When AI was the shiny new object, it took a while to really come out and play in the legal space – at least for most teams. For a while, it lived in pilots and proof-of-concept projects, carefully safeguarded from the risks of applying it to real matters.
But, if you ask our 2026 AI Visionaries, that era of experimentation before implementation is coming to an end. Legal teams are starting to really dig into AI and get their hands dirty (in the best, most defensible way, of course).
Three members of Relativity’s 2026 AI Visionaries class – Andre Golueke, Melissa Weberman, and Adrian Agius – are deep in that work. And, though they’re obviously passionate about AI, it’s not something they’re doing just for kicks and giggles. In their view, AI adoption in legal is an inevitability.
“I anticipate that AI will be woven into every stage of the discovery workflow within the next five to ten years,” says Andre, discovery strategy & operational support lead at Google. “With oversight from human teams dictating workflows and assuring quality and adherence to policy, I believe AI can improve all aspects of the discovery process.”
As counsel and head of eData and legal technology, Melissa Weberman has helped her firm, Arnold & Porter, deploy generative AI in practical, defensible ways. She shares Andre’s view on the not-so-distant future, saying that in five years AI will no longer be a separate initiative, but an embedded component of everyday legal tools.
“Firms that do [adoption] well will treat AI like any other high-impact, high-risk system,” Melissa says. “They will define what it is allowed to do, test it, monitor it, and document it.”
These visionaries are prime examples and excellent guides in the art of adopting AI with intention. So today, we’re letting them lead the way for any legal intelligence professionals navigating the AI transition. Here’s what they’ve seen – and what they suggest for you.
That Age-Old Advice: Start Small
One of the clearest themes across all three Visionaries’ perspectives is a preference for deliberate, staged adoption. In other words, don’t expect – or even go after – a big and bold transformation right away.
“Start by thoroughly testing the tools,” Andre suggests. “Focus on ensuring quality (not perfection) and defensibility. Design a validation protocol that meets those standards and have the metrics to back it up. Once you have achieved that level of comfort, you can expand into real world use.”
Melissa shares a similar approach, encouraging legal teams to pick one repetitive, weekly task to test an AI workflow on. Firms that succeed, she argues, are the ones that don’t treat AI as a shortcut; they make it part of their standard process with clear rules, documentation, and a record of all the good (and not-so-good) moments along the way.
“Document the steps, define what ‘good’ looks like, and keep a log of when the tool helped and when it didn’t,” she says. “The most valuable innovations are the ones that make the work more consistent and defensible, not just faster.”
After all, “fast” in the legal industry is next to useless if it introduces inconsistencies or undermines defensibility.
Own Every Outcome
Adrian Agius, director of legal informatics at Gilbert + Tobin, works in a particularly thorny market. In Australia, judicial scrutiny is high, regulatory and client expectations are arguably even higher, and reckless experimentation tends to get punished quickly (while careful adoption is rewarded).
With that, Adrian’s view on AI is clear: “I’m comfortable with AI as an assistive layer inside governed systems, where outputs are observable, challengeable, and reversible. Used well, it amplifies expertise and helps curate ideas,” he says. “I’m far less comfortable when it is treated as an authority. History suggests that outsourcing judgment rarely ends well.”
Adrian shares how AI will support time-intensive, repetitive work, but it cannot replace accountability.
“[Legal work] is high stakes. When something goes wrong, and it generally does, the system isn’t on the hook with the standards set by regulators,” he says. “For as long as we are required to meet these standards, people still have to own outcomes. AI just changes where their time is best spent.”
Invest in AI Literacy
All three Visionaries treat AI literacy not as a one-time training exercise or a backburner project, but as an ongoing professional responsibility.
“AI is here to stay and is being used across the legal industry, so sooner or later it will come across your desk,” says Andre. “It’s best to be prepared when it does.”
Adrian echoes that, noting how tools and processes have been evolving, and how that constant evolution will not change. “The moment you stop learning is usually the moment you should start looking for something different,” he says.
He also returns to the theme of accountability: “We need to practice and execute the skills of verification on the basis that anything can be convincingly wrong. The risk isn’t in experts doing the wrong thing; we have professional obligations to manage that concern. The risk is in well-intentioned people doing the wrong thing confidently. Our obligation is to design systems and education that make that harder, not easier.”
For Melissa, hands-on experience – whether it’s through pilot programs or continuing education classes (or, preferably, a combo of both) – is key to AI literacy. She also recognizes the importance of keeping her colleagues up to speed and is careful to remove any perceived barriers to learning the tech.
“No one needs to be an engineer to [use AI], and creating space for ‘basic’ questions and translating concepts into everyday language makes it easier to validate and document the work responsibly.”
Go Beyond “Legal Work”
It seems counterintuitive, but the legal professionals who thrive with AI, Melissa argues, won’t necessarily be the most technically fluent. They’ll be the ones who can run the project.
“This profession is deeply people-facing in practice. There’s constant opportunity to engage with others, translate between different perspectives, and help teams accomplish difficult projects successfully,” Melissa says. “The work that moves the needle is often not strictly ‘legal work.’”
She points to a few shifts worth making: build strong project-management instincts; learn the technology well enough to ask smart questions; and partner across disciplines to turn complex problems into workflows that are both efficient and defensible.
In other words: AI handles more of the work, but someone still has to hold the whole thing together.
If you’re reading this, chances are you’re already deep in the work of implementing AI. Follow these Visionaries’ path, and you’ll be in good shape.
Graphics for this article were created by Kael Rose.
