Download (PDF)
Talking to lawyers about the topic of AI, one generally encounters a palpable fascination and a general willingness to try it out, as well as internal law firm task forces and legal tech labs that are experimenting with AI. This is usually followed by a general statement about how impressed they are with the capabilities of AI and its potential to make our work easier and potentially increase efficiency. But the high-end creativity of our legal advisory business cannot, so the common understanding among most lawyers, be replaced by machines, and certain human skills (empathy, negotiating skills) will certainly not become superfluous due to AI either. So don’t be afraid!
At the same time, these discussions are usually quickly followed by the story of the danger of AI hallucinations and that the whole thing is not yet ready. In general, this attitude towards the topic of lawyers and AI is prevalent: AI makes us lawyers more efficient, but doesn’t replace us. Or: It’s not AI that replaces us, but the lawyer who uses AI. Most of the comments made by the panelists at last December’s “class reunion of the legal market in Germany”, the InhouseMatters conference in Frankfurt, went in this direction (see review). And Bruno Mascello, in his outlook on the key trends in the legal market in 2024 in issue 01/2024 of Deutscher AnwaltSpiegel, is not worried that lawyers will be replaced by AI, but is more concerned about legal professional secrecy and advises using 2024 to consolidate (see here).
Lawyers may face the same fate as translation service providers
There is no reason for such confidence. In reality, the current discussion about law and AI mostly misses the core of the problem. The old world of lawyers and law firms will be over in the foreseeable future. We lawyers haven’t really realized this yet. In reality, lawyers are in a similar position to providers of translation services in times of DeepL. They are in danger of becoming obsolete to a large extent.
Richard Susskind may still be right when he states that the short-term impact of AI on lawyers is greatly overestimated. But above all, he is right when he goes on to say that the long-term impact of AI on lawyers is greatly underestimated (for example, see here). He refers to Ray Kurzweil’s statement that the performance of neural networks doubles every 3.5 months, i.e., in six years it will be 300,000 times greater than it is today. Anyone who has tried out Microsoft Copilot and its already impressive performance with regard to the Outlook, Word, PowerPoint and Excel programs bundled in MS Office 365 can imagine that exponential increases in the performance of AI will then not leave much room for human improvements to the results obtained by AI.
What does this mean for lawyers? At present, the strength of AI, at least in Germany, does not yet lie in answering legal questions. It still lacks sufficient training with the necessary data. However, AI can already be used surprisingly well by lawyers in certain core areas of their work, namely document creation.
Attorney-client privilege and data protection are, of course, issues that must be kept in mind. Uploading non-anonymized data to ChatGPT is out of the question without appropriate safeguards. However, today this is already not an insurmountable hurdle. For example, when using Microsoft Azure, it is possible to ensure that the relevant data is not used for general training purposes. And with the right partners, the necessary agreements on confidentiality protection with regard to § 43e (3) BRAO and on commissioned data processing can already be concluded today. If such contractual safeguards are in place, AI may then be used by lawyers even on non-anonymized data sets.
Impressive increase in efficiency thanks to AI
Summarizing, translating and comparing documents are the least impressive achievements of AI. Things get more interesting when it comes to analyzing documents with the help of AI. In this way, data points can already be easily extracted from documents and used for further processing. Draft contracts from the other party can be compared with the company’s own standard in its own playbook in seconds.
Even more impressive is the assistance that AI can already provide today when drafting documents. For example, I can already use the AI functions of the PatternBuilderMax program from NetDocuments to create a customized draft based on a template or similar document from my sample collection if I only specify a few key points that I want to have reflected. A reasonably tech-savvy notary may probably already have 95% of their notarization preparation work done by AI in this way.
Efficiency gains will not benefit lawyers commercially
However, anyone who primarily emphasizes that AI makes lawyers’ work easier and more efficient has not heard the shot. Even at a superficial level, AI will be a bitter disaster for lawyers’ business models. Notaries may be able to pocket efficiency gains because they charge according to the (for the time being) compulsory statutory fee schedule. However, under the rule of the billable hour, lawyers do not commercially benefit from efficiency gains (if they honestly record and bill the work involved). Instead, such gains end up with their clients who are billed for less legal work.
