AI Is Changing Law, Not Eliminating Lawyers

Artificial intelligence in the legal industry is no longer an experimental add-on but a tool that is already part of everyday legal work. It can draft letters and contracts, support strategy development, and take over many small and repetitive tasks that until recently required substantially more time. “Apart from IT, it is difficult to find an industry where such a tectonic shift has already happened over the past three or four years,” Marko Porobija, Managing Partner at Porobija & Špoljarić, tells ICTbusiness Media - ICTbusiness.info. The change is particularly relevant to law because generative models, unlike traditional software, can work with probability, nuance and context that fit legal reasoning better than simple yes-or-no logic.

Porobija confirms that AI can genuinely produce a legal letter or a contract. “It can write a letter, it can write an entire contract, it can devise a strategy, and it can act as an agent for a hundred small things you do,” he says. The crucial question, however, is not whether the system can generate convincing text, but whether that output is good enough for the specific legal situation and jurisdiction.

This is where legal expertise remains essential. Clients increasingly arrive with a document already produced by AI and ask for a quick review. Only an expert review may reveal that a professional-looking text contains clauses that are inappropriate for the transaction, market or court procedure. “AI does a lot of repetitive work for us, but we are the quality control,” Porobija says. The lawyer decides whether the document can actually be sent to a client, court or counterparty and whether it serves the purpose for which it was created.

The difficulty is that a non-lawyer often cannot recognise the error. Porobija compares it with an apartment design that looks excellent at first glance while an architect immediately notices a structural problem. “Those are details that a layperson has no idea are happening,” he says. In law, such mistakes may remain hidden for years because many contracts are never tested through a dispute. Once a dispute occurs, individual words, punctuation and subtle distinctions can become important.

His firm therefore uses AI extensively but cautiously. “We use it a great deal; we work hard to know how to use it, but we are very careful about what is confidential,” he says. The rapid growth of legal-tech tools shows how strongly the industry is looking for technology that can accelerate research, document production and the processing of large volumes of material.

Data protection is the most sensitive issue. Legal documents often contain personal data, trade secrets, transaction details, disputes and strategies that must not leave a controlled environment. Porobija says they use the highest levels of commercial subscriptions that promise protection and no training on customer content, but they do not rely solely on contractual assurances from technology providers.

“We anonymise absolutely every document,” he stresses. Removing names or identification numbers is not enough. Contextual details that could reveal the identity of a person or company also have to be removed. If a document describes an organisation so precisely that it can be identified without being named, that information also has to be stripped out before processing.

Even stricter rules apply to privileged and particularly sensitive information. “Privileged information has no chance of going in,” Porobija says. In those situations, a person remains a mandatory control point. A closed AI system running entirely inside an organisation’s own infrastructure and never sending data outside could change part of that relationship. Still, he expects truly practical and broadly accessible versions of such solutions to take more time.

The speed of development makes long-term forecasting difficult. Porobija notes that not long ago it would have seemed impossible that AI could support such a large share of legal work, yet that is already happening. “If someone had told me eleven months ago that I would be able to do around 70 percent of legal work with the help of AI, I would have said there was no chance,” he says. That is why he avoids categorical predictions about where the technology will be a year from now.

He nevertheless does not expect the human role to disappear. “Without a human in the loop, it will not work,” he says. A lawyer provides more than a formally plausible answer: the job involves context, interests, risk, the behaviour of the other side and non-verbal signals that appear in conversations and negotiations. Current systems cannot reliably replace those elements.

The lawyer’s role is therefore changing, not disappearing. “We will still be here, deciding, checking quality and looking at what clients actually want, while AI can do its part,” Porobija concludes. The biggest change may be in the division of labour: machines will handle more preparation and routine work, while human value will increasingly lie in judgement, responsibility, context and quality control.