we have now and that are coming, we have to work and think in a different way because data is what we're dealing with. Is it going to fuel that? And we need to make sure that our data is in a place that we're managing our data.

Right. And that we are, you know, if we talk about being a strategic asset, but if it's not managed well, it's not an asset, it becomes a liability. On many fronts, right? And so understanding the data at a very core level and as our business that we're in will help us to make the decisions across the data life cycle, but also how we do intelligence going forward.

And we're already seeing that shift and change, right? 

[00:07:34] Ryan Connell: Yeah, that makes a lot of sense, and something you said kind of my mind spinning a little bit on the cost of collection, I think that's an interesting point, um, curious like, which way it goes, or maybe the answer is both, like, is there a point in time where you're thinking, Oh, interesting. This piece of data can be collected.

Let's make a concerted effort to collect it really without the purpose in mind. Or do you always start with the purpose and say like, what piece of data am I missing in order to help me get there? is there a direction that makes sense there?

[00:08:00] Lori Wade: Well, I mean, what I've came into this role because I'm not new to the, you know, intelligence community, of course, I've been working as a contractor from the private sector side and then flipped over to the government side, but I've been working in the intelligence community since 1998.

So I've seen kind of across this, right. And I've worked at multiple different of the, uh, IC elements and what I have noticed over time, right and the decisions. I don't think, you know, I think we don't make the decisions around the data piece. And I, what I mean by that is I came up with a set of five smart data questions to ask ourselves before we do any collection, and that's what really became the basis of a data management template.

That gets into, you know, for what, intelligence value or mission is this going to fill? Who needs this, right? You know, what is the, what's the frequency of this? What is the volume, right? To get at the formula that we need to make better decisions about how much compute we actually do need, right?

You know, what's the storage decision on this, right? Meaning not only just the, what is the classification level of it, but how are we going to store it? Are we going to do cold storage, you know, or on prem all these decisions that can be made up front, because think about the level of details that goes into a collection, how many details are laid out, but there was no plan or deliberate plan for the data throughout its life cycle.

[00:09:24] Ryan Connell: Got it. That's super interesting. Yeah.

[00:09:26] Lori Wade: Right. So then we don't know what the total cost of the ownership of having that data all the way through. we're in this place now, we're pretty enamored by all the glitzy technology without taking a step back and understanding how we're actually going to use it.

And then what state does our data need to be in? How do we architect for data quality, which is a key component of being able to do AI at scale, right? And then the second thing is that there are a lot of, focus on how we need a lot of compute, right? Compute is a big discussion right now and I think people, are having that discussion around compute and stockpiling, chips and compute and all these other things.

I hear this all the time, right? And there's not a discussion in that about the data piece. And I think that's a mistake, right? And so then we're making investment decisions and prioritizing that over. We need to do it right. We need to be innovating. We need to be thinking about that. We need to be prepared for X, Y, or Z outcome, but we can't do it at the expense of not having the foundational pieces done.

And I heard something recently. I didn't come up with this, but I heard something and I liked it. Uh, recently on the correlation between where we are now in the, compute discussion and the, building out of data centers and the amount of money being spent on data centers to where we were in the mid 1990s, as we were going into the dial up journey to the internet versus for streaming, right. And what the streaming did and how, where we are today with 4k streaming, right. There was all this idea that we need all these big pipes and we, and, you know, we needed all this and people started building toward that. and, but as that was happening is when we went into the background, what was happening around HD and flash and then, you know, real time messaging protocol and then compression technologies.

That was all happening in the background, so when it got to a place, we didn't even need that anymore, right? So, what is it that we need to do to our data to get to a place where maybe we have different decisions about how the actual volume and how fast we want to process it, and have a formula by which we would determine how much compute is actually needed?

Or what's happening to the data along in the background here that we need to be doing so we can get to a place that we don't need that much compute and we won't need that much storage, right? Like how are those, what is the evolution on the data and the innovation on the data side is what I'm interested in and continuing to put a focus on.

