Without a crystal ball, it’s impossible to predict with absolute certainty the outcome of any legal process. But AI is getting pretty close.
A startup in Israel has developed an algorithm that is currently 85 percent accurate in assessing liability and in predicting how a case will conclude. That number will only increase as the AI handles more cases.
“Litigation is a very unknown territory,” says Yariv Lissauer, CEO at Haifa-based Canotera and a former lawyer with vast experience in the startup world.
“There are a lot of unknowns, a lot of complications, a lot of regulations and decrees. But at the end of the day, most cases are based on facts that happened in the past and that need to be put into a regulatory framework that already exists. When you step into court, you have no idea how you will get out. That’s an anomaly we’re trying to fix.”

Canotera scours the available documentation for every comparable case – a task that could tie up a team of lawyers for weeks – analyzes patterns and provides its best assessment of how the case will unfold and be resolved.
It’s a potential gamechanger for anyone involved in litigation who agonizes over whether to fight or settle.
Canotera AI does the heavy lifting. It presents a solid, objective, evidence-based prediction about the outcomes of legal disputes. This helps lawyers decide which cases to take on, what to charge, how to proceed and how to manage risk.
Lawyers’ best friend
“We are not replacing the experienced knowledge and professionalism of the lawyers,” Lissauer tells ISRAEL21c. “We are their best friends because we massively expand the breadth of precedents or facts that eventually a litigation outcome is based on. And we do it in a very scalable manner.”
There’s still space for the instincts of a seasoned professional. It’s just that decisions can be based on a far bigger pool of facts — a vast database of similar cases and how they’ve been resolved in the past.
“When you step into court, you have no idea how you will get out. That’s an anomaly we’re trying to fix.”
A typical case could involve a shopper falling and getting hurt in a mall. The shopper claims the mall was negligent in preventing such an accident and is seeking damages.
“The mall needs to decide whether it wants to settle, or whether it wants to counter my arguments,” says Lissauer. And if it counters, whether to use its own lawyer or to outsources the work.
“If they have our system, it will retrieve all the applicable data from the claim form, and compare it with all the historical cases that have been resolved, either through a court ruling or through a settlement, so that we can also identify or predict the outcome.
“It could be that based on the historical data none of that will happen and the parties will settle after the claim form is submitted. But assuming that we have to go through the full dance of the litigation process that will end up in a court ruling, we outline all the various phases and how long they will take,” says Lissauer.

Perhaps most importantly he adds, the system can predict whether there will be payment from the defendant to the plaintiff and, if so, how much it will be.
“And we also provide the reliability ranking for the results that we predict.”
First client
Canotera uses large language models (LLMs) — the kind of tech that powers ChatGPT – as well as geometric machine learning (which can compare similar cases despite their differences) and conformal analysis technology, which can quantify uncertainties without making strong assumptions about the data distribution.
The company has incorporated existing technologies into what Lissauer calls a “rather robust and complicated technological funnel.”
The company was founded in 2023, has attracted $1.6 million in funding, and has its first client in place: an insurance company in New York. Canotera’s SaaS (Software as a Service) is available for a monthly charge.
Insurers deal with huge volumes of compensation claims and constantly have to decide whether to settle or fight. Canotera has taught its AI everything it needs to know about New York law.
But every US state has its own legal system, so it’ll need to learn about the specifics of any other state where it’s deployed. That’s not a straightforward job because there is no centralized repository for all the relevant data. The AI will have to start more or less from scratch for legal systems in other countries.
Lissauer says he is “not too concerned about competition,” since only a handful of companies make predictions based on the letter of the law, and there’s no shortage of opportunities for all.
But here’s a final and intriguing thought: What if both sides in a legal dispute were to use Canotera?
Both the defendant and plaintiff would get the same prediction (allowing for slight variations in the data feeds), says Lissauer.
“If I know what the result is, and you know what the result is, and both of us trust the generator of the outcome, then we may just shake hands and decide on it and save all the litigation expense.”
For more information, click here.