Modernization in Defense

The Department of Defense has begun the massive process of modernizing its information systems, which cover the globe and handle administrative functions as well as power advanced weapons systems. What are some of the issues DOD managers have to confront as they balance the use of new technologies with organizations that have traditionally been reluctant to change?

Defense Systems hosted a discussion on the topic late last year that offered a group of managers from DOD an opportunity to discuss their experiences with introducing emerging technologies into the process of modernization. The discussion was on the record, but not for attribution. Quotes have been edited for length and article flow and clarity.

A new way of looking at infrastructure

A fundamental change is happening in DOD that moves the focus from the individual system to the data or the code in that system, one participant said. “That's going to be an interesting dynamic and a change that affects everybody. It affects how PEOs operate. It affects how program offices operate. It affects how integrators come in and support the government.”

“It's going to really affect our infrastructure decisions. What I mean by that is when we start talking about cloud -- great, infrastructures as services -- awesome. Those are pretty easy things to work out.” But when you start focusing on the data itself, there is a lot of uncharted territory.” There is not a lot of policy covering how it is handled or shared, the executive said.

Another executive said he was a business guy in a room with a lot of technologists. He offered a different perspective. “I don't think it's about cloud. I don't think it is consolidation. I don't think it is AI. I think it's about exploiting data--period.”

But what is the best way to accomplish that? One participant said that a lot of our organizations have not determined appropriately what the access and dissemination rules are for the data individually, let alone in compilation, or how vulnerable we are on those perspectives.

“We are going to have to learn and think through this as we go,” said another participant. “The propelling factor for us will be consolidation of technology. Cloud will be in part of the answer. AI will certainly be part of the answer. They're all part of it.”

The question is how do we exploit data that we have? How do we understand new business models? How do we do business differently because the threats are ever increasing?

Another executive said, “What we found was that by sharing and collaborating across spaces and creating marketplaces for that, we might be able to start that journey. It took an awful lot of work from the organization to do that.”

“I think it's taken so long just to get two disparate data systems together, get data from the systems together.

“We've got people exchanging hard drives in the parking lot, said one participant, because we can't do it digitally, and just to perform analysis. It's taken us one year to get two data systems to do descriptive analytics so that we can better inform predictive and prescriptive analytics.”

“That's not good stewardship of DOD's money,” the executive said

I think from a larger DOD perspective, the data piece is going to be hugely important, and kind of opening that up to a broader audience is what should happen, said another.

“A question for my colleagues here,” said another participant. “We are seeing the innovation and the radical pace of applying data to solve problems. But there is an opposing force in place that represents risk adverse system authorization,” said the participant.

“If you're risk averse, you can always narrow it down and always find a 'no' answer somewhere in the policy framework. That's what we're seeing right now. Very archaic old school processes can slow the innovation, and any single person in the authorization chain can pull the system and halt it.”

Protecting the data

If you are going to double down on exploiting data, then it becomes very important to protect that data. One participant said, “Honestly, protect data, not systems.”

“I don't think it’s realistic to keep pace with the threats continually evolving,” said another participant. It's being able to have that agility and flexibility so the organization can react more quickly. Trying to beat every threat or end them one at a time is impossible. You're not going to be able to stay ahead of the threat environment.”

“If you have the false positives, the challenge is to apply AI to reduce it, or to apply expert analysis to reduce it. If you have the false negatives, the only way you can do that is triangulations,” said another executive.

“I thought one way to look at it from a cost effective standpoint is for every agency to look at its own threat. If you're in a military hostile situation versus a civilian organization in a regulatory environment or if you're in commerce or healthcare, then you can better understand your individual threat environment.”

"I would be interested in some sort of a consolidation,” said a participant. “Where can we go to a place in the government where we can triangulate? Where you can consolidate the various sensors' global intelligence, triangulate to the false negatives? The battle is in the false negatives. The battle is not in the false positives.”

“The threats are changing so quickly that by tomorrow the current tools may all be useless. We've seen that time and time again," said one participant. "How do we set up a framework within DOD and the IC so that we can move very quickly?”

The modernization workforce

“Human capital is a huge challenge. You have a whole generation of folks that are graduating from universities right now and even coming out of high school who code better than some of the best developers we have in DOD right now. That is a big deal,” said an executive. “Some suggest retraining the work force. That’s going to be a huge challenge.”

“Our industry partners tell us that also,” said another participant. “In conversations with industry, even the integrators that support DOD say their ability to retain talent on the big data side, on the DevOps side, and in a bunch of different areas is a big challenge.”

“We have to think differently in terms of the workforce that we want to attract, as well as those we want to retain,” said another executive.

“In some ways, unless we change our approach to how we hire, as well as how we structure our pay, that challenge is going to persist -- and probably get worse. I don't know if those changes are going to happen any time soon.”

“We need to manage differently” said another executive. “We need to reconstitute how we think about our problems, meet our workforce — not how are we going to get our workforce to meet us, because they are brought up differently with access, tools, and other capabilities.”

Emerging tech

“There's a journey here, because here's the reality,” said a participant. “A lot of this is on legacy systems. That causes its own issues. The data's not uniform across the force, and that's OK. We can work around that, and the questions being asked there are fundamentally different, too. It's like a little bit of an evolution. You got to get some data...crunch it a little bit, answer some questions, then evolve to meet some of the rest of your requirements.”

“You have to start somewhere," said another executive. “In my eyes, you're going to have to start with predictive analytics, and you get your algorithms and everything else. Now you have to train a computer to understand what those things are. Then you eventually get to AI."

“Then from AI, you have machine to go on its own, and start consuming stuff, not randomly, but in a way, and that's machine learning,” the participant said.

“One of the things to consider is how long does it take for people to generate trust in that AI process.”

Another participant suggested, “If a machine creates a product, it's going to be five years before I'm going to believe that product is actually as good as a human being could produce. That product is produced, and someone's going to go back and check that thing again, and again, and again, and again, until there is such consistency.

“There's an interaction. It's almost a system reward interaction between the operator and the system. What the system provides is that ability to channel those massive amounts of data and help chew through it to present a problem that a human can actually handle, and then provide the reinforcement.”

“We will evolve and teach those systems as we go along said an executive. “One of the things we may have to worry about at some point is that there's a potential for working backwards from the AI to determine what the training steps were.

“I think we need to focus on how we exploit the information we have because AI is a technique for predictive analytics. The issue is how do we take advantage of what we have today. Not what we wish for five years from now.

“The question is can we exploit through analytic techniques the information we have to bring value to whatever problem we face,” said the participant.

And the question that ended the conversation was, “How fast can we move into that? We need a work force, we need infrastructure so tools are commercially available, and we need to learn to manage through any challenges.”

Note: The November 14 gathering was underwritten by OpenText, but both the substance of the discussion and the recap on these pages are strictly editorial products. Neither OpenText nor any of the roundtable participants had input beyond their November 14 comments.