Can machine learning make sense of government's IT spend?
Connecting state and local government leaders
As the federal government looks to embrace the Technology Business Management framework, Apptio believes intelligent automation and analysis can speed those efforts.
The Trump administration wants the Technology Business Management framework -- a detailed system for tracking IT investments and measuring the business outcomes they deliver -- to be embraced governmentwide by 2022. But since most agencies lack clean and consistent data on the IT they buy, tackling TBM often starts with a tedious and time-consuming inventory process.
"We want to streamline the reporting of IT," Federal CIO Suzette Kent said at a July 11 event hosted by the TBM Council, and " we want to use technology to do that."
Deputy Federal CIO Margie Graves elaborated on that point, noting that the Office of Management and Budget is working with federal contractors to ensure their bills to federal agencies come with detailed data in a TBM-friendly format. "Our vision is that this becomes more automated [so that] people are lifted up out of the data calls to the actual analysis," Graves said.
At least one firm believes artificial intelligence can automate much of that data-gathering and make it easier for agencies to jump to TBM. Apptio, which offers a suite of TBM solutions, announced on July 11 that an "early adopter program" is now open to federal agencies looking to leverage machine learning to automatically extract and map technology costs from their general ledger and invoicing systems for TBM.
The data in those federal financial systems is too broad and inconsistent to be insightful, Apptio CEO Sunny Gupta told GCN at the event. "Our machine learning technology can automatically ingest the ledger and the expense structure and the budget files, which are not at the level of granularity that is required.… We are pattern matching millions of words to map all that data to technology."
That machine learning tools can shrink a months-long process down to "a matter of days" and get the data mapped at the TBM "cost-pool" level, Gupta said. An "intelligent insights" service, meanwhile, will automatically flag outliers in the cost structure and suggest changes for agency IT leaders to consider.
"Both tools are geared to helping agencies … adopt the TBM model faster," he said.
Of course, agencies don't have to have intelligent automation -- or even a dedicated TBM application -- to implement the framework.
"We did TBM through brute force for the first few years," General Services Administration CIO David Shive said at the event. "We didn’t have a system.... And that's a perfectly viable way to do it." In fact, he added, there are dangers in trying to scale the effort too quickly or trying to "make the data perfect" from day one. GSA had to essentially reboot its TBM effort after six months, Shive said, having learned those lessons the hard way.
At this stage, however, Shive agreed that "automating the process was absolutely the right way to do it." (Note: Shive, Graves and Kent all spoke strictly about the larger goals and challenges, not about Apptio's new offerings or other vendor solutions.)
And while OMB is pushing to require TBM-ready data from IT vendors as part of future contracts -- something that presumably would reduce the need for machine learning-driven categorization -- Gupta said that would simply make services like Apptio's more useful.
"Additional granularity would make our cost models even more accurate," he said. "It gives us the opportunity develop even stronger machine learning capabilities around the vendor invoices."
"The role we want to play with the federal government is to be very prescriptive," Gupta said, "and say to GSA or OMB, 'Hey, if every vendor invoice had these 10 fields, the entire model will light up.'"
Apptio's platform received FedRAMP authorization in April 2018.