Every legal team rolling out AI eventually notices the same pattern, even if nobody names it right away. Junior lawyers pick up new tools fast. They have the least to unlearn and the most incentive to work quickly. Senior leaders, once they're bought in, delegate freely and set the tone for everyone underneath them. But somewhere in between, adoption stalls. The people with up to ten years of experience are established enough to have a way of working that already succeeds, though not senior enough to simply hand the work off. We often see them hesitating.
This "frozen middle" came up directly in a fireside conversation among in-house legal leaders at a recent AI symposium, moderated by Shashank Bijapur of SpotDraft with David Tudor (Group GC, Nasper) and Annabelle Thomas (Legal Director and Company Secretary, Heineken). When Shashank put the pattern on the panel, it produced one of the more concrete, practical exchanges of the session.
It's tempting to assume resistance comes from stubbornness or fear of obsolescence. The panel pushed back on that framing. Mid-level lawyers are often the busiest people in a legal department. They're the ones producing the work product, managing client relationships, and running matters day to day. Learning a new way of working isn't a low-cost decision for them; it's one more thing competing for time against deadlines that don't move.
There's also a trust gap. A junior associate has less at stake if an early AI-assisted draft needs heavy revision. It's expected that their work gets reviewed. A senior leader can absorb an early misfire because their reputation is already established, and they can also afford to build slowly, testing tools without pressure. A mid-level lawyer, by contrast, is often the one whose name is directly attached to the final work product delivered to a client or the business. The perceived cost of a bad AI-assisted output feels much higher, even if the actual risk is manageable.
Rather than mandating adoption from the middle down, the panel described a different lever: visibility. One lawyer showing another how they used a tool, what it produced, and how it changed their output turned out to be more persuasive than any policy memo.
Annabelle offered a specific example. Her team was deep into a lengthy RFI response seeking thousands of supporting documents and multiple drafting iterations, the kind of project where it becomes genuinely difficult to keep track of whether the response is actually answering the question being asked. Heading into a review meeting, she asked a colleague to check whether they'd missed the mark anywhere. What came back wasn't a quick reassurance; it was a detailed compliance breakdown showing exactly where the response was strong, where it fell short, and specifically what needed to change to close the gap. Annabelle described being genuinely surprised by how sophisticated and strategic the output was, especially as someone who doesn't consider herself technical.
That kind of moment reframes the ask from "learn a new tool because leadership says so" into "here's what became possible once someone figured out how to use it well."
David offered a structural counterpoint: his organization doesn't really experience a frozen middle at all, because the organization itself is flat, and building is expected at every level. When the founder builds constantly and expects the same from the whole team, "middle management" stops being a distinct behavioral category. Everyone is expected to be experimenting, regardless of seniority.
That's not a universally replicable structure since most legal departments aren't built around a founder who codes. But the underlying principle travels: the more building and experimentation is treated as normal at every level, rather than something senior people delegate and junior people execute, the less room there is for a middle tier to quietly opt out.
For legal leaders trying to move their own frozen middle, a few things stand out from the discussion:
The frozen middle is a structural problem with a mismatch between who has the most to lose from a misstep and who's being asked to experiment first. Solving it means changing that structure, not just repeating the instruction to adopt.
Want a full framework for getting your team past AI hesitation?
Get SpotDraft's AI Adoption Playbook for practical steps to build a legal team that's ready to embrace AI at every level — not just at the top.