Good to have all that, but don't know that it might be a similar type of journey as we go through here.

[00:11:55] Ryan Connell: Yeah, that makes a lot of sense. That's interesting. I appreciate you sharing. you mentioned a few times the importance of data, but also have talked about the, I'll say the funness or the experimentation with artificial intelligence. I personally found it interesting 'cause I, you know, being in DODI see a lot of organization have a CDAO, and you being a CDO, curious, uh, what that journey was like and the decision of someone I assume made a decision at some point to divide the artificial intelligence away from data. So, curious your thoughts on that.

[00:12:25] Lori Wade: Yeah, I, this is a interesting one for me and I've had a lot of conversations and done a lot of look into this. Uh, one of the things I did coming into this role is I continue to work is to elevate the role of a chief data officer because we, again, we are a data organization. The fact that, you know, architecting for data quality data interoperability. And all of that work, and I'm talking about the technical work around down to, you know, like I said, the decisions around what is the actual purpose of this collection or this data, how do we make it discoverable for the intelligence community, we have to also think about what is the attorney general guidelines bending decision about that data, which then drives how long we can actually keep the data.

we did a great deal to get a lexicon. Uh, there's a lot of work we're doing. We did a joint signed memo last year with the CDAO on a joint common core ontology, right? All the work that has to happen, uh, civil and privacy, all the things that we have to do to the data to tag it and label it in a way.

So that, and, and the classification decision right on the data. All of those things are not ai. Decisions, right there, nor are they technical decisions, right? And so I've been working to elevate the role of the CDO and several of the deputies of the elements have asked me, well, where would you have that role report to?

And I said, well, where does your CFO report to? Right. Because throughout the life cycle of your budget and your funds, there's a lot of tagging of those funds, reporting on those funds and watching, you know, spin plans and watching it all the way through, right? You have someone looking at that. Well, the same would be for our data because our data is, you know, again, there's only an asset.

If it's managed well, otherwise it's then becomes a liability and the data. So that's how I think I, so it started a discussion for me about the CDO does not belong under a CIO and then you get to the CAIO and I think the DOD model of combining it with. The CDO, and I see this across the federal government.

There's a trend toward that way, like state department, it has a CDAO. I think that is where it belongs with the data driving the implementation piece. Now there's other parts of the AI portfolio that belong in the CIO. And then they belong in like the research and development side. So like in your S& T portfolio, but the actual implementation it's going to require, it really cuts across multiple areas, right?

Because it's AI is a, is a capability, right? It's like any technology, right? You're just, you're trying to advance analytics, right? You're trying to advance your data. So A data professional responsible and looking at throughout the data life cycle and getting our data to a place where it can be consumed, right, by machines, machine readable, and it has all that other parts I just talked about on it is very critical.

What I see by separating it currently is because people are, you know, Focused on the glitzy part of the AI, who they're putting into the roles. I'm not sure it is mapping right. And some of them, it's a third, third, third, right? Some of them that I know of across the intelligence community and the federal government right now.

They have it be, it's, you know, it's the CIO is dual hatted as the CAIO, right? And then you have, some of them are CDOs and then some of them areseparate out into themselves, right? In the intelligence community, they're definitely making a conscious decision to have it separated out.

and we'll see how that goes. Right. that's brand new. I don't even, you know, that's just a very new, piece. I will work with that person just like I do what I call the digital C suite. Because to me, it's going to take a, I have to work as a digital C suite to make this work, right? We need quality data.

So that's the CVO. We need,the infrastructure and the compute and all those pieces. That's the CIO part, right? That I work, there's a data centric filler into the ICCIO five year roadmap, right? We need zero trust architecture. So it's a CISO. Right. This is so peace. So this is so important.

