How A.I. Invoice Review Actually Works

Sinead Kenny, Director, Customer Insights at Brightflag

Updated December 25, 2023

It’s happening again. A genuine innovation is morphing into a marketing buzzword and appearing in product descriptions where it doesn’t belong, leaving those who may benefit the most understandably confused.

This time it’s artificial intelligence (A.I.) in legal operations.

So to help get the conversation back on track, we’ll clarify the application we know best: A.I. invoice review. That way legal departments can better assess the technology’s business value.

But before we do…

A.I. Defined: Key Terms to Know

Artificial intelligence (A.I.)

The programming of computer systems to perform actions typically requiring human intelligence, such as sensory processing and cognitive problem-solving.

Natural language processing (NLP)

A specific branch of A.I. that directs computer systems to autonomously recognize, interpret, and respond to human language.

Machine learning (ML)

A specific branch of A.I. that directs computer systems to autonomously interpret data, learn through experience, and improve the accuracy of their outputs over time.

Supervised machine learning (SML)

A subset of machine learning focused on training a computer system to recognize inputs and predict outputs by supplying it with examples of known input-output pairs (also called “labeled” data).

A.I. Applied: Legal Invoice Review

Most corporate legal professionals would love to reduce their role in the tedious process of invoice review. But what does it look like when A.I. picks up some of the slack?

By combining NLP and SML, Brightflag developed a solution that interprets legal invoices with the speed of a machine and the intelligence of a lawyer.

It starts with NLP, as our software reads line item narratives and recognizes key phrases. Next, that text is classified into corresponding legal tasks and activities. And it’s this step where the depth and quality of our training make all the difference.

Brightflag has been refining its SML model since 2014, presenting our software with an ever-expanding set of labeled training data so it can learn to automatically classify an ever-expanding range of real invoice text into the correct categories.

The pace of its learning has been exponentially accelerated by a vital layer of human insight. Our internal team of lawyers act as expert supervisors, reviewing cases where the A.I. is not yet confident in its interpretations and providing corrective feedback that helps the software grow smarter every day.

A.I. Invoice Review Advantages

The initial advantage of A.I. invoice review is two-fold: Complete accountability achieved in record time. The software reviews every character of every invoice while the humans work on more creative, interesting, and valuable tasks. (#NoMoreSpotChecks)

But increasing the scope and speed of oversight is only a winning proposition for in-house legal teams if quality is maintained along the way. Brightflag customers consistently meet (and typically exceed) that standard for three main reasons.

The first A.I.-driven improvements to invoice review accuracy stem from the simple fact that software doesn’t tire. Whereas human eyes might get heavy and human minds might wander toward the end of a 23-page invoice (or 80-hour workweek), A.I. reads every line item with the same precise logic.

Brightflag’s solution also improves accuracy by eliminating any dependency on invoice coding from lawyers. Instead of relying on limited and error-prone task-based billing systems (ex. UTBMS), our A.I. goes straight to the primary source and reads exactly what was written in the full narrative description. This captures the truest portrait of the services performed.

And once again, it’s essential to acknowledge that the quality of all this automated decision-making has been bolstered by seven years of extensive training and continues to be refined with ongoing human supervision from our internal team of lawyers.

That all adds up to a radically improved legal invoice review experience for Brightflag customers. Scope, speed, and quality go up. And thanks to the resulting time savings, operating costs go down.

General Counsel Tara Young explains how Brightflag's invoice review capabilities have helped level up her legal team's ability to track legal spend.

The Value of A.I. Invoice Review

The ultimate goal of legal invoice review is not merely quick, accurate, and cost-effective reading. The business process is only a means to an end, helping corporate legal departments answer four critical questions:

Who worked on the matter?
How long did they work?
What exactly did they do?
Did the work align with expectations?

The value of A.I. invoice review is its ability to provide clear and confident answers faster than ever before.

Once you have those preliminary answers, suddenly you can start asking much more strategic and impactful questions like:

Which vendors aren’t honoring our billing guidelines?
How much of our litigation spend goes to discovery?
What is the most cost-efficient way to resource IP matters?
When does it look like we’ll hit our annual budget number?

And that’s how legal departments start advancing away from merely paying bills and toward proactively managing spend.

On the Horizon: Generative A.I.

In recent years, the rise of generative A.I. has taken the world by storm, led in large part by the popularity of OpenAI’s ChatGPT model.

Generative A.I. solutions take what they learn from their ML training and use it to create original content, identifying patterns within their inputs, and extending those patterns into new outputs that can include text, audio, or visual media.

The impacts of generative A.I. on legal departments will likely extend well beyond just invoice review in the years to come, as the technology is very much still in its infancy.