It gets even worse. Things become really tense for lawyers if you ask the question differently and change your perspective. Instead of asking what AI means for lawyers, what does AI mean for their current clients? From their perspective, not only will legal advice become cheaper in view of the efficiency gains described above, but they will also ask themselves whether they still need a lawyer to answer legal questions and draft contracts at all. And the answer is clear. In view of the expected exponential increases in the performance of AI and assuming sufficient training under German law, this will no longer be the case for a large proportion of the work currently performed by law firms. This is already beginning to emerge with regard to the involvement of external lawyers by corporate legal departments. In the not-too distant future, contracts will only be drafted by external law firms in exceptional cases. The same applies to consumer law. Soon hardly any consumers will need to ask a lawyer the question: What is the legal situation? What should I do? Instead, people will ask AI. And get an answer that is at least as good, but certainly faster and cheaper.
Privileged treatment of lawyers in the Legal Services Act must be abolished
At best, one may argue about when this point will be reached. Will it be in one year, three years or five? That will to a large extent depend on how quickly the existing hurdles in Germany are removed. One such obstacle is the German Legal Services Act which prohibits non-lawyers from providing legal advice. Answering legal questions relating to individual cases using AI undoubtedly constitutes legal advice and would under the current legal rules be reserved for lawyers as a result. There will no longer be a reason for this privilege in the future if AI is superior to lawyers. Secondly, in order to be able to answer legal questions with sufficient certainty using AI in Germany, we need a sufficient data basis. This makes free online availability of court rulings essential. The days in which only a small percentage of court rulings are available online, as is the case today, and a large proportion of these are only accessible after overcoming access barriers such as beck online and Juris, must come to an end as quickly as possible.
Legal protection from hallucinations caused by AI?
What remains? Perhaps we lawyers will still be needed to check the plausibility of the results obtained by AI. Here too, however, skepticism is warranted. Presumably, we will only be called upon for a short transitional period to validate the results obtained using AI and to protect people from AI hallucinations. In view of the prospective capabilities of AI, the question will tend to be the other way round. It will then be like autonomous driving. In reality, humans are worse drivers than AI-driven cars without human drivers. In the foreseeable future, AI will not answer legal questions worse and in a more error-prone manner, but better and many times faster than most lawyers. It will then no longer be a question of whether the use of AI in legal advice constitutes professional negligence. Rather, we will find ourselves confronted with the claim that answering legal questions and preparing contractual documentation without the aid of AI is professionally negligent per se.
People may also continue to instruct us because they merely want to have a liable party in the event that something goes wrong. In the future, however, this will be more the domain of insurance service providers than of lawyers who will merely be middlemen in this regard.
Little consolation from remaining fields of activity
Of course, there will still be areas in which the involvement of lawyers remains necessary, either as a legal requirement or out of factual necessity. For example, notaries will probably succeed in defending the existing notarization requirements for a few more years without any material justification. The obligation to be represented by a lawyer in court may also continue to exist. The time will soon come, however, when the question arises as to whether and to what extent court proceedings can be replaced by AI-driven dispute resolution procedures. This is likely to be the case with the mass litigation proceedings that are currently felt to be overburdening the courts.
That leaves situations in which classic human skills are required, such as empathy and negotiating skills. However, this is not much consolation for lawyers. In reality, these skillsets often have little to do with the application of the law. In any case, none of these are learnt in law school. And it will never be possible to increase the hourly rate for these remaining activities to such an extent that they can even remotely economically replace the activities that are no longer required.
What can be done? We may regret all these developments, but we will not be able to stop technological progress. We have to say goodbye to our old world and get used to a new, different role instead. We need to discuss what this role might be in the year that has just begun. In this respect, we don’t have much time for consolidation.
Author
Dr. Matthias Birkholz, LL.M.
lindenpartners, Berlin
Lawyer, Partner