And then there's the AI, right? So in the IC, I'm just making that another part of the digital C suite. That we have to work with to make this happen, but as a data professional, as someone who has a master's degree in this, I see it very differently. And I don't see just the data piece, just being functional, or, just a technical, this is just like you would have, you know, like your chief financial officer, a chief data officer, especially in a data organization is the one who has to look at that all the way through from that point of collection, all the way through to disposition.

And that's what I see my purview as right.

[00:17:11] Ryan Connell: Yeah, you kind of drew something in my mind that I never really connected and I'm going to say it out loud and you can, agree or tell me no, Ryan, you don't know what you're talking about, which is fine. You know, we're commonly here talking about the cake layers of infrastructure data and kind of that top layer of app or platform.

And as you just described it, I was thinking, gosh, you kind of just explained that bottom piece being the CIO slash CISO, the middle piece being the data and the top piece being the glitzy AI, am I drawing incorrect analogy or is that something that you visualize it as well?

[00:17:40] Lori Wade: No, in fact, what's so funny is it wasn't called AI at the time, but I worked, in around the 2015 time I was the mission capabilities group chief. For eyesight for the intelligence community, right? So the IT, the vision for the IT enterprise for the intelligence community. And at the time, right, it was a pyramid that we drew a pyramid out and it was the, sort of that foundation.

Infrastructure, you know, that's where we have, you know, that's where the cloud journeys started. Right. And then you had that data layer, right? So the service is a common concern. Doing common, what was commonly done across the 18 elements, the IC data services came about so that we still are working that right.

And that's where we have. The IC Data Catalog, we have a lot of those services, and then as you go up the stack, it goes up to that analytics stack, right? Now, that is now where that can be powered, right? That's where, in the intelligence community, one of the actions that we're working on, and hopefully will be signed in the next year, a few weeks here, supposed to be signed next week, but I, uh, I just, I noticed right before I left, it was just still in review with the legal side, which we love our legal team.

but we have a data reference architecture, right? That's, we have worked across the intelligence community with that C suite I'm talking about, with the DOD. With our private sector partners to look at how are we architecting for data quality, which takes us up to a data mesh principle, which is really a data management approach when you're looking at distributed ecosystem and really getting us to that semantic layer.

Which is where on your graph, you're talking, that's where we're going to, an imperative will be the AI capabilities, right? So that we can scale,

[00:19:23] Ryan Connell: Yeah. 

[00:19:24] Lori Wade: right? And so all of that other work is, and that's why I view it as a, I don't own all of that stack. I know what part I'm focused on. It goes throughout our intelligence life cycle.

but I need all those other parts to be, you know, I need the demand signal from the. If the chief AI officers remain separate, then I need them to have a demand signal back to me, right, for quality data. I need them to have a demand signal back to the CIOs for infrastructure. And compute and all the other factors we need.

Right. So all of that needs to be working. And right now the demand signal it's kind of, they're still focused out here on the research and development part. Right. and then like, I have a demand signal where I need zero trust architecture, right? I know what the demand signals are and we just need to make sure that where are those things intersect that we're working for that to solve for that.

[00:20:16] Ryan Connell: Yeah, makes a lot of sense. right. Because if, the. AI team or that top layer isn't demanding a specific data point or even the amount of data or the clumsiness of data, it doesn't really steer your ship. and they also need to know how much on the compute side and infrastructure. So no, that makes a lot of sense.

Uh, I appreciate you opining on that, uh, reference as it popped into my head, as you were saying, and I was like, I think this is the same thing. so thanks. you know, you mentioned, you know, your, uh, Really, you know, intense breadth of experience in, how long you've been supporting some of these efforts.

I'm curious, like we all know that data is growing, you know, immensely in the world. so I'm curious, like how much has, I'll say like mission reliance on data, in terms of like outcomes and mission over the last couple of years, compared to maybe some of your experiences historically, when it comes to national security concerns.

