From Courtroom to Codebase: Are You Ready for AI in Adjudication?

This is the eighth article of a 10-part series on how technology is transforming evidence, litigation, and dispute resolution. In this installment, we look at the benefits and risks of using AI in adjudication. Other articles in the series can be found here.

Are we ready for AI judges? Are we ready for algorithms to start passing judgment and rendering decisions?

Artificial intelligence, large language models (LLMs), neural networks—these tools have evolved rapidly in recent months and years to take on more and more tasks. They are spurring innovations and reshaping workflows across countless industries, including the legal sector, but also generating significant errors in certain situations. Notably, there have been at least 733 cases worldwide where AI hallucinations featured in court submissions as of January 2026.

Given this mixed track record, can we envision a future where AI might play the ultimate role in dispute resolution?

Well, yes. In fact, it’s already happening. In Canada, Australia, China, and Mexico, various forms of Automated Dispute Resolution are already being used. 

Examples from around the world

In Canada, British Columbia’s Civil Resolution Tribunal (CRT) is considered a pioneer of online dispute resolution with appealable automated decisions on small claims up to $5,000, and motor vehicle injuries up to $50,000.  It relies on decision trees, expert systems, and automated triage to resolve small claims and strata disputes, with many cases concluding before human adjudication is required.

Australia’s state-based civil and administrative tribunals are considered some of the most advanced public users of algorithmic dispute-resolution workflows. State tribunals such as NCAT, VCAT, and QCAT rely on rule-based intake, automated triage, and structured decision pathways to resolve tenancy, consumer, and small-value claims. While human adjudicators retain final authority, algorithms increasingly determine how disputes are often categorized, processed, and resolved without a hearing.

In the United States, algorithms already resolve entire categories of disputes, particularly high-volume, low-value matters. Courts and municipalities use rule-based online dispute resolution systems for traffic citations, parking violations, and small civil claims. Administrative agencies rely on automated eligibility and decision engines to adjudicate benefits and regulatory disputes. While human judges retain final authority, algorithms increasingly determine how disputes are categorized, processed, and often resolved without a hearing.

In the private sector, companies like Amazon and eBay are using online dispute resolution (ODR). As early as 2010, eBay disclosed that it was resolving roughly 60 million disputes per year, with approximately 90% handled entirely by software. Notably, eBay no longer publishes these figures. This omission reflects how normalized large-scale, algorithmic adjudication has become. Subsequent statements by Colin Rule and other ODR experts continue to describe the system as operating at massive scale, with human review reserved for a small minority of cases.

Part of the answer lies in realizing that not all cases are created equal. AI adjudication seems well suited for cases that are high-frequency and low-complexity. This is where efficiency is the primary goal and the outcomes do not involve significant human rights or liberty concerns. Combine this with remote access to the hearing, and you have a formula for speeding up the docket and improving access to justice. 

For example, debt collection and eviction cases represent a large percentage of municipal dockets and are great candidates for ODR. A 2025 Stanford Law School report about the Los Angeles Superior Court docket report stated that Los Angeles has 12,000 evictions (writs of restitution), 45,000 debt collections, and 30,000 debt-buyer cases that are defaulted due to lack of  engagement. AI also seems applicable to resolve minor traffic violations, which are minor in scale but not minor in quantity.

Other candidate cases are traffic cases, unemployment eligibility, standardized contract disputes, and even portions of international commercial arbitration, especially where case histories provide a large body of data that can used for training, where the cases are very repetitive, and there is a minimal or zero need to interpret human motivations, credibility, or ethical considerations.

High demand for adjudication resources

Globally, many cases that could be adjudicated are not due to costs and delays. The World Justice Project found that in at least 62% of countries, a majority of people with legal problems who needed a mechanism for resolution could not find one.

Generative AI and other forms of decision algorithms are still in their infancy. Yet they have progressed from struggling with natural language, to taking high school exams (2019), to passing the Uniform Bar Examination (2023). The future promises even more capability.

AI adjudication is already underway, whether you support it or not. And even if your cases are never decided by AI, the foundations it requires—clarity, consistency, and accountability—strengthen traditional decision-making as well. Building processes that work for both humans and machines leads to better outcomes, regardless of whether AI ever enters your courtroom or organization.

Another benefit of AI is that systematizing filing systems produces cleaner, more uniform data, which improves algorithmic performance. At the same time, advances in generative AI and natural language processing make it easier to analyze unstructured, messy data. Together, cleaner inputs and more capable tools lead to stronger, more reliable algorithmic analysis.

Preventing hallucinations

We are also developing systems to control for hallucinations. Although “hallucinations” are currently inherent to the way LLMs work, there are systems already in the marketplace that control for hallucinations—especially where citations and references are concerned. Consider Harvey, Perplexity.AI, or Matey.AI , or newer alternatives such as Reliath.AI. These systems can limit the universe of documents considered, reinforce proper research and citation, or even establish a global gold-standard for certain facts (full disclosure: this author uses Perplexity.AI and resells Matey.AI).

There are millions of disputes, every year, that are candidates for some degree of algorithmic management or adjudication. When the right processes and guardrails are in place, we can use advanced AI tools to give millions of people greater access to justice and make more resources available for those with more complex needs.

AI adjudication is already part of our present. It will surely be a part of our future. In the words of author William Gibson, “The future is already here—it’s just not very evenly distributed.”

Closing thoughts: Join the conversation

This is just one piece of the bigger conversation on the future of evidence. As legal professionals, we need to stay on top of emerging technologies.

Let’s continue the discussion on this LinkedIn post.