AI is going to give us lawyers our time back. That was the pitch.
Fewer hours on review. Faster turnaround. Less of the repetitive work nobody went to law school for. And to a real extent, it delivered. Contract review that used to take a senior lawyer half a day now takes an hour, maybe less. First-pass redlines get generated before anyone on the team even touches the document. These are real, tangible gains.
So why do us lawyers still feel like we’re busier than ever?
Here’s the reality for most legal teams: AI handles the first pass now. The senior lawyer still reviews it exactly the way they reviewed the manual version, because the approval process never changed. A contract gets summarised in minutes. It still sits in the same queue, waiting for the same person to sign off, because nobody redesigned the workflow around it.
The way the work moves hasn't actually changed. The AI sped up individual tasks, but the approval chains, review cycles, and the escalation patterns all stayed the same. So teams end up running the tool on top of the process the tool was supposed to reshape. The work got faster in places. But the function itself still operates the same way.
We're supposed to be working smarter. Instead we're working twice: once the old way, once the new way, then calling it transformation.
That's the gap I want to dig into here: why speed and change aren't the same thing, and what has to shift underneath the tool for AI to actually earn its ROI.
The data is starting to confirm what many of us already suspect. 87% of general counsel now report using AI within their teams, which is nearly double the rate from just over a year ago. Companies are deploying faster than they are governing or measuring, because the pressure to act is real.
Some of that pressure comes from watching peer legal departments move first and not wanting to fall behind the curve. Some of it comes from higher up: boards and C-suites who’ve watched AI reshape finance and marketing, now asking legal why it hasn’t caught up.
But adoption for the sake of adoption doesn’t hold up to scrutiny forever. That’s why analysts like Forrester predict 2026 to be the year the AI hype period ends. Companies are even increasingly deferring their planned AI spend into 2027 due to unresolved ROI concerns.
The rush to deploy has produced a specific kind of problem, and to grasp why, we have to look at what legal functions were actually designed around.
The traditional legal function’s operating model was built on the concept that expertise was expensive. Expensive in time, expensive in skill, expensive in money.
Contract review? Too costly to happen continuously. It happened when someone needed it, not before.
Obligation tracking? It was tracked in a spreadsheet…or more often, it simply lived in someone’s head. After all, what other option was there?
Compliance monitoring? Most organizations could only check periodically, so issues were identified through audits, reporting cycles, or when something went wrong.
Outside counsel was engaged case by case because keeping that depth of judgment on staff full-time was not realistic for most organizations. And legal ops, where it existed, was designed to help legal function do more with limited resources. Because limited capacity was the defining constraint.
Before the AI era, these were reasonable choices for a GC to make. Legal expertise was a finite resource amid constant budget pressures. Now, AI has dramatically reduced the cost of reading, analyzing, summarizing, classifying, and reviewing commercial documents..
It’s an exciting, if uncertain, time. The problem I’m seeing is that organizations are responding to this shift by adding AI to the workflows that the old constraint produced, without stopping to ask whether those workflows still make sense.
In practice that looks like:
This is what I meant earlier when I said the work moved faster, but the way it moves hasn’t changed. And I understand the appeal of that. Faster feels like progress. But speed alone doesn’t fix a function that was never redesigned. That’s the real work ahead.
When the AI underdelivers, the instinct is almost always to look at the technology. Maybe it’s an unsuitable platform, the wrong vendor, or something that needs revisiting in the configuration.
In my experience working with clients and running our Legal Center of Excellence, the problem isn’t typically the tool. More often, the tool is doing exactly what it was asked to do. The issue is that the function never gave it a different job. That’s a harder thing to sit with, because it means the constraint isn’t the technology itself. It’s the way the technology has been embedded into existing processes.
This isn’t a new mistake. Decades before anyone was talking about generative AI, Michael Hammer argued in the Harvard Business Review that companies were spending heavily on technology while achieving little more than faster versions of the same broken processes. He famously described it as “paving the cow paths”. Change the technology and the diagnosis remains largely the same.
The data supports this. 69% of legal professionals are using AI individually, but only 34% of firms have formally adopted it at the organizational level. Many still lack clear ownership for what happens after the tool goes live. Without that ownership, adoption slows, accountability gets blurry, and the expected returns fail to materialize.
Which brings us to the harder part.
The first step is recognizing that the real challenge is in redesigning the environment around AI: the data, the workflows, and in some cases the operating model itself. The second step is recognizing that there is no universal blueprint. What works for a lean legal team supporting high-volume commercial contracting will look very different from what works for a heavily regulated function managing complex multi-party agreements.
Fixing that means deliberately redesigning the function, rather than layering AI onto existing ways of working. That is the harder conversation, and the one legal leaders need to be having now.
I'll be honest. "Rebuild the function around AI" is one of those phrases that sounds important and means almost nothing until you define it. From what I’ve seen, the conversation usually comes back to three areas.
The foundation is your data. Getting contract data into a state where AI can work with it reliably (structured, clean, consistent) is a significant undertaking on its own. It requires leadership, sustained attention, and usually far more effort than most implementation plans acknowledge.
Next comes the workflow itself, not just the data feeding it.
Look at the workflow the tool is supposed to be improving, and ask a genuinely open question: does that workflow still make sense, given what's now possible? Or are we just doing the old thing with a better engine? A contract that still needs sign-off from the same three people, in the same order, even though risk flagging now happens instantly. Compliance monitoring that still runs on a quarterly cycle, even though the tool could flag issues the moment they appear. That's the old thing with a better engine. One useful test: are approvals still there because they add value, or simply because they’ve always been there?
Then, go back further still.
Look at what you're managing, look at who owns it, and look at what you're measuring as success. This is the layer most functions skip entirely, and it's the one that determines whether any of the above sticks. In my experience, these efforts rarely succeed without clear executive ownership. Someone has to be accountable for how the function changes around the tool, and not just getting the tool live.
What happens after the shift
When redesign is done thoughtfully, the difference is visible. Information that was previously trapped in documents becomes available continuously, and legal teams spend less time finding answers and more time acting on them. But getting there takes longer than any implementation timeline suggests, because it requires changes to data, workflows, and ownership, not just technology.
If the gap between what you've invested in AI and what your function actually looks like day-to-day feels familiar, you're not behind. You sit in the same room with other legal and procurement leaders right now.
Adding AI to a function that was designed around an old constraint makes it faster. It doesn't make it different. And faster, on its own, is not the ROI we are looking for.
At this point, AI adoption is no longer the relevant question. The more pressing question is: has AI really changed the way we work? If it has only accelerated existing processes, the real opportunity may still be ahead.
Comments? I’d be interested in hearing your perspective. Feel free to reach out at hello@execo.com
Because most teams add AI to the workflow the old constraint produced, without redesigning the workflow itself. The same approval chains, the same batch schedules, the same escalation paths stay in place. AI makes each step faster. It doesn't change how the work moves.
The function never gave the tool a different job. AI is usually doing exactly what it was asked to do, running the same review-on-request model or the same periodic compliance check, just faster. The constraint isn't the technology. It's how that technology gets embedded into a process that was never re-examined.
Three places, in order. First, contract data: structured, clean, and consistent enough for AI to work with reliably. Second, the workflow itself: ask whether each approval step still adds value or just persists out of habit. Third, ownership: someone accountable not only for the tool going live, but for how the function operates afterward.