Our Take
Law firms think they're buying AI platforms; they're actually signing up for open-ended token consumption that will blow past their tech budgets once adoption moves beyond the current 1% of active users.
Why it matters
Most lawyers still barely use AI, so today's costs feel manageable. But the moment a firm's workforce shifts to routine agentic workflows, the token bill becomes the dominant line item in legal tech spend—and most firms haven't modeled for it.
Do this week
CFO: Model your law firm's token costs for 2027 and 2029 based on projected power-user adoption rates (not today's actual usage) before renewing any platform contracts.
Artificial Lawyer estimated frontier model costs for the ten largest law firms
Artificial Lawyer asked ChatGPT to calculate total token costs plus enterprise licensing fees for large law firms under two scenarios: current usage patterns and usage three years forward, assuming increased adoption of frontier models as primary AI providers (per the analysis). The estimates carry wide ranges because law firm AI adoption is highly non-uniform; a 3,000-lawyer firm may see only half its attorneys using AI intentionally, and of those, only a quarter are power users.
The core finding: token costs alone will grow substantially as usage shifts from occasional document review to routine, data-heavy agentic workflows. The analysis assumes law firms continue paying for existing tech stacks—Microsoft 365, document management systems, legal research platforms, billing software—alongside new AI spend. Token costs become the accelerant, not the bulk, but they expand total tech budgets beyond where firms have currently budgeted.
For context, Artificial Lawyer estimates fewer than several thousand lawyers globally out of roughly 12 million are true AI power-users today. That percentage is tiny, which means current token spend is low. The scenario modeled shows costs rising sharply as that ratio inverts over the next three years.
Platform vendors are absorbing token costs now; firms will pay later
Legal tech companies offering all-in platform deals today may be profitable on low-usage customers but will face margin pressure as those same customers mature into power-user cohorts. Token costs will not disappear; the question is only who bills whom.
The analysis points to an inevitable shift: legal tech vendors will need to either pass token costs directly to firms (analogous to buying an iPhone and separately paying for data) or embed them into tiered, usage-based pricing. A firm that thinks a $1 million platform deal is expensive today may find it was actually a good price once its lawyers are running agentic workflows at scale.
Single-vendor lock-in with Claude or OpenAI amplifies this risk. Platform vendors can route work across multiple models to optimize cost; firms locked into one provider cannot, and building internal routing systems is expensive for firms without dedicated ML infrastructure.
Start modeling token economics now
Law firm finance and technology leaders should project power-user penetration rates and token consumption per user for 2027 and 2029, not assume costs will remain flat. Include scenario planning for agentic workflows, which consume more tokens than single-request document review.
When evaluating platform contracts, ask vendors explicitly how they will handle token cost inflation and whether they offer fixed pricing caps. Audit current legal tech spend to understand where token costs will be additive versus replacing existing tools. If a platform is genuinely cheaper than rolling your own Claude or OpenAI subscriptions, quantify the model-routing and governance value it provides; that margin buys you time as token prices rise.
Most firms do not need to panic today. Adoption is still sparse. But waiting until half your workforce are power-users to model these costs will mean painful budget expansion on short notice.