Podcast: Leaping the valley of death with Matt Steckman

Matt Steckman joined me on the Acquisition Talk podcast to discuss about a wide range of issues facing growth-stage companies in the defense industry. He is the Chief Revenue Officer of Anduril Industries. Topics include:

• Edge computing and the future of artificial intelligence
• How to think about risk taking in product development
• The beauty of the Air Force’s approach to enterprise tools
• How cost and pricing data becomes self-reinforcing
• Why the security clearance process is a barrier to entry

The podcast features a discussion of transitioning technology across the “valley of death” to a program of record. Matt applauds government efforts to make it easier for new entrants to get small government contracts. However, it often takes years to get funding lined up through a program office. While large incumbents can manage the system by having a portfolio of projects at various stages, new entrants may face a gap in revenues for multiple years. As Matt reminds us, Anduril didn’t even exist when this year’s budgets were being planned more than two years ago.

In order to bridge the valley of death, Matt recommends small companies field capabilities early and have honest conversations with government about what it takes to stay viable. You will often find an empathetic partner. He also provides three things the DoD can do that will allow industry to rise up and participate. (1) recompete programs more often; (2)  if the winner is new, be willing to move larger amounts of funds; and (3) increase flexibility to move funding within the fiscal year.

Podcast annotations

Anduril is a defense products company. We focus on getting products 50, 60, 70, 80 percent of the way there, and then going to the government and trying to start a project together to get over the hump to fit the perfect need…

 

In a lot of current contracting models, the government bares all the risk. What we’re saying is we’re going to shoulder a lot of that, and we’re going to spend a lot of money and time to prove this thing out and bring it to you when it makes sense. We don’t just do that because we’re crazy. We do that because it allows us to move extremely fast to rapidly prototype and iterate, because we can control aspects of it up to a certain point.

One of the ways new firms may be able to break into defense is by riding a wave of commercial technology. They can start by using small SBIR dollars and private investment to build a minimally viable product that is designed from the start using new platform paradigms, such as artificial intelligence. It won’t have high capability at first, but it can fill niche needs within military operations that major systems cannot. Best of all, the product doesn’t have to wait several years to win a program of record. It can access unprogrammed funds in the O&M budget account so long as it is better than the available alternatives.

That lower-end product, however, rests on a foundation that allows future iterations to achieve rapid advances that would eventually supplant the incumbents in terms of high-end capabilities. Having sustained itself on niche capabilities, new firms can prepare itself for winning programs of record. Incumbents will find their old processes difficult to align to the new way of doing things. This is but one re-telling of Clayton Christensen’s disruptive innovation story.

Here is Matt on why it might be difficult to sprinkle AI on top of existing system applications:

If you don’t consider autonomy and artificial intelligence from the very start of your project, you probably end up making design choices that inhibit the ability to use that toward the end state. Good example is the Anduril Sentry Tower. It allows you to effectively create a sensor bubble around any fixed point on the Earth, where you’d know anything happening within that bubble and the AI is able to identify, characterize, and track those things.

 

If we didn’t assume that was operating in absence of a human looking at a screen or controlling it from the very start, we would have made many design choices along the way where at the end we wouldn’t have been able to backtrack and add the AI later. Multiplying that by all they complexity you see in aircraft platforms and ships, tanks, and things like that, you start to understand how difficult it will be to try to go back in time and add this as opposed to thinking about it fresh.

When disruptive innovations appear, the DoD should be on the leading edge of that. The new concepts are unlikely that they are reflected in military requirements already in motion. Not missing out on the next generation of systems requires providing government officials with untrammeled funds that can be redirected within the year of execution onto promising new projects. In effect, well a well funded technology transfer account.

While the DoD is starting to move fast on projects of around $1 million or less, there needs to be options for experimentation at the levels of $10 million and even $100+ million without entering the multiyear planning phases of the budget. As Matt tell us:

The discretion that they can move money around fast at those magnitudes is limited. There are processes for urgent needs. Someone might look at the statement and say, “oh, it exists already.” I’d say, have you ever actually tried to do it? It is painful.

Here’s a good line:

Will there ever be a day in the history of Anduril where we do a cost-plus or something like that? Probably not. But maybe. We wouldn’t say ‘no’ if it made business sense.

I’d like to thank Matt Steckman for joining me on the Acquisition Talk podcast. I recommend you read his articles on Medium, “Pilot to Production, Leaping the Valley of Death in Defense Contracting,” and “Access Impedes Innovation in Defense — There is a Simple Solution.” Here is an NBC news segment on Anduril Industries.