Why lawyers overestimate translation ease

This week we welcome Paula Reichenberg, founder of Neuron, for a sharp and thoughtful conversation about legal translation, artificial intelligence, and what happens when professional expertise collides with tools that look polished but still miss the mark. Paula shares her path from M&A and capital markets law into business school, legal services, machine learning, and finally legal tech entrepreneurship. What started as frustration with inefficiencies inside law firms grew into a translation business, then evolved again as machine translation improved and forced a harder question about survival, adaptation, and quality.

Paula explains how her early company, Hieronymus, found success by handling sensitive, high-stakes legal translations in Switzerland, especially where precision and confidentiality mattered most. But as machine translation improved, the market for average work started to disappear. Clients began doing more on their own, leaving only the hardest, highest-value assignments for specialists. Rather than ignore the shift, Paula leaned into it. That decision led her back to university, into data science and machine learning, and toward building Neuron, a company focused less on replacing expertise and more on improving the process around imperfect AI output.

A central theme of the discussion is the uncomfortable truth that many users do not care as much about excellence as professionals do. Paula makes the point with refreshing honesty. AI often produces work that is mediocre, but for a large share of users, mediocre is enough. That creates both a market shift and a professional dilemma. In legal translation, as in legal drafting more broadly, the issue is rarely whether AI produces something flawless. The issue is whether the user notices what is wrong, has the time to fix it, and has the systems in place to improve the result efficiently. Paula argues that the real value is not in claiming perfection. It is in helping experts find the mistakes faster, correct them with less pain, and avoid wasting hours doing work that feels like cleanup on aisle five.

The conversation also digs into trust, user behavior, and the strange authority people give to AI-generated answers. Paula recounts how, in one negotiation, a party trusted ChatGPT’s answer more than a human tax lawyer’s detailed explanation, even when the AI response was wrong. That anecdote opens up a broader discussion about confidence, presentation, and why polished outputs often feel more persuasive than expert judgment. Greg and Marlene connect that idea to legal systems, translation quality, and access to justice, especially where technology might offer better service than overworked and underfunded human systems. The result is not a simple pro-AI or anti-AI position. It is a grounded look at where human excellence still matters, where automation fills gaps, and where the future may split between mass-market convenience and premium, highly tailored expertise.

Looking ahead, Paula sees consolidation coming to legal tech, along with a growing push toward seamless interfaces that bring best-in-class features into one place. For Neuron, that means becoming an embedded layer inside other legal tools rather than forcing lawyers to juggle yet another standalone platform. Her crystal ball view is both stylish and sobering. The legal industry is not simply moving toward automation. It is sorting itself into tiers of service, quality, and expectation. And if Paula is right, the future belongs to those who understand where “good enough” ends and where true expertise still earns its premium.

Listen on mobile platforms:  ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Apple Podcasts⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ |  ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Spotify⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ | ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ | Substack

[Special Thanks to ⁠Legal Technology Hub⁠ for their sponsoring this episode.]

⁠⁠⁠⁠⁠Email: geekinreviewpodcast@gmail.com
Music: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Jerry David DeCicca⁠⁠⁠⁠⁠⁠⁠⁠⁠ Transcript