The feedback paradox: Why AI critique lands differently

Key highlights:

• • • AI can help remove interpersonal friction during feedback conversations — Attorneys have taken naturally to soliciting substantive critique from AI that covers tone, argument strength, and clarity in ways they might hesitate to request from a colleague or supervisor.

    • The role of feedback-giver remains irreplaceable — The developmental core of feedback still requires someone who knows both the work and the person receiving the feedback, even if AI is used as a sounding board.

    • Feedback culture can be accelerated using human-AI mentorship — AI makes the conditions for receiving feedback easier and can spark shared human conversations about the work rather than replacing those conversations.

---

It’s one of those things upon which most people on any legal team would agree — feedback matters. It’s essential to professional development, to mentorship, and to the long-term growth of attorneys at every level. And yet, for all the emphasis placed on its importance, feedback remains one of the more nuanced and inconsistently experienced dimensions of working in the legal profession.

At my firm, Seyfarth Shaw, in particular, there has long been a deliberate focus on building high-performing teams and developing legal talent in a way that is both structured and intentional. We have designed systems that measure and reinforce feedback because we believe it is central to how attorneys grow and how high-performing teams consistently deliver exceptional client service.

Within the legal profession, however, there are dynamics and circumstances that can often make feedback more complex to navigate in practice.

At its core, a lot of feedback still relies on the comfort level of both supervising and junior attorneys to proactively provide it and seek it out, and that is often easier said than done. Time pressures, the structure of legal teams, and the challenge of balancing candor with constructiveness all contribute to making this an often-difficult task.

---

One of the most common use cases we began seeing was attorneys taking their own drafts and asking AI to critique them — not just for grammar or formatting, but for more substantive reasons such as tone, argument strength, clarity, and impact.

---

Attorneys are trained, as an actual skill set, to question everything and that can naturally extend to how feedback is processed, depending on how it is delivered and by whom. The amount of effort, intelligence, and judgment that goes into legal work is significant, so when that work is challenged, it can feel personal. And on the receiving end, actively seeking out substantive feedback is not a muscle that gets consistently developed in most educational settings leading up to working within a law firm.

The behavioral shift

This is the backdrop against which something genuinely interesting started happening when generative AI entered the picture. Almost immediately, one of the most common use cases we began seeing was attorneys taking their own drafts and asking AI to critique them — not just for grammar or formatting, but for more substantive reasons such as tone, argument strength, clarity, and impact. Traditionally, this is the kind of feedback many of these same professionals might hesitate to request directly from a colleague or supervisor.

What struck me was how quickly and naturally asking this of AI became a default behavior. There was an almost instinctive willingness to let AI review and analyze work product in a way that felt qualitatively different from how feedback had traditionally been experienced in interpersonal settings.

Landmark, meta-analysis research on the effect of feedback interventions on performance found that feedback interventions actually decreased performance roughly one-third of the time, particularly when that feedback shifts attention to the self and triggers anxiety rather than a renewed focus on the work. More recent insights from Harvard University identify three triggers that cause people to reject feedback, including the relationship trigger, in which the reaction is not to the substance but rather to the person delivering it.

And LawyerBrain’s Dr. Larry Richard’s profiling of attorneys adds a profession-specific dimension: Lawyers tend to score lower on resilience, defined as how one reacts to criticism or rejection, while also ranking high in skepticism, an instinct to question assertions rather than accept them at face value. The combination can make feedback both more essential and, at times, more complex to deliver effectively.

Using AI neutralizes many of these dynamics — there is far less perceived pressure, no interpersonal dynamic to navigate, and no relationship to manage in the moment. We are constantly giving AI feedback about what it did right, what it did wrong, and how to improve. Indeed, we often are already flexing that muscle without thinking twice about it.

What the paradox reveals

This is the part to which I keep coming back. The problem was never that lawyers cannot handle feedback; if that were true, they would not be seeking it so readily from AI. The opportunity lies more in the conditions under which feedback is delivered and received. AI did not make people more open to feedback, rather it removed some of the perceived barriers that can accompany feedback in traditional settings.

---

What AI has done, perhaps unintentionally, is create a clearer line of sight into how feedback could work even better.

---

However, removing that friction is not the same as providing what lawyers actually need to grow. AI can tell you that your argument has a structural gap, but it cannot tell you why that gap matters in the context of a particular client relationship, a judge’s known preferences, or the broader strategy of a case. It cannot replace the judgment that comes from a supervising attorney explaining not just what to change, but why it matters to change it, and how to think about it differently next time. The developmental core of feedback still requires a person who knows the work, knows you, and is invested in your growth.

What AI has done, perhaps unintentionally, is create a clearer line of sight into how feedback could work even better. It has identified behaviors — such as seeking input early, iterating quickly, engaging with critique — that firms like ours have long been working to encourage, and made such behaviors easier to access in the flow of work.

Yet, there is also something more subtle happening. With AI, attorneys retain a clear sense of autonomy over the feedback itself. They can take it or leave it, focus on it, or set it aside without any interpersonal consequence.

AI creates a different dynamic: It functions more as a sounding board, essentially another set of eyes on the work. That makes it easier to engage with feedback in a more exploratory way by incorporating what resonates, questioning what is unclear, and testing ideas before bringing them back into a human conversation. In that sense, AI can help build the muscle not just of receiving feedback, but of engaging with it more thoughtfully — including developing the confidence to ask the “why” that is often where the real learning happens.

At Seyfarth, this is where we see a meaningful opportunity to build on an already strong foundation. Our focus has long been on creating a culture in which feedback is expected, measured, and part of how work improves. What AI allows us to do is take that a step further, making feedback more continuous, more immediate, and easier to engage with as part of the workflow.

In the next part of this series, we will look at what we call SEYmultaneous Advancement, our way of intentionally bringing AI into the feedback process — not as a replacement for human interaction but as a progression — and in true AI form, iterating with the human-in-the-loop along the way.

---

You can find out more about the impact of AI and other advanced technologies on the legal profession here