Scoping a knowledge graph project for the first time is more than figuring out the technical components and making a TODO list of your deliverables. You’re actually most interested in how implementing this new technology, which leverages a web of connections between unstructured data instead of strictly-defined tables, provides business value to your organization.
But when people hear the term knowledge graph, they often make an misinformed assumption that they’re designed to replace people, because isn’t generating and acting on knowledge a uniquely human talent? How can a piece of technology store or generate knowledge itself?
But knowledge graphs aren’t designed to be a standalone solution for your business problem, and they can’t just magically generate knowledge from your data. And, most importantly, they don’t replace people. Instead, knowledge graphs, and the analytical engines developed on top of their connected, relationship-based data, are perfect for augmenting and enhancing the unique and invaluable skills of your organization’s heroes—your knowledge workers—not replacing them.
Your talent gets even better with new gadgets at your day-to-day disposal.
5,000 financial agents and too much data to deal with
To illustrate how knowledge graphs empower, and not replace, your organization’s most talented minds, we have the tricky real-world scenario leaders at the United States Department of the Treasury found themselves in.
They’re responsible for not just processing financial reports and collecting taxes, which is already an enormous task—every year, they handle 225 million reports, plus 68 million cases over the phone or in person, to discover the rare case where someone is operating illegally or trying to defraud the government. Even needle-in-haystack metaphors don’t convey the complexity of rooting out these frauds.
With over 5,000 agents, the Treasury had assigned to these cases, they routinely ran six-month investigations that involved an inordinate time spent gathering, reading, and organizing data about their subjects. The vast network of relationships between entities and transactions between them lived only in an agent’s head, making collaboration impossible, even when the volume of data was far too much for any single agent, however talented, to handle.
That overwhelming complexity didn’t even account for agents needing to sift through even more incoming data constantly. They couldn’t possibly read, think about, and categorize everything fast enough to find what’s relevant to their ongoing cases. And because of the Treasury’s complex data warehouse, agents were routinely missing relevant links to their cases because the data came from facets of the warehouse they weren’t yet familiar with.
The Treasury didn’t need to hire more agents. They needed to give their agents the right tools to get the job done.
Augmenting experts with financial knowledge graphs
The Treasury’s foray into knowledge graphs began when data and analytics teams worked with GraphGrid to move subsets of relevant data on high-priority cases from their data warehouse into the knowledge graph. Together, they converted thousands of tables into an easy-to-follow network, rich with context, built from unstructured data and relationships:
- Nodes: The people, places, and things the Treasury knows about.
- Edges: How nodes are related, like:
- People who are members of known organizations
- The addresses of people or businesses
- Who submitted a specific document to the Treasury
- Spousal or familial relationships between people
- Business transactions codified by financial filings
- Labels: A node’s category, like Person, Organization, Document, and more.
- Properties: The metadata related to a specific node or edge.
This knowledge graph was fielded to a “small” team of 100 agents responsible for solving urgent fraud cases, hoping they could empower this talent with better tools to add context to data and reveal patterns. By expanding each of their valuable skills, the Treasury could close cases faster than ever, recover millions more than projected, and even accelerate research into new fraud detection techniques they didn’t have time to tackle before.
GraphGrid provided the technological ecosystem that connected data from across the Treasury’s siloed warehouses and silos, giving agents a more comprehensive and accurate view of the reality behind any given case. That meant these 100 agents could suddenly collaborate directly on their new knowledge graph, providing a single truth source and convenient visual interfaces. They could work in tandem to trace clues and to make connections they wouldn’t have, as individuals, known to look for.
This team instantly went from comparing rows on spreadsheets and struggling to interpret complex queries to their old data warehouse to exploring their knowledge graph in ways that reinvigorated their creativity and analytical prowess. Which, in turn, led to them closing $10 million in recovery orders within the first 30 days.
For more of their stunning results, check out the full case study: How to Hunt A Cheat.
They haven’t replaced a single agent, but rather empowered each of them to be far more effective at their jobs. Not just to work faster, but to better leverage their unique skills for identifying red flags, recognizing the patterns of fraudulent behavior, and making the kinds of cognitive leaps that solve cases.
Empower your knowledge workers with GraphGrid
Wherever you have complex analytical problems that need solving, knowledge graphs are your answer—not to derive solutions all on their own, but to give your people every possible gadget to help them get there faster and with confidence.
And with GraphGrid, you get an entire suite of graph and artificial intelligence, which break down data silos, absorb the time-consuming tasks of analyzing and ingesting new data, and provide a collaborative workspace for solving the toughest of business challenges.
Or, read the full story of GraphGrid’s work with the United States Department of the Treasury in our case study.