87% of Legal Departments Use AI. So Why Does Nothing Feel Faster?

The number everyone is quoting

In March, FTI Consulting and Relativity's seventh annual General Counsel Report delivered the stat that has since appeared in roughly every legal tech keynote on Earth: 87% of general counsel now report generative AI use within their teams, up from 44% a year earlier. The ACC/Everlaw survey tells the same story from a different angle, with active GenAI use among in-house professionals jumping from 23% to 52%.

Congratulations, everyone. We did it. AI is adopted.

Except... if AI is everywhere, why is the contracts queue the same length it was last year? Why does the NDA that "should take five minutes" still take three days? Why does your team still keep a spreadsheet called Matters_FINAL_v7_ACTUAL.xlsx?

Broad is not the same as deep

Here's the part that doesn't make the keynote slides. Bloomberg Law's 2026 State of Practice survey reportedly found that only about a quarter of in-house lawyers use AI daily, and a similar share hadn't touched it in the previous six months. Adoption is wide. It is also about an inch deep.

That's not a knock on legal teams. It's what happens when AI shows up as a chat window instead of as part of the work. A chat window asks every lawyer to remember it exists, decide it's appropriate, copy the right context into it, and trust what comes back. That's four chances to say "eh, faster to just do it myself."

"We have AI" is a procurement status. "AI does the first pass on every inbound NDA" is an operating model.

Three signs your AI adoption is a mile wide and an inch deep

• Nobody can name the workflow. Ask five people what AI does for the department and you'll get five answers, all starting with "Well, I sometimes..."
• Your success metric is licenses. Seats purchased is a vendor's KPI. Hours returned, cycle time, and outside counsel spend avoided are yours.
• The power users are self-taught. If your best AI users learned from YouTube and each other, you don't have a program. You have a hobby club.

What the teams getting real value do differently

1. They pick boring, high-volume work first

The FTI report found that summarization is the most common use case (83% using or experimenting), followed by identifying contract clauses (63%). That's not an accident. The quickest wins live in work that is repetitive, well-defined, and easy to check: NDAs, vendor paper triage, clause extraction, first-pass policy questions. Glamorous? No. Measurable? Very.

2. They put AI inside the workflow, not beside it

The difference between a tool people forget and a tool people rely on is usually where it lives. AI embedded in intake, in Word, in the CLM, or in the ticketing flow gets used because it's in the way. If you're evaluating options, start with the systems your team already lives in, like contract lifecycle management or legal intake and workflow, before adding yet another standalone assistant.

3. They govern it like they mean it

Trust is the gating factor. Teams with high confidence in AI output report positive returns far more often than low-trust teams. Confidence comes from structure: an approved tool list, clear rules on what data goes where, a verification standard ("AI drafts, a human signs"), and vendor terms that prohibit training on your data.

4. They measure something, anything, before and after

Pick one workflow. Measure cycle time and touch time for 30 days. Turn on AI. Measure again. You now have more hard data than most departments presenting at conferences this year.

A 30-day plan to go from wide to deep

1. Week 1: Survey the team. Who uses what, for which tasks, how often? You'll find the real use cases in the answers.
2. Week 2: Pick one high-volume workflow and document the current process and baseline metrics.
3. Week 3: Embed AI into that workflow with a playbook, a named owner, and a verification step.
4. Week 4: Measure, share the result with the GC, and pick the next workflow.

The bottom line

The adoption race is over, and everyone got a participation trophy. The next race is about depth: which departments turn AI from something people can use into something the work runs through. Browse the generative AI legal assistants on CorporateLegal.tech to see which tools are built to live inside your workflows, not next to them.