Why the AI Spending Spree Might Be LegalTech's Best Thing Yet

Jan de Vries · 2026-08-25

Remember that initial AI hype? The one where every demo felt like magic and budgets seemed limitless? Well, the party's getting a bit more sober. Rising AI costs have become the main conversation in LegalTech, and honestly, that's probably a good thing. We're moving past the 'wow' factor. These days, vendors and clients are sitting down to talk about real business cases and hard numbers—return on investment, or ROI. It's less about potential and more about proof. ### The Wake-Up Call From Big Tech This shift in tone is actually welcome. It comes after a period of what you could call unfettered enthusiasm, which led to some pretty wild spending. Take Uber, for example. Back in April, reports said they'd blown through their entire annual AI budget in just a few months. They had to cap employee spending on hot agentic tools like Claude Code and Cursor. Then there was the 'tokenmaxxing' trend. For a minute there, high AI token usage was treated like a badge of honor, a proxy for developer performance. Amazon reportedly had an internal leaderboard for it. But they dropped it fast when employees started optimizing for usage instead of results. Microsoft sent a clear internal email in early August. They told engineers the goal was outcomes, not tokenmaxxing. They also put caps on employee AI spending. It's a natural step in the technology adoption cycle. As these models get more powerful, they also get more expensive to run. Suddenly, talking about costs and value isn't just prudent—it's essential. ### Getting Seriously Savvy About ROI Flashy sales demos worked in the early days. Now? Buyers are getting rigorous. I've heard of legal teams bringing their own documents and specific questions to sales meetings. They're stress-testing products against the exact scenarios they face daily. Asking for evidence of ROI upfront isn't being difficult; it's being smart. It sets clear expectations from the start. It makes it way easier to later answer the big question: Is this thing actually delivering? Here's the thing—many in-house legal teams never really tracked baseline metrics before. Things like turnaround times or task volumes were just... part of the job. AI tools change that. They create visibility. They make outcomes quantifiable. And it's not just legal teams anymore. Business units are using legal tech too. Sometimes the gains are even clearer there because these teams are directly tied to revenue and costs. Let me give you a concrete example. One insurance claims team was processing about 150 claims per person per month manually. After implementing agentic AI, that number jumped to around 700 claims in the same period. That's not just efficiency; that's a transformation. ### The Vendor's Tightrope Walk Companies are now demanding ROI and proof points after getting those first shocking AI bills. For vendors, the cost pressure is just as real. Their solutions run on large language models (LLMs), and processing those isn't cheap. For LegalTech startups, these costs hit at the core of their business model. Can they scale sustainably? Can they monetize effectively? - **The Per-Seat Dilemma:** Most vertical AI solutions charge per seat. Procurement likes that—it's predictable. But it creates tension. Underneath, the vendor's costs are based on token usage, which can skyrocket. That eats into their margins. - **The Usage-Based Risk:** Switching to a pay-per-use model protects the vendor's margin, sure. But it risks terrifying customers who fear spiraling, unpredictable costs. That's a fast way to lose them to a competitor with a simpler pricing plan. It's a classic tightrope walk. ### Making Smarter Technical Choices So, how do vendors navigate this? They have to get technically clever. It's about using the right tool for the job, not just the shiniest, most expensive one. Not every task needs the most powerful LLM on the market. It's about matching the right-sized model to each specific task. Think of it like tools in a workshop—you don't use a sledgehammer to put in a tiny screw. You can build multiple AI agents for different workflows. Each one could be powered by a different LLM, or a different version of one, optimized for cost and performance. It's a more thoughtful, surgical approach. As one industry insider put it, *'The era of blank-check AI spending is over. The era of strategic, value-driven implementation has begun.'* This new focus on cost and ROI isn't a setback for LegalTech. It's a sign of maturity. It means we're moving from a toy to a tool, from fascination to function. And that's how real, lasting change actually happens.