Microsoft said an artificial intelligence project cut Northern Trust’s client onboarding data preparation time by 83%, in customer results published on Nov. 12 by Rodrigo Kede Lima, president of Microsoft Frontier Company.
The task previously took about two and a half weeks. It now takes three days, Lima said in “From AI ambition to business impact.” Microsoft, which sells the AI software and cloud services used in the projects, published the account on its corporate blog.
Lima described customer projects undertaken with Microsoft engineers, although he gave no common evaluation period for the results. Microsoft had worked on nearly 500 projects with more than 200 customers across 28 countries, he said. Most customers with completed initial projects had returned for more work, generating hundreds of additional projects.
At Northern Trust, Microsoft engineers and the financial services company’s specialists built an AI onboarding tool within Excel and SharePoint, Lima said. The tool recorded a first-pass quality score of about 97% in testing, and analysts retained control of the work. Lima did not give the testing period or number of clients involved.
South Korean telecommunications provider KT cut the time needed to detect network problems to two to three minutes from more than 10 minutes, Lima said. Engineers built its AIO-Net platform using operating procedures and experienced network staff’s knowledge. People continue to control decisions and actions, he said.
At construction materials maker Saint-Gobain, engineers used Microsoft Discovery to assemble 62,000 patents and 44,000 research abstracts into a searchable collection within hours, Lima said. Saint-Gobain estimated that researchers would have needed a year for that work. The team identified possible cement additives in about six weeks, but scientists are still testing those ideas in laboratories.
Other projects had expected benefits rather than quantified results, Lima said. Microsoft and 3M are building AI software to help process customer orders, including credit checks and overdue-payment assessments, with human approvals. Lima said the system should reduce manual work and speed up payments, but supplied no measured results.
Lima also cited November research from IDC, sponsored by Microsoft, on how widely businesses were benefiting from AI. IDC found that 66% of organizations reported major improvements or changes to their businesses, while 48% were expanding use or getting consistent value across their businesses. Another 71% expected AI spending to rise at least 10% over the next year.
The IDC figures came from “The Real Business Value of AI,” but Lima’s post did not specify its sample size or survey dates. Lima said businesses needed to change work processes, establish controls and help employees learn new ways of working to turn individual projects into lasting gains. He said Microsoft’s success would depend on customers being able to run, oversee and improve their own AI systems after the joint work ended.
