In Legalweek 2026 coverage and related commentary, one statistic moved quickly: AI adoption among corporate legal departments reportedly jumped from 44% to 87% in a single year.
That number deserves scrutiny before it deserves celebration.
The underlying survey was a March 11, 2026 report from FTI Consulting and Relativity based on 224 general counsel and chief legal officers at companies with at least $100 million in revenue and 1,000 or more employees. That means the headline is about large-enterprise legal departments. It is not a measure of the broader legal market. And "adoption" in this context means reported use within teams, not deep workflow integration.
That distinction matters.
What Does Adoption Mean?
If adoption means corporate legal departments that have purchased or provisioned AI tools for their teams, 87% is believable. Enterprise procurement cycles for AI tools accelerated sharply in 2025. Vendors made it easy. Many organizations bought seats the way they buy software: evaluate, approve, deploy, move on.
But purchasing a tool is not the same as using it, and using it is not the same as changing how work gets done.
The more useful questions are downstream of the headline number:
- How many of those departments have changed a workflow?
- How many have retired a manual process?
- How many attorneys use the tools daily rather than occasionally?
- How many have measured whether the tool's output is actually better, faster, or more reliable than the process it was supposed to replace?
Those numbers would be smaller. Probably significantly smaller.
The methodology makes that even clearer. The most common reported use case in the survey was summarization at 83%, followed by contract clause identification at 63% and transcription at 53%. Only 39% said AI was a strategic priority. That is real momentum, but it is not the same thing as saying AI is embedded in legal workflows at scale.
The ROI Question Nobody Can Answer
Another recurring theme in Legalweek coverage was return on investment. Baker McKenzie's Danielle Benecke described 2026 as the year value comes under the microscope, following what she framed as discovery in 2023, experimentation in 2024, and early use cases in 2025.
That framing makes sense as a description of where the industry is. As a strategy, it has a problem.
In the Legalweek recaps and commentary I reviewed, I did not see a clear, credible framework for measuring AI ROI in legal work. The conversations were earnest, but they circled. Firms want to see the return. They are not sure what to measure. They are not sure when to measure it. And the tools have not been deployed long enough, or deeply enough, to generate the kind of data that would make the case cleanly.
This is not surprising. It may even be fine.
Nobody asked for the ROI of email when firms adopted it in the 1990s. Nobody asked for the ROI of document management systems when they replaced file cabinets. Some technology changes are infrastructure. You adopt them because the work requires them, and the organizations that wait for a spreadsheet to justify the transition often adopt later at higher cost with less institutional knowledge.
The firms demanding hard ROI from AI in 2026 are asking a reasonable question at the wrong moment. The answer will come, but it will come from organizations that have been building workflow integration for two years, not from organizations that spent two years waiting for someone else to prove it was worth starting.
Licenses Are Not Workflow
The gap between the 87% number and the actual state of legal AI adoption is the gap between procurement and integration.
A corporate legal department that buys Copilot seats for 40 attorneys has adopted AI by most survey definitions. But if those attorneys use Copilot to summarize emails they would have skimmed anyway, the workflow has not changed. The tool sits on top of the existing process. It makes some tasks marginally faster. It does not restructure how legal work moves through the organization.
Workflow integration is different. It means the AI tool is embedded in the process itself: intake systems that classify and route before an attorney touches the matter, document workflows that draft from structured data rather than from scratch, review pipelines where the AI's output is a step in the chain rather than a standalone suggestion.
The distinction matters because the value of legal AI is not in making individual tasks faster. It is in changing which tasks exist.
A tool that helps an attorney draft a contract 20% faster is useful. A system that eliminates three steps in the contract workflow because classification, first draft, and conflict checks happen before the attorney opens the file is transformational. The first shows up as time savings. The second shows up as capacity.
Most legal organizations are still in the first category. The 87% number counts both. It should not be read as proof that workflow transformation has already happened.
The Pricing Conversation Tells the Real Story
If AI adoption were genuinely transforming legal work at scale, the pricing conversation between firms and corporate counsel would be further along than it is.
In practice, those conversations are still stuck in familiar territory. Firms and clients are negotiating who bears the risk of efficiency gains the same way they negotiated alternative fee arrangements a decade ago. If AI makes the work faster, does the firm charge less? Does the firm keep the margin? Does the client demand a discount? Does anyone actually know how much faster the work is?
These are lawyers. They bill in six-minute increments. They have some of the most granularly measured professional output in the world. And yet there is still very little confidence around how to quantify AI's impact on legal work in a way both sides of the engagement letter accept.
That is not a failure of measurement. It is a signal that the workflow change has not happened deeply enough for the economics to shift. When it does, the pricing conversation will resolve itself because the data will become harder to ignore.
What Actually Matters
The 87% number is not meaningless. It reflects real momentum. Procurement barriers have dropped. Institutional resistance has softened. The tools are inside the building.
But the work that matters now is not buying more tools. It is integrating the ones that are already there into workflows that change how legal work is structured, reviewed, and delivered.
That is a harder problem than procurement. It requires systems designed for legal workflow, not general-purpose tools adapted for it. It requires governance frameworks that scale with agentic capabilities, not chatbot-era policies applied to agent-era systems. And it requires platforms that measure adoption by what changes in the work, not by how many seats are provisioned.
The firms that figure this out will not need to ask about ROI. The firms that do not will still be asking in 2028.
FlowCounsel is the AI-native operating system for legal teams. FlowLawyers is the consumer-facing legal help platform with attorney discovery, legal aid routing, state-specific legal information, and document tools. Neither provides legal advice. Attorney supervision of all AI output is required.