Can Artificial Intelligence Make Bad Decisions About Contract Analysis?

How attorneys review, summarize, and abstract contracts has undergone some extensive changes in recent years. Technology-assisted structured data reviews changed the game, and now artificial intelligence promises to finish the transition into faster, more efficient, and more detailed contract insights.

However, there are limitations to what artificial intelligence can do.

At the bottom of any ChatGPT thread, you’ll see this message:

It’s a reminder that this technology, even with the power of large language models (LLMs) and massive amounts of textual data, cannot deliver accurate and cost-effective results without some guidance. Here’s what generative AI can get wrong and why attorneys are still relevant in contract review.

How Artificial Intelligence Is Limited for Contract Review

Artificial intelligence isn’t an easy button that resolves challenges in an instant. It’s a very potent technology, but one that still requires humans to wield it. Users need to ask questions, assign tasks, and set parameters before AI can undertake them. We’re still years away from proactive algorithms fully anticipating which contracts you need to review in advance of your needs.

Then there’s the matter of AI hallucinations. When ChatGPT was first released, journalists and researchers were bombarding LLMs with prompts in the hopes of testing the limits of the technology and producing false information. Often these issues are caused by inaccuracies or biases within the training data which prompt the models to establish incorrect patterns.

Though AI has made significant leaps forward, some experts suggest hallucinations happen about 3% to 10% of the time, even after testing models are fine-tuned and adjusted. As Mohamed Elgendy, co-founder and chief executive of Kolena Inc., put it, “As more models are built, new hallucination types will appear, and scientists will find themselves racing to stay up to date with the latest ones. The rate of hallucinations will decrease, but it is never going to disappear — just as even highly educated people can give out false information.”

In our own projects, we’ve encountered AI misinterpreting or even completely missing key legal terms from a contract. And there are a few different reasons why that happens. Something as simple as bad scans can cause AI to miss entire provisions within a contract. Again, poor training data is a common reason why AI extracts the wrong information or misinterprets pivot legal clauses.

These types of errors can lead to missed deadlines (if they’re caught before delivery, they can often require rework) or bad decision making across the business since the underlying data is wrong. In the legal function, even a single mistake can have significant consequences.

Why Keeping a Human Review Is Still Important

Rather than throw the overwhelming good out with the bad, legal leaders need to adjust how they think of this technology. Artificial intelligence is an incredible way to accelerate your reviews, and you can trust the accuracy of your insights if you keep a human in the loop.

When attorneys find incorrect information in structured contract data, they can rework and remediate the errors or false positives. This can save your organization money as you avoid basing decisions on factually inaccurate reports.

At an organizational level, when you add humans (or specifically trained attorneys) to a review, they add a layer of trust and provide comfort in the accuracy of the output, which can transform change management efforts when integrating a new CLM platform or other enterprise-level technology.

Want to hear more about what happens when AI goes wrong and why you need humans to avoid contract analysis mistakes? Watch our webinar, “The Human Advantage: Why Expertise Matters in AI Contract Review” for more insights.

Related Articles

4 Ways Machine Learning in Law Boosts Accuracy & Compliance

AI Contract Review Has Its Limits: Learn How to Maximize Your ROI

How the Right Contract Data Simplifies Your Quest to Reduce Legal Costs