Three Guiding Principles to Unlock AI’s Potential

AI is going to change the way you work.

This sentiment has been splashed across headlines, conferences, think pieces, marketing – it’s unavoidable in the wake of the release of ChatGPT, which gave mainstream visibility to AI (specifically large language models) and made this complex technology accessible to everyone.

It’s also true. No matter what your line of work, AI-powered tools can make us more productive while freeing us up to be more creative and strategic. At Bloomberg Industry Group, we see countless opportunities to help our clients in law, tax, and government reimagine how they do their jobs with AI.??

?

Years of AI experience

While the allure of AI is sending many organizations rushing headlong into the fray, we’ve been successfully leveraging AI in our products for more than a decade.

We’ve done the most work with AI in the legal market, under the leadership of Bloomberg Law President Joe Breda, who has been intimately involved in guiding the technology that underpins Bloomberg Law since its launch 15 years ago.

Bloomberg Law already uses artificial intelligence to power some of its most innovative workflow technology, such as the award-winning and patented litigation tool, Points of Law, which seamlessly and transparently connects legal concepts to accelerate research, and Draft Analyzer, an automated contract review tool.

For our government clients, we’re using AI to answer their questions about bills and contracts. In the tax market, we’re putting the finishing touches on an AI-powered Q&A function to augment search results. And in one of my recent articles on this platform, I wrote about how our news organization is using automation to cover more routine stories like press releases and hiring announcements so our reporters can focus on more in-depth journalism.

?

Core principles for AI development

Our obsession with improving our customers’ workflows led us to explore and implement AI functionality in our products, but given the expectations of the customers we serve, the key ingredient in our work is a healthy dose of careful consideration. Our approach to AI has been guided by three core principles, informed by feedback from our clients.

First and foremost, reliability is critical. Something that provides the right answer 80 percent of the time will not cut it in our markets. Lawyers need answers they can rely on in front of their clients or in front of a judge, so we’ve taken unique approaches in our use of AI to ensure they can do just that. For example, Points of Law lets users isolate the most-cited language essential to a court’s reasoning by locating and extracting legal standards verbatim from language used in millions of court opinions.

Secondly, we put transparency front and center. Our clients expect us to be transparent about the source of the information we’re relying on to provide answers to them. Lawyers are trained to rely heavily on citations to ground legal arguments. ChatGPT, for example, makes no effort to cite sources and does not provide footnotes or links to where it derived the information it provides. More troubling, it’s unclear where ChatGPT gets its information (when I asked ChatGPT, it informed me: “…The specific sources used to train me are not publicly disclosed by OpenAI...”). Bloomberg Law is in the business of providing authoritative legal information, so the provenance of our data is a big deal to our customers.

Beyond understanding the underlying source of information, our users have expressed that they want transparency into why our algorithms are making suggestions to them. For example, our Brief Analyzer tool provides a bulleted list explaining in plain English why additional opinions, briefs, and other content are suggested to the user by the AI.

Finally, privacy and security are non-negotiable. Our clients deal with highly confidential information and want to use services they’re certain will protect their clients’ confidentiality and their own work product. For example, our Draft Analyzer tool not only uses strong encryption of user documents in transit and at rest, but also ensures that any document uploaded to the tool is deleted at the end of the session.

?

What the future holds

Bloomberg Industry Group has nearly a century of rich, annotated content and data, a decade of experience with AI, and a team of technologists and data scientists with extensive domain expertise, so we’re excited to explore new uses for these technologies.

Our latest solution for storing, searching, drafting, and negotiating contracts, Bloomberg Law Contract Solutions, uses AI to analyze agreements and extract key contractual terms, helping our customers to identify and flag potential risks and obligations. Bloomberg Law Contract Solutions also leverages a patented semantic clause analysis algorithm to offer suggested clauses to improve an agreement during the drafting process. Our team is already working on ways to augment this brand-new product leveraging large language models.

Our parent company, Bloomberg, developed BloombergGPT, the first large language model trained specifically for the finance domain. We’re working closely with the team that developed this model, looking at how we can adapt it and other models for the markets we serve. As always, our solutions will incorporate the hallmark reliability, transparency, security, and privacy we’re known for – and our clients demand.