In the late 2000s, financial crime committed at banks by traders garnered a lot of media attention. Those behaviors also eroded customer trust in financial institutions and led to massive fines. To help banks detect undesirable activity, KPMG created Magna, a risk analytics solution. The company recently added more AI capabilities to Magna using Microsoft Azure Cognitive Search, Azure Cognitive Service for Speech, and Azure Translator. The sophisticated AI detects subtle clues about misbehavior and alerts KPMG customers in a fraction of the time, boosting trust and decreasing risk.
“With support from Azure AI, we give analysts better insights and only highlight the most relevant conversations, which leads to effective risk detection at a reduced cost.”
Steven Wells, Director, Forensic Data Analytics Team, KPMG
Increased risk complexities
As a large professional services company serving many customers a variety of services, KPMG has both breadth and depth of expertise. One of its offerings focuses on identifying and preventing financial crime, including trader surveillance, using technology and data analytics. Customers such as banks and other financial institutions rely on KPMG for solutions to identify fraudulent transactions or other misconduct by financial traders.
But identifying potential fraud is harder than it sounds because the signs aren’t always obvious. For example, two traders might try to manipulate the market by using insider information but subtly hint to continuing the conversation offline. KPMG created Magna, a risk analytics solution, to detect these signals and help banks monitor risk. It first launched the solution on-premises in 2014, and by 2017 started collaborating with Microsoft to move the solution to the cloud.
Recently, KPMG wanted to consolidate growing volumes of unstructured data and offer its customers better search. It decided to take advantage of advanced AI capabilities to bring together trader communications across various channels such as email, phone calls, and chats to identify potential risks quickly and comprehensively. “Over time, we perform market scans for the best technologies available to support our services,” says Steven Wells, Director, Forensic Data Analytics Team at KPMG. “Traders speak in code to each other, so using AI helps us pinpoint signals that are intentionally hard to identify.”
Sophisticated and unified AI capabilities
As part of its ongoing upgrades, KPMG adopted Azure Cognitive Services for its many AI offerings. The company added Azure Cognitive Service for Speech to Magna to automatically transcribe conversations, resulting in searchable transcripts. It uses Azure Cognitive Service for Language to extract and sort text from documents. With Azure Computer Vision, KPMG better finds and analyzes images and videos, including using optical character recognition (OCR) APIs. “We’ve found Azure Cognitive Services to be highly accurate and a better fit for our ecosystem and our customers’ ecosystems than any other available services,” says Wells. “The technology is easy to use from an onboarding and integration perspective, which we found really compelling.”
KPMG serves banking customers around the world, so its employees speak a variety of languages, and sometimes multiple languages in the same conversation. Traders might use local slang or colloquial language that basic AI might not detect. The company adopted Azure Translator for Magna to gain prebuilt and custom translation capabilities. “With Azure Translator, we capture multiple languages within a single audio file, and we can customize our models to capture jargon, acronyms, and pronunciations,” says Athanasios Pavlou, Senior Manager, Advanced Analytics and Cloud at KPMG. “These capabilities are incredibly important for us and have led to Magna’s increased performance.”
To further unify capabilities, the company adopted Azure Cognitive Search, which contextualizes and searches unstructured data across customer data ecosystems and improves the quality of the search results. KPMG also uses the preview capabilities of semantic search, powerful machine learning models trained on billions of Bing queries to return nuanced and semantically relevant search results across the data ingested in the company’s search index. “We think of Azure Cognitive Search as a central pillar for our engagement model because it consolidates the Cognitive Services that we already use so that we can effectively tackle unstructured data,” says Pavlou.
Faster response time and reduced risk
KPMG updated Magna in part because it needed to offer faster insights to customers, while also searching more data from wider sources than ever before. “In general, data volume always grows, and we needed to drive insights through that volume of data much faster so we could find the proverbial needle in the haystack,” says Wells. “Customers want to take action sooner, but we’re not talking about just one system—we’re talking about lots of connected systems, which means catching risk signals earlier.”
The company achieved this goal by adopting Azure Cognitive Services and Azure Cognitive Search and making more content searchable. It consolidates data from documents and other data sources and automates how the risks are found so that analysts know exactly where to focus their attention instead of spending time manually combing data. “With Azure Cognitive Search, we give customers relevant information and provide time to insight much quicker,” says Wells. “When we first launched Magna, it would alert an analyst to an issue within 30 days. Now, most of our customers want alerts within two days, and we can deliver that. The sooner they can take action against a developing issue, the further they reduce their risk.”
Increased trust and ROI for customers
KPMG achieved its top priority of helping customers decrease risk. First and foremost, decreased risk leads to increased trust and reputation for customers. “I’ve seen how important transparency is to our customers,” says Wells. “They want to be trustworthy custodians of their clients’ money and data. We help them achieve that because our solution reduces risk and improves compliance.”
Second, decreased risk also helps customers save money, both by avoiding issues and breaches in the first place, and by eliminating the subsequent fines. “Financial institutions get fined serious money,” says Wells.
Finally, KPMG provides this tremendous value to its customers while keeping its product and pricing very competitive for the industry. “These programs have traditionally been very expensive to run and operate, but we uncover relevant risks at a good cost point,” concludes Wells. “With support from Azure AI, we give analysts better insights and only highlight the most relevant conversations, which leads to effective risk detection at a reduced cost. We deliver more value to customers and create a competitive differentiator for our solution.”
Find out more about KPMG on Twitter, Facebook, YouTube, and LinkedIn.
“We think of Azure Cognitive Search as a central pillar for our engagement model because it consolidates the Cognitive Services that we already use so that we can effectively tackle unstructured data.”
Athanasios Pavlou, Senior Manager, Advanced Analytics and Cloud, KPMG
Follow Microsoft