The Pentagon’s AI Shop Takes A Venture Capital Approach to Funding Tech

Nand Mulchandani, chief technology officer, Joint Artificial Intelligence Center, speaks Dec. 12, 2019, at the AFCEA NOVA-sponsored 18th Annual Air Force Information Technology Day in Washington

David Vergun, DOD

AA Font size + Print

Nand Mulchandani, chief technology officer, Joint Artificial Intelligence Center, speaks Dec. 12, 2019, at the AFCEA NOVA-sponsored 18th Annual Air Force Information Technology Day in Washington

The Joint Artificial Intelligence Center will take a Series A, B, approach to building tech for customers, with product managers and mission teams.

The Joint Artificial Intelligence Center will take a Series A, B, approach to building tech for customers, with product managers and mission teams. 

 Military leaders who long to copy the way Silicon Valley funds projects should know: the Valley isn’t the hit machine people think it is, says Nand Mulchandani, chief technical officer of the Pentagon’s Joint Artificial Intelligence Center. The key is to follow the right venture capital model.

Mulchandani, a veteran of several successful startups, aims to ensure JAIC’s investments in AI software and tools actually work out. So he is bringing a very specific venture-capital approach to the Pentagon.

Here’s the plan: when a DoD agency or military branch asks JAIC for help with some mission or activity, the Center will assign a “mission team” of, essentially, customer representatives to figure out what agency data might be relevant to the problem. 

Next, the JAIC will assign a “product manager” — not DoD’s customary program manager, but a role imported from the tech industry. 

He or she handles the actual building of the product, not the administrative logistics of running a program. The product manager will “gather customer needs, make those into product features, work with the program manager, ask, ‘What does the product do? How is it priced?’” Mulchandani told Defense One in a phone conversation on Thursday.   

The mission team and product manager will take a small part of the agency’s data to the software vendors or programs that they hire to solve the problem. These vendors will need to prove their solution works before scaling up to take on all available data.

“We’re going to have a Series A, a seed amount of money. You [the vendor] get a half a million bucks to curate the data, which tends to be the problem. Do the problem x in a very tiny way, taking sample data, seeing if an algorithm applies to it, and then scale it,” Mulchandani said on Wednesday at an event hosted by the Intelligence and National Security Alliance, or INSA.

“In the venture capital industry, you take a large project, identify core risk factors, like team risk, customer risk, etc. you fund enough to take care of these risks and see if you can overcome the risks through a prototype or simulation,” before you try to scale, he added later.

The customer must also plan to turn the product into a program of record or give it some other life outside of the JAIC.

That’s very different from the way the Defense Department pays for tech today, he said. “The unit of currency in the DoD seems to be ‘Well, this was a great idea; let’s stick a couple million bucks on it, see what happens.’ We’re not doing that way anymore” he said on Wednesday. 

The JAIC is working with the General Services Administration Centers of Excellence to create product manager roles in DoD and to figure out how to scale small solutions up. Recently, some members of the JAIC and the Centers of Excellence “participated in a series of human-centered design workshops to determine essential roles and responsibilities for managing data assets,” across areas that the JAIC will be developing products, like “cybersecurity, healthcare, predictive maintenance, and business automation,” according to the statement. 

Mulchandani urges the Pentagon not to make a fetish of Silicon Valley. Without the right business and funding processes, many venture startups fail just as badly as poorly thought out government projects. You just don’t hear about them.  

“When you end up in a situation where “there’s too much capital chasing too few good ideas that are real, you end up in a situation where you are funding a lot of junk. What ends up happening [in Silicon Valley] is… many of those companies just fail,” he said Wednesday. “The problem in DOD is similar. How do you apply discipline up front, on a venture model, to fund the good stuff as opposed to funding a lot of junk and then seeing two or three products that become successful?”

Close [ x ] More from DefenseOne