Translating documents overnight, studying files for hours on end, time-consuming research into contracts – many lawyers have already had to deal with unpleasantly tedious tasks in the course of their career. At least some of these tasks can be increasingly automated. Legal tech applications are not a new topic for lawyers, but since the OpenAI bot ChatGPT began its triumphal march just over a year ago, applications with artificial intelligence (AI) have received a further boost.

"The offerings are surprisingly powerful when it comes to summarising legal texts and documents and explaining them in a structured way," says Tom Braegelmann. The lawyer from the law firm Annerton has been involved with the use of technology in law firms for many years and has himself worked for a legal tech company for a time. A hundred-page decision by the Federal Network Agency, for example, can be summarised by ChatGPT in a very short time – with one major caveat: "You have to check all the results," emphasises Braegelmann. "However, familiarising yourself with the facts of a case is quicker if you already have a summary."

The business law firm Allen & Overy uses several AI tools, including Harvey, which has been specifically trained for the legal context. Harvey, like the latest version of ChatGPT, is based on OpenAI's GPT-4. "You can ask Harvey legal questions such as 'What is the legal nature of banking secrecy in German law?' and have it provide sources to verify the result, similar to the traditional process without the use of AI," explains Associate Niklas Germayer. Feeding the AI with existing documents is also becoming more effective. In this process, search results are not created based on the AI model in connection with the internet but primarily based on the information in the document itself. The tool is significantly more powerful in using GPT-4 compared to its predecessors.

However, Germayer also believes that careful monitoring is essential. For Alexander Behrens, Partner at Allen & Overy, the rapid developments in AI are a double-edged sword: "On the one hand, the applications are a good help. But of course, some lawyers are worried that the technology will one day completely replace legal work." But that is a long way off. The lawyers are convinced that even powerful AI models cannot replace legal work at the present time. Rather, it is about supporting and automating highly standardised activities.

Eric Wagner, Partner at Gleiss Lutz, sees the broader range of applications as the greatest advance of the latest generation of AI tools. "Until now, many legal tech offerings have been heavily focussed on individual areas of application such as text recognition or data analysis," he says. The new generation can fulfil different tasks without the lawyer having to switch between different applications. "This increases acceptance within the team and leads to more routine handling – and therefore better results," says Wagner.

Marc Geiger heads the Legal Operations & Business Technologies department at the law firm. Around 80 employees work there, providing training on AI and legal tech, for example, but also developing their own digital solutions and keeping an eye on cyber security. For Geiger, the latest generation of applications with generative AI is a kind of universal platform: "Lawyers use them like an assistance system. The interest and potential applications are huge."

AI could change not only the way lawyers work but also the way lawyers are paid. Until now, the hourly rate has been the benchmark for payment for legal services. However, AI is particularly good at preparing lengthy routine tasks, such as translations. This reduces the billable hours. "In the future, the technology could lead to law firms increasingly switching to fixed prices," expects Tom Braegelmann.

AI tools have not yet reached the point where it is possible to realise significant efficiency gains, says Eric Wagner from Gleiss Lutz. However, the better the systems become, the greater the pressure for change in the market. On the one hand, law firms have high investment costs to acquire the technologies. But how much is a client prepared to pay if at least some of the pleadings have been created by an algorithm? "There is currently an imbalance," admits Wagner.

However, he believes that some mandates could probably no longer be won without investing in new technologies: "Many clients specifically ask in pitches which legal tech and AI tools the law firm uses." Wagner would not rule out that the price structure will shift more towards fixed prices that include the use of AI tools for certain use cases. "We may even get to the point where clients accept a certain error rate in the interest of increased efficiency."

It will not work without lawyers in the future – however, the areas where they can provide added value are shifting. For instance, linguistic clarity may become less important as a differentiation criterion. This is because an AI can handle the linguistically appealing presentation of facts. "The performance will lie in using a suitable AI tool, for example, for repetitive tasks to work more efficiently and create space in other areas," expects Wagner. AI ultimately remains to be a technology. "AI can't do law. It can interpret data well," emphasizes his colleague Marc Geiger.

However, if an issue cannot be broken down into data, AI quickly reaches its limits: "Some issues have to be assessed without a database. This then requires empirical knowledge," says Braegelmann. "Lawyers are experienced in dealing with authorities, they know the local labour, insolvency and other courts, and they can mediate in disputes between shareholders. Clients will continue to be willing to pay for this added value," the lawyer is certain.

Despite all the interest in the new possibilities – lawyers can't just try it out like private users do with ChatGPT. Data protection and legal professional privilege place strict limits on the use of AI tools. At Allen & Overy, AI tools are programmed in such a way that they do not store any data; the principle is called "zero data retention". "The tool does not learn from the data," explains Partner Behrens.

Today's AI models are no longer primarily concerned with large volumes of data, explains Marc Geiger from Gleiss Lutz. This is due to the way they work: Early approaches in machine learning used numerical tables to represent words. The tools could not map relationships between words – for example, they did not recognise when two words had a similar meaning. In large language models (LLM), which are used today, such relationships are represented by word embeddings using vectors. Put simply, words that have similar meanings are closer together in the vector space than others. "With machine learning, the principle was: the more data, the better the algorithm," says Geiger. "With LLM tools, on the other hand, you can also improve the quality of the results by weighting certain vectors differently. The amount of data is less decisive." Many projects are currently underway at large law firms, in particular, to test use cases for artificial intelligence. Hogan Lovells, for example, has just developed the chatbot "Eltemate Craig" with its legal tech subsidiary Eltemate. It should be able to draft legal texts and summarise large amounts of information in a structured manner. KPMG Law is also working with its own AI-supported chatbot.

However, the world of justice is not regarded as a digital pioneer; hardly any other sector still uses the fax machine so intensively. In order to establish AI on a broad scale, lawyers need to utilise the technology. "There have been a few cases in recent years where law firms have euphorically invested in legal tech tools that were then left untouched by employees," reports Braegelmann, who also offers training courses on the use of AI for lawyers. This reluctance was often due to uncertainty or fear of embracing something new. However, nobody needs to worry about being exposed by artificial intelligence: "The algorithm can do no more than the lawyer. But it can help lawyers save time."

At Allen & Overy, 800 to 900 of the 3,000 lawyers currently use an AI application every day. "That's a pretty good rate," says Behrens. A simple trick has driven the usage figures up significantly: "The first AI tools ran in a separate browser and received far less attention," recalls Behrens. In the meantime, the law firm has developed its own AI application called "Contract Matrix" and integrated it directly into Word.

Access to the contract negotiation tool, which the law firm also provides to clients, has significantly increased since then, and the tool can be more efficiently utilized in transactions. For Behrens, this is an important realization: "It's not about bringing the user to the technology. It is more promising when the technology comes to the user."