Microsoft has been digging deep into the data created by its $120bn treasury unit in search of real-world use cases where blockchain could cut costs.
Culling through this trove of information, Microsoft's newly appointed principal program manager and lead blockchain engineer, Marley Gray, searched for pain points and labor intensive processes that were adding complexity and error to his company's operations.
But the fact that an executive like Gray would be allocating time to such an endeavor is perhaps not surprising. Since helping establish Microsoft Azure's blockchain-as-a-service platform in November 2015, Gray has been working with his company's partners to find new ways to use distributed ledger tech to solve real-world problems.
But, in this case, Gray was also looking for was a problem he could solve quickly.
Gray told CoinDesk:
"We wanted to pick something that was big enough to matter, but small enough to do quickly. In like, a month."
The blockchain engineer was speaking from the Sibos financial conference in Geneva, where he had just demoed the prototype that resulted from his searching: a stand-by letter of credit solution built on the ethereum blockchain with Bank of America Merrill Lynch.
Letters of credit are what banks give clients when they're dealing with new customers who may not know their client well. It's a payment of last resort in which the bank agrees to pay the tab in case the client is unable to do so.
On average, Microsoft says it uses about two letters of credit per week, but that figure increases dramatically on a seasonal basis, according to Gray.
To build the prototype, Gray and his blockchain engineers first worked alongside several Bank of America team members to calculate the number of steps the counterparties in a standby letter of credit transaction need to conduct, from handwritten notes to phone calls and faxes.
In total, the two teams calculated 15 steps that would need to be conducted by four counterparties, including a customer, a beneficiary and each of the counterparty banks.
From beginning to end, the application process takes about five business days, and the team identified an error rate as high as a 50% due to factors like poor data entry and illegible handwriting.
The resulting cost per letter of credit is between $2,500 and $15,000 depending on the number of errors.
Gray summed up the current transaction delays:
"Those things can't be resolved other than picking up the phone and with the time-zone difference you won’t hear back from them until the next day. When you're dealing with international trade it becomes very painful."
Banking on blockchain
Microsoft's $120bn treasury is dispersed over 1,000 bank accounts held by multiple providers — one of which is Bank of America Merrill Lynch, and Gray had been consulting with the bank to keep it appraised of his company's blockchain roadmap in the months leading up to the project.
In conversation with CoinDesk, Bank of America's head of North America global trade and supply chain finance, Chris Bozek, said his team had been looking for a way to learn more about how blockchain tech could help his own company and its clients.
To help ensure the proof-of-concept integrated with Bank of America's existing system, Bozek deployed members of Microsoft's core product team, technology team and operations team to validate each of the encoded processes.
"Each step of the way is important when we are doing a proof-of-concept," said Bozek, "that we can see the connection points to the broader enterprise."
Built on a multi-node, private installation of the ethereum blockchain, the prototype simplifies the standby letter for credit process from 15 to four steps. And with a 0% error rate, it cuts the time down from five days to between five and 10 minutes, according to Gray’s numbers.
The benefits of such a streamlining include not only helping Microsoft and Bank of America save time by reducing errors, but increasing transparency, resulting in a more efficient means to manage cash flow and working capital, according to Bozek.
While the prototype demoed to CoinDesk earlier this week shows that "the plumbing works," as Bozek put it, he added that it only "scratches to surface" of how it can be used.
"Now that we have that down, we move to some of the more complex products."
The next step is to build a pilot that runs parallel to the existing analog application, with other counterparties and new partners participating.
Currently, the business model has yet to be determined.
But Marley told CoinDesk Microsoft is unlikely to seek to take ownership of the tech. "We’ll probably turn it over to an existing consortia and say 'Here you go,'" he said.
As time goes on, Gray expects to add new features from Microsoft's Project Bletchley, including identity management, integration with the Azure Active Directory, a blockchain identity service, key management and data services such as data analytics dashboards.
Among other applications that the standby letter of credit "plumbing" might be able to serve are commercial letters of credit, and other even more complicated trade financing products that involve numerous counterparties.
"But the value or promise of blockchains is that the value of the network itself goes up by each counterparty you add," said Gray.
During side-by-side tests, his team added additional counterparties to the credit application process to see what happened to the system's efficiency.
While the number of steps in the analog process increased to more than 15, "the blockchain line stayed flat," he said.
"That kind of proved our point, the near exponential value of the network relative to the number of counterparties. It was a good experiment for us to figure out, is the promise real?"
Image via Michael del Castillo for CoinDesk