AI and Enterprise Knowledge Integration: Part 3
The starting point in Part 1 of this series was the fragmented “knowledge landscape” of most big companies. Information is everywhere but it mostly lives in autonomous silos, in different formats and suffers from “semantic incoherency”.
This is a huge problem for extending the use of AI in business, beyond the many “narrow” (i.e. highly specific) tasks that Machine Learning does so well today and into the digital transformation of complex knowledge intensive work.
To meet this challenge, I argued in Part 2 that we need to “connect up” different forms of enterprise knowledge, with the help of semantic technologies - such as ontologies and knowledge graphs - from the “Symbolic AI” tradition where meaning and reasoning take center stage.
This concluding post will focus on business outcomes – the benefits that leading-edge companies around the world are already beginning to achieve, leveraging semantic graph technologies to integrate enterprise knowledge and transform knowledge work.
Business outcomes of Enterprise Knowledge Integration
The business benefits of knowledge integration are to be found across at least the following four areas:
eBay is a leader in “conversational shopping” with ShopBot, a personal shopping “bot” which chats with customers via text, voice or photo search, to help determine what they want and what eBay can offer.
While the NLP interface was trained by Machine Learning, the heart of the system is a knowledge graph which contains both the catalog and the attributes of shopper interactions while seeking products.
Starting from requests like “I want a brown purse”, ShopBot uses the graph to pose more precise follow-on questions (material, size, style, budget…) before offering a choice of products. It also remembers and stores in the knowledge graph the interactions with individual shoppers for increasingly customized recommendations in the future.
Siemens is a pioneering user of industrial knowledge graphs, with a variety of use cases. Let’s take one example: generation of turbine configurations to meet customer requirements.
Previously, configuration information was scattered across spreadsheets, inconsistent and redundant, wasting the time of skilled engineers to pull it together.
With an industrial knowledge graph, it is now managed as product configuration knowledge with rich semantics, and used in an integrated and accelerated design process across all components and technologies.
Thomson Reuters was previously known primarily for its feed of market quotes to traders and financial institutions. In 2017, the company launched a Knowledge Graph Feed, transforming itself from a data provider into a supplier of knowledge.
The graph can be installed on premises or accessed in the cloud, continually updated by a feed of semantically “linked data” about companies, value chains, financial instruments and quotes.
According to the company, “the 2 billion relationships and data structures that power the Knowledge Graph enable financial firms to understand the complete picture of the ecosystem around their investments, targets and prospects.”
Trust and compliance
Credit Suisse, a financial sector leader in using semantic graph technology and Machine Learning, had tens of thousands of data silos including 45.000 databases plus various Big Data projects. Siloed data makes life very difficult for Compliance Officers, who need access to all sorts of information - a typical problem for large banks, subject to as many as 50 compliance regimes with constantly evolving rules.
The crux of the Credit Suisse approach was not stand-alone knowledge graphs (although the bank presumably has some) but rather semantic graph technology to link the silos.
The entire infrastructure was connected on a semantic graph with standardized models and taxonomies, creating a “virtual knowledge graph” for provable regulatory compliance, but also for many other use cases in the bank’s digital transformation.
In this blog series, I’ve only just sketched out how “connecting up” enterprise knowledge opens new vistas for the transformative force of AI.
Many issues call for “drill down”. How will Machine Learning and Symbolic AI increasingly work together for business benefits? How can cybersecurity be strengthened by leveraging these complementary technologies? How can semantic AI bring greater value to enterprise investments in Big Data and Data Lakes? And what are the best practices for implementation? The list goes on.
Keep an eye on this space as the conversation continues.