A $135 million contract signed this week extends IBM’s support of cloud services and software for the Army’s Logistics Support Activity (LOGSA).
The deal also gives LOGSA access to IBM’s artificial intelligence product Watson. This follows on the heels of a successful proof of concept in which LOGSA used Watson to deliver maintenance information on a portion of the Army’s Stryker vehicle fleet.
“It proved that in a very short period of time, Watson could consume an incredible amount of unstructured data, something like a billion bytes of data, and be able to identify what’s important in that raw data, and then use that to figure out ways to go forward,” said Kevin Aven, partner and co-account lead, Army and Marine Corps, IBM Global Business Services.
Going forward under the new contract will likely mean expanding the use of AI beyond the Stryker fleet, as a means to streamline and modernize other aspects of logistics.
“Capabilities like Watson make it a very exciting time for us,” said LOGSA Commander Col. John Kuenzli. “This can be a way to free up our analysts from some of the more technical work, to let the machine do some of that and to leverage the analysts’ professional expertise to look up and look further toward where the Army is going.”
One example comes in a just-launched Air Clearance Authority pilot project. IBM officials say Watson will help the Army to better analyze the tens of millions of dollars it spends to transport spare parts, leveraging AI to determine whether air or land is the more rational mode for a given shipment.
“The Army would like to see if Watson could look at all of those transportation decisions and be able to optimize that to reduce cost,” said Greg Souchack, partner, managed services and cloud solutions, IBM Global Business Services.
At present, seven LOGSA analysts save the Army about $100 million a year by auditing 10 percent of all transportation requests and routing material away from air transport and onto less expensive modes, Kuenzli said. His aim to have Watson audit 100 percent of requests. “Air is obviously the most expensive mode of transport, so we want to use it only when we absolutely need to,” he said. “We think artificial intelligence can audit those air cargo requests at a faster rate than a human analyst could possibly do it.”
The exploration of AI builds on a longstanding promise by Army to tap the emerging technology as a mean to simplify soldiers’ work.
Speaking at the Association of the United States Army Annual Meeting and Exposition last fall, Lt. Gen. Robert P. Ashley Jr., then deputy chief of staff, G-2, said the volume of senor and communications data has “almost become a burden” for soldiers and that “placing it on a machine” through machine learning and artificial intelligence systems could help soldiers make meaningful use of all that information.
In the Stryker proof of concept, LOGSA put Watson to work analyzing data generated by the 17 transmission sensors built into every Stryker vehicle. A small cadre of maintenance warrant officers taught Watson how to interpret the data, and then AI pored over millions of data points looking for early signs that engines might be facing trouble.
One simple but telling outcome: Watson noticed instances when transmission fluid got too hot, even at slow vehicle speeds.
“These are two very simple sensors — one captures fluid temperature and the other captures RPM — but a human wouldn’t see the two together. A human could see the temperature but not necessarily capture the speed,” Aven said.
By noticing the correlation, Watson could potentially pare back transmission failures — a significant win, considering Army spent $11 million to replace 120 transmissions in 2014. “What if we could have predicted that failure and maybe changed the fluid more often?” Aven said. “That could help prevent that cost. Then from a prescriptive standpoint, you could also look at how soldiers operate that vehicle and how it is being maintained.”
The Stryker pilot was “very beneficial,” Kuenzli said. “We learned that we could teach it our processes and turn it into an advisory tool to inform commanders. That’s very valuable: It reduces risk for commanders as they are getting ready to take vehicles into combat situations. It can also be used to inform the supply chain, and to inform our training programs around maintenance capabilities.”
At the same time, Kuenzli isn’t looking to AI as an outright replacement for human analysts, many of whom will still be needed to oversee these emerging systems. “You need technical experts, and you also need functional experts who understand the business process, people who can apply an understand of logistics,” he said.
“You need people involved who have an understanding of how we compute readiness to make sure we’d given the system the right equations and that the computer is hitting all the right data sources,” he continued.
The expanded AI capability comes in a contract meant generally to cover cloud services and software. That’s not coincidental. IBM officials describe the Army’s use of AI as the next logical extension of its ongoing migration to the cloud.
“Five years ago, the discussion around cloud was about infrastructure as a service, a way to get cheaper storage,” Souchack said. “This moves up the stack in cloud adoption, leveraging powerful tools in a cloud-based environment to do those analytics.”
As its cloud maturity has evolved, Army also has deployed more sophisticated computing techniques to bring down the costs associated with such capabilities.
IBM’s cloud support originally cost Army an average $60 million a year, Aven said. Under this contract it drops to $45 million thanks to techniques such as agile software development, which helps to streamline processes and reduce the cost of supporting a cloud deployment.