The Next 10 years in Legal AI

Author: Carina Negreanu, Vice President of Artificial Intelligence

Today’s AI marketplaces are overwhelmed with hype-driven messaging. Generative AI is evolving so rapidly that seasoned tech professionals find it hard to keep up, and experienced decision-makers in other domains find it difficult to separate truth from fiction.

So how can you understand what to expect today and tomorrow from Legal AI tools? We believe it’s important to have a structured framework for understanding progress in Legal AI — to keep both ourselves and our competitors honest, and to help users make purchasing decisions that work for them.

Robin AI proposes five milestones, or pillars, of Legal AI progress.

‍

Where are we today?

In 2024, AI is smart enough to extract insights from legal documents, allowing AI to take on the role of a partial administrative assistant, allowing legal professionals to cut down the time taken to put together documents.

The technology is reliable enough that lawyers and paralegals can extract textual content from important documents for research purposes, using citations to help them check if an AI hallucination has occurred.

Contracts have become a focal point of legal AI tools. That’s because AI can help extract new insights from these documents – scanning them at scale to spot new business opportunities, and delivering this analysis much quicker than humans, creating scope for cost-savings.

Legal AI cannot yet offer fully customized recommendations to its user. The parties to a contract often have complex, detailed, and nuanced positions and also legal teams have strict in-house guidelines. In these situations it is not sufficient for an AI contract edit to be accurate, it also has to be stylistically acceptable for a given lawyer or legal team.

Such prescribed views make it difficult to scale suggested edits. Solving this is a priority for Legal AI providers today.

‍

Where we see legal AI going in 2025 and 2026

AI systems will continue the trend of greater accuracy and greater personalization, across a growing range of business and client types. That personalization will come through work at both the foundation model and application layers of the market.

The increasing role of AI in legal workflows will start to transform legal roles in the course of 2025 and 2026, particularly in legal departments of large companies and financial firms of all sizes.

Legal AI tools will demonstrate clearer and higher returns on investment — by simplifying day-to-day work and free up time for tasks that require higher-level thinking.

Legal professionals will be forced to adapt their skills.

The role of paralegal will shift from its traditional form. Time-intensive tasks previously owned by legal assistants and paralegals such as filing basic court documents and drafting and triaging emails will be undertaken by AI assistants under human supervision.

Paralegals will update their skillset to become competent legal prompt engineers through training and ongoing use of AI tools to maintain experience.

Those who become highly competent legal prompt engineers will begin to gain capabilities which allow them to outperform even senior lawyers who do not use AI across some legal tasks.

New generations of AI models will enable more complicated and multi-modal legal content to be reasoned over which will help businesses identify income opportunities and improve workflows tremendously.

‍

Where will we be in 2030?

Many businesses and individuals adopt five-year growth plans: what if we adopted the same framing for Legal AI?

We’d see that AI will move from assistant to active partner in legal work.

AI systems will be able to self-orchestrate to resolve a number of complex tasks, plan next steps and handle the kind of contract drafting and reviews that typically slow lawyers down in 2024.

These AI agents will be aware enough to know when to pull-in human lawyers and paralegals for reviews and feedback.

That means AI agents will have fully replaced human paralegals on specific tasks such as scheduling meetings or taking notes.

The more tasks legal AI systems can complete on their own, the greater the pressure on the billable hours business models of law firms.

Billable hours will remain for some, but it will be challenged as the leading system for valuing legal services output.

‍

Legal AI in 10 years

A decade from now we can assume that AI capabilities will have reached the stage where they can be considered fully-fledged assistants or partners in legal work — depending on what is required.

Human lawyers will focus on creative and strategic workflows and how to manage AI agents.

Those without any legal academic qualifications will routinely use legal AI tools to obtain legal advice and support with a high degree of competence and confidence.

Error rates in legal services work will be greatly reduced, while the availability of legal services will be greatly expanded.

For example, AI systems will be able to better assist areas of law that are poorly funded today: such as human rights law, and pro-bono work to ensure a greater form of equity in these practice realms.

The way we practice and learn the systems of law will be unrecognizable –  legal education will have transformed from memorization and research skills to critical thinking and new skills related to engaging with AI systems and verifying AI output  in order to succeed.

We’ll know at this point where we have incorrectly predicted legal AI trajectories, giving us a clearer ability to forecast the next 10 years in Legal AI, than the 2025 to 2035 window.

‍

Conclusion

Benchmarks for Legal AI are important for understanding how the technology can better serve society and for easing anxieties about its implementation.

In the legal industry, these four stages of AI show us a way to determine success and where to push for progress.

AI leaders across other industries can create their own benchmarks for success in tandem with a product roadmap to ensure they are building and innovating at the highest level while taking a human-centric approach to AI.

‍