According to Financial Times News, in-house legal teams at major tech and finance firms are now actively building and deploying generative AI to automate core tasks. At Salesforce, Chief Legal Officer Sabastian Niles is developing “AI agents” that can make decisions and take action with less human involvement, focusing on compliance checks and handling routine sales queries. At Nasdaq, Chief Legal Officer John Zecca has integrated the risk team under his purview and predicts AI will shift more work like discovery and due diligence in-house, changing the dynamic with outside counsel. Meanwhile, CrowdStrike’s legal chief Cathleen Anderson says her team of over 80 has trained on generative AI and has more than 100 legal AI agents in development, speeding up patent work and contract review. These teams are also grappling with how to build ethical safeguards and manage the blend of human and automated work.
The move from assistants to agents is a big deal
Here’s the thing: everyone’s been talking about AI summarizing documents or drafting emails for a year. But what these legal teams are describing is different. They’re building “agents” – systems that can execute a multi-step process, like running a full compliance review or handling a customer query from start to finish. That’s a shift from a tool that helps a lawyer to a system that is the lawyer for specific, bounded tasks. At Salesforce, they’re literally coding legal and ethical guardrails into autonomous AI products. That’s not just using tech; it’s becoming a core part of the product development lifecycle. It turns the legal department from a cost center into a value-driver for the entire business’s AI strategy.
This will fundamentally reshape the legal ecosystem
John Zecca’s point about the balance of work is probably the most profound insight here. For decades, the trend was to bring specialized legal expertise in-house. Now, AI is poised to bring entire processes in-house. Think about it. If an AI agent can reliably handle the first pass of document discovery or due diligence, why pay a law firm associate hundreds per hour for it? The outside counsel’s role gets pushed “up the stack” to pure strategy, high-stakes litigation, and complex negotiation advice. The corporate lawyer becomes a product manager of AI systems and a strategic advisor internally. That’s a massive change in the economics and structure of legal services.
The new skillset and the hidden data grind
So what does this mean for the lawyers themselves? They’re not being replaced, but their job description is getting a total rewrite. Cathleen Anderson says lawyers now need to work “more like a product manager.” That’s huge. It means understanding development cycles, user needs, and iterative improvement. Sebastian Niles talks about learning to manage the mix of automation and human insight as a critical decades-long skill. But there’s a less glamorous, crucial challenge they all mention: data. Anderson pointedly talks about collecting and managing legal data from everywhere – contracts, briefs, billing software – to fuel these AI systems. The real bottleneck for corporate legal AI won’t be the technology; it’ll be getting their own historical data into a clean, usable format. That’s a tedious, unsexy, and absolutely vital grind.
A quiet revolution from the inside out
Look, this isn’t a flashy consumer AI launch. It’s a quiet, operational revolution happening inside the legal departments of some of the world’s most regulated companies. They’re not waiting for law firms to sell them AI tools; they’re building their own, tailored to their specific risks and workflows. And because they operate at the intersection of fast-changing tech and heavy regulation, they’re essentially becoming the R&D labs for the future of legal work. The fear is always about job losses, but the reality seems to be a brutal prioritization. Routine, process-driven work gets automated. Human judgment gets reserved for the complex, strategic, and truly novel problems. The lawyers who can adapt to that new world – who can manage agents and interpret their output – will be in high demand. The rest might find themselves on the outside looking in.
