The rollout had been carefully designed.
Start small. Build confidence. Let people experience value before introducing structure.
No overload. No pressure to get it right. Just space to try.
And it worked.
People began using the tools in low-stakes ways — drafting, summarizing, organizing their thoughts before a call. Not perfectly. But enough to see what was possible. You could feel the shift—and see it in the metrics. How people approached their work was changing.
Structured GenAI workflows
Then came the structured workflows.
A new feature entered the picture, designed to solve a familiar hesitation:
What if I prompt it wrong?
The intention was clear. Make the experience more guided. More predictable. Reduce ambiguity, provide a path, make it easier to get started.
And on the surface, it worked. New adopters used the predefined workflows. Tasks got completed. Adoption metrics held up.
But something quieter was happening at the same time.
Hesitation didn't go away
Here’s the thing about the question what if I prompt it wrong? — it isn’t only a workflow problem. It’s a signal. A signal that attorneys haven't yet developed a feel for how GenAI actually works — and that the underlying way of working hasn't shifted yet.
With most legal technology, there’s a correct way to proceed. You run the search. You select the filters. You generate the output. Consistency equals competence.
GenAI doesn’t work that way.
There isn’t a single correct prompt. There isn’t one predefined path. The output depends on how the question is framed, what context is included, and what “good” looks like in that specific moment.
Which means the skill to develop isn’t following the steps.
It’s developing judgment.
When structured GenAI rollouts arrive before that judgment is built, they give people a way to use the tool—without ever needing to think with it. The range of use stays narrow. The tool becomes associated with a handful of predefined tasks rather than something that can reshape how work is approached more broadly.
More options made the problem worse
The hidden cost of structured GenAI rollouts isn’t that they fail.
It’s that they succeed—within a much narrower boundary than what’s actually possible.
Then the workflows multiplied. And things got worse.
Here’s where the story takes an instructive turn.
The natural response to “attorneys are only using a few workflows” is to add more — to show what else is possible, to demonstrate the tool’s range, to give people more paths to follow.
It made the problem worse.
This dynamic is well-documented outside of legal tech. In a now-classic study, shoppers presented with 6 varieties of jam were far more likely to make a purchase than those presented with 24. More options created less action. The researchers called it the paradox of choice.
The same effect shows up when the workflows multiplied.
A handful of workflows feels manageable. A growing library sends a different message: there are so many pathways — what if I pick the wrong one? The breadth that should feel empowering will more likely feel overwhelming. The easiest decision becomes not to choose at all—especially when the cost of choosing wrong feels high.
Too little structure creates hesitation to start.
Too much structure, introduced too early, limits learning.
And too much structure, added too quickly, creates paralysis.
Timing and design matter
So the challenge isn’t structure vs. no structure.
It’s timing—and design.
The question isn’t whether to introduce workflows. It’s when—and on what foundation.
A useful test before adding more structure:
Can attorneys use basic GenAI tools to:
- Address a daily task like emails or summaries
- Iterate to get close to the desired results
- Recognize what improved—and why
- Know when to take over as a human—because more prompting won't make the results better
If not, more structure won’t build confidence.
It will create dependency.
The antidote isn't less structure. It's better sequencing.
Sequence matters more than structure
The firms that sustain momentum past early adoption tend to follow a different sequence.
Not less structure—but delayed structure.
Experience first.
Principles second.
Features third.
Because once attorneys develop the judgment to frame a question, iterate on an output, and recognize what “good” looks like, then workflows don’t limit them.
They accelerate them.
Hesitation won't go away by itself
The hesitation to use genAI doesn’t resolve itself automatically.
People have to stay with the tool long enough to work through its quirks.
What makes the difference between those who persevere and those who drop out early is surprisingly consistent:
People need a reason to start.
A reason to continue.
And visible evidence that they’re making progress.
Without those elements, even the most thoughtfully designed rollout sees drop-off long before real capability develops.
We’ve seen this play out in our own online self-directed Edge program. Most online programs see completion rates below 10%. Ours is above 50%. And 80% of the participant become daily users of GenAI.
That’s not a coincidence.
It’s the result of design: a deliberate sequence of small wins, clear feedback, and visible progress that makes the learning feel worthwhile before it feels easy.
The goal isn’t to remove uncertainty.
The goal of GenAI programs isn’t to remove uncertainty.
It’s to make working through it feel worthwhile.
GenAI tools only become truly valuable when people trust themselves to use them.
That trust doesn’t come from more workflows. It comes from having worked through enough uncertainty — in a supported, well-designed experience — to know how to deliver valuable results despite the quirks of these tools.
If you're designing structured GenAI rollouts in your firm and want to build that kind of foundation, the Edge program was built specifically for this — the judgment, the progression, and the experience of figuring it out in a way that actually sticks.