[00:21:06] Lori Wade: Well, I mean, definitely, I have worked so many efforts that have happened, you know, 9, 11, we just had the anniversary of 9, 11. Right. And so I've worked a lot of these efforts, you know, in the aftermath, to look back and see. And so from that point to where we are today, and even just in the last two and a half years that I've been in this role, it's been phenomenal to see the shift, even of the leadership to look at the importance of data and how having our data, not only discoverable by default, but then how are we making it available? And then the use of that data and the demand signal from the mission has been incredible. and just even in the last year, I've been working with, different mission elements.

On some of their biggest data challenges and one I have worked with a big one that we worked with the on the D. O. D. side, was an end of pay. Come example. And what I have done as a response to that. I've recently just re org'd the office of the ICCDO, and one of the groups that we set up is a data operations group.

And there, what we're doing is identifying real mission use cases, or I call them mission challenges, and taking those and turning them into mission sprints. That we work, we bring in data, you know, data experts, some of the C suite, right? The C suite people I'm talking about. And we bring them into. Work a real example to see what are the gaps, you know, how can we fix a problem now and just turn and burn as I like to say, and move forward.

Right. And when we're doing that, we have the mission telling us, Hey, you know, I'm using these national tools, I'm using this work, but I'm sitting here and I call it, operation swivel chair fat finger, right? It's manual data transfer, and you're coming in here talking to me about, you know, artificial intelligence.

well, you can't scale stuff that's, manually transferred like that, right? So how do we help them solve those problems? you know, these are these, places around mission. I have mission coming, and they've now built. end to end data management plans that they are starting to implement, right?

I inserted, myself into this process where all major system acquisitions, all new collection, platforms come for funding, for national intelligence funding. And in there, they have just put in a milestone B to have an end to end data management plan as a deliverable.

[00:23:44] Ryan Connell: Wow. 

[00:23:45] Lori Wade: And why that's important is because it will ensure that What is collected then gets to a place that it's put into the IC data catalog and, Discoverable. And then all of that tagging and labeling I talked about, the attorney general guidelines, civilization, privacy, the correct classification. So that's put into, the proper tool. all these decisions can be made to get data to where it needs to be. And then also what I've heard from the mission, whether it be mission, you know, from our, from a human.

Perspective versus a signals versus a, you know, a geospatial, whatever it is, I've had mission. Ask me. over the years. How do I even know what I collected was of value? Well, we could do things with the data and with the technology to be able to show them that provenance and lineage, to be able to answer those questions.

But I actually have mission, asking that, right? So they care about the fact that they're doing this work and they want to see it make a difference. They want the, they want to understand the value of what they're doing, right? And then you have those who want the experience. That they have when they're at home working.

Right. And I call it the day they want a data centric experience. They want that user experience. Like they would, if they were playing a game online, like they don't want to go data dumpster diving, like, how can I have this available to me? And then trust it, understand that, you know, It's where it needs to be when it needs to be, right?

We have the right data available to us when we need it, right? So all of those things, because I've worked the mission, I'm not just a data professional. I've worked the mission. I understand that and I understand the challenge, right? And so that's what I've been trying to solve for as well.

[00:25:24] Ryan Connell: Yeah, that's super interesting. So I, I like, operation, swivel chair, fat finger. 

[00:25:28] Lori Wade: I've said that a lot.

[00:25:29] Ryan Connell: yeah. and I like the milestone plan B, a concept of needing a plan in place. Like, during requirement development. Is there any thought, or is there a connection in looking at what I think you call the IC data catalog of like what already exists that I could potentially use to make this to change my requirement?

Is that part of the equation?

[00:25:48] Lori Wade: That's one of the questions that's in the template.

Yeah, and all 18 during this last, last year's, action plan for the strategy, all 18 elements developed an end to end data management template. And multiple of the agencies, of course, have multiple different types of collection.

So they had to, they have a master template and then they have some even developed templates down to the data set level, and there is already helping to make some of those decisions. and I have an example that I, I. Run people through where we did this for, it was a new type of collection that was coming online and I can't get in, of course, to all the details, but I'll just say when they came to me and ask, Oh, Lori, you know, you talk about data a lot, help us out here.

We're trying to do this and, you know, here's our goal. We want to be able to get this data from the point of collection. and exploited and disseminated within 24 to 48 hours and I go, what's your plan for that? Granted, they had a massive operational plan for every other detail, right? Which included

a lot of significant lift, right?

If you will. And, they talked about their plan where we're going to do this. So we're going to do, you know, we're going to collect, we're going to collect it here and then we're going to bring it down here and we're going to take it across these pipes. And. Then we'll put it there into this repository and then, you know, and then it'll get exploited.

And I said, well, what is the, basic question is, well, what is the classification of the data that, you know, when collected? oh, it's unclassified. Well, then why are you using the most exquisite transport? Putting it in the, hardest place to where the repository, you know, where it's the hardest to get to, and only like two tools can hit it and not, and you want five missions to see it, but they don't even have access to that.

You know, all these decisions. It was a non decision, right? It was as a default about how we were going to run this. And so when we did it end to end, we brought data management professionals in, we brought in the IT elements, we brought everybody together for that kind of sprint mentality idea to work it and do a data management plan end to end in the first run, they got it down to like 40 hours.

And then they got it down to minutes, right? So, we call it, you know, the whole wrapper around that kind of really became how do we get something, you know, if we make deliberate decisions, we can get things that may take months down to minutes.and that worked very well and it worked, we did that in a non crisis, timeframe.

And then right after we had done that, we went into a, the period of time with, the Afghanistan withdrawal. And that is when they were able to deploy that end to end data management and get that data shared in, in minutes. Right. And so that's why I say the demand signal from the mission understanding.

How they need to have their data and the speed at which they need it and the format that they need it in and that all of that matters. And if we do a planning and deliberate planning, then we can make that happen and make it possible.

[00:28:46] Ryan Connell: Yeah. That's an incredible story. 

[00:28:47] Lori Wade: And then also they tagged and labeled that data. On ingest and so cataloged it in the IC data catalog, and then it went into an integrated data layer.

So all these decisions mattered, and it also used all of the capability and things that we've already funded, right? And that exist, we didn't have to create new repositories or new databases. We were able to use what already existed. And that's a big part of that deliberate thinking behind it as well, is use the capabilities that we have.

[00:29:17] Ryan Connell: a hundred percent. No, that's awesome. I appreciate that. Um, I do want to ask you one more question before we wrap, uh, sort of the, the Superman super woman conversation. But, if you, you know, had one big thing you could change about data or data collection, snap your fingers. what would it be?

[00:29:32] Lori Wade: I mean, one of the things that I think coming into this role two and a half years ago is to get to. Uh, get that end to end data management, planning, as part of how we do it. And the intelligence community directive 504 and getting everyone to do this, where the work is happening, elevating the CDO role, looking at it end to end, understanding that total cost of ownership, all of those things were my snap the finger, Well, I'm not leaving it to chance. That's why I'm embedding it in every place I can so that it just in the future for all new, this'll be a part of how we do intelligence as a data organization. so the the snap, the finger was when I came in and snapped that finger, but then it became a, how do we put action against it?

And that's where we are and what's what we're doing now. We've got work to do. we're not there. but the fact that this is already starting to be embedded in the And these major muscle movements of the intelligence community across these 18 elements that's it. We will get there by doing it.

And it's already starting to be embedded for all new data. So that's good.

[00:30:36] Ryan Connell: That's awesome. Hey, Lori, I appreciate your time here today. amazing insights. I appreciate you reframing everything into you all being a data organization. So just wanted to say, thank you.

[00:30:46] Lori Wade: Okay. Thank you. I've enjoyed the discussion. I probably could talk another two hours, but no one wants to listen to that. So

we're 

good. 

[00:30:53] Ryan Connell: No, it's all good. I appreciate it. Um, thanks again.

[00:30:56] Lori Wade: Okay. Thank